Home

1986 , Volume , Issue June-1986

1. Al Tykulsky Development Engineer Andover Division a TT ET Fig 2 To extend the volume a the sampler oper ates a number of times b at intervals of T for each sampling period The results are accumulated JUNE 1986 HEWLETT PACKARD JOURNAL 39 Copr 1949 1998 Hewlett Packard Co signal In HP s system this is performed by FFT al gorithm which requires the DDC to digitize the Doppler signal At first glance it would seem reasonable to digitize the forward reverse flow information coming out of the rotational filter and perform separate Fourier transforms on each signal However the rotational filter has nominal phase errors of up to two degrees which will induce some mirroring on the spectral display although it has little effect on the degree of stereo separation in the audio output Therefore it was decided to digitize the Doppler data before separating the forward and reverse components and to allow a single complex FFT to perform this requirement Two eight bit SAR type analog to digital converters ADCs digitize the in phase and quadrature outputs of the multi pliers and send the data to the FFT card Flow Direction Decoding Each of the in phase and quadrature signals contains the Doppler information for both forward and reverse flow Each by itself is incapable of differentiating the direction of flow but considered together the two sig
2. 4 b Copr 1949 1998 Hewlett Packard Co internal structures but also tiny echoes scattered from mov ing blood The primary scattering site in the blood that produces echoes is the red blood cell The size of the echoes from red blood cells is related to the scattering cross section of the cell which is on the order of four micrometers The scattering cross section of a platelet is 1000 times smaller and does not contribute significantly to the Doppler shifted energy Leukocytes are not present in sufficient numbers to influence the backscattered signal from the red blood cells The incident ultrasonic pulse extends over many red blood cells and the reflected signal received at any given range is the sum over the effective resolution cell of the complex vectors produced by each red blood cell There are five million red blood cells in a cubic millimeter of blood give or take a few and these phasors do not arrive at the transducer in phase although they are coherent be cause they vary in unison This coherent integral is analo gous to the same phenomenon that results in the speckled appearance of laser light images see reference 1 for more details about ultrasonic speckle If one assumes that the position of a given red blood cell is a random variable with a uniform distribution the amplitude of the backscattered signal will also be a random variable distributed according to a Rayleigh distribution
3. HEWLETT PACKARD JO N gt U JUNE 1986 my ced ceed ced cee hanna 5 id VAPA WO PI VA PA MIO VODA A OI BET uhmm EE iiit ai pay oe oe 1 2 tail ae aes AA Copr 1949 1998 Hewlett Packard Co HEWLETT PACKARD OWI Se Articles Integrated Circuit Procedural Language by Jeffrey A Lewis Andrew A Berlin Allan J Kuchinsky and Paul Yip This in house Lisp based VLSI design tool accommodates different fabrication processes and functional cells with variable characteristics 8 Knowledge Assisted Design and the Area Estimation Assistant 1 1 Software Development for Just in Time Manufacturing Planning and Control by Raj K Bhargava Teri L Lombardi Alvina Y Nishimoto and Robert A Passell This software product provides streamlined management information for the just in time environment 13 Comparing Manufacturing Methods 1 8 Authors 2 0 The Role of Doppler Ultrasound Cardiac Diagnosis by Raymond O Connell Jr Data about blood flow anomalies can be obtained by observing the shift in frequency of ultrasonic imaging pulse echoes 2 Doppler Effect History and Theory by Paul The shift in the frequency of sound for movin
4. Doppler lines have different timing requirements from image lines If the Doppler gate is set for a shallow location the Doppler line will be shorter than the image line Dop pler lines do not require the timer to initiate new mixer focal zones for dynamic focusing Duplex Doppler mode involves alternating between Doppler and image lines Since reprogramming the timer after each line would be prohibitive a dual timer was designed The individual counters that generated timing signals have been removed and replaced with a large RAM A master counter sequen tially reads the contents of each RAM location The RAM outputs are used as timing signals Appropriately program ming the RAM makes timing signals appear when needed The RAM contains two pages one for image lines and one for Doppler The page address can be toggled easily and quickly between lines Processor Enhancements A number of other hardware changes supported the vamping of the software While scanning the main task of the MC 5 microprocessor is to assemble control bits and a line number for the next acoustic line to be fired The processor must also monitor the HP 77020A Imaging Sys tem s internal bus based on the for new commands and take care of housekeeping background tasks This over head was previously performed between sector frames Synchronous Clock Bus Doppler mode requires the highest possible line rates to detect high blood flow velocitie
5. Looking at the signal in the frequency domain the pro Magnitude PRF 1 N 1 PRE Baseband IF Fig 3 Sampling the IF signal translates and sums each of the harmonics of the PRF and their respective spectrums down to baseband frequencies 36 HEWLETT PACKARD JOURNAL JUNE 1986 cess of ideal sampling will replicate and translate the orig inal spectrum by integer multiples of the sample rate i e the PRF Since one is only interested in the baseband frequencies dc to the PRF in pulse Doppler mode the process of sampling can be restated as the translation and summing of each of the harmonics of the PRF and their immediate spectrums down to baseband This is shown in Fig 3 In ideal sampling however the frequency spectrum is periodic with a period of PRF Hz see Fig 4 This implies that the output of an ideal sampler has infinite power and energy which is realistically impossible In the samplers on the DDC a zero order hold is implemented Once a sample is taken the value of that sample is held for the duration of the PRI In the frequency domain this is mathematically equivalent to multiplying the spectral out put of the ideal sampler by that of a sinc function A sinc function is defined as sinc x PRF sin 1 From Fig 4 it can also be seen that the spectrum is mirrored about the frequency PRF 2 this is referred to as the Nyquist rate For a singl
6. An HPJIT screen for enter ing select criteria for an MRP 14 HEWLETT PACKARD JOURNAL JUNE 1986 terial requirements planning run Copr 1949 1998 Hewlett Packard Co HP divisions that practice JIT so that we could quickly develop a product that would suit HP s internal needs as well as those of the marketplace To meet both internal and external needs several impor tant requirements must be satisifed There must be both Management commitment and individual commitment throughout the project teams at divisional sites participat ing in the development process More than one external division must be involved and each must dedicate some one full time to the project In practice the partner division teams communicated among themselves so effectively that many disagreements about product requirements were re solved without intervention by the development team in the software lab The partner divisions played an important role in all phases of the project During the design phase increased productivity in the software development lab resulted from their review of the prototypes and specifications eliminat ing a significant amount of guesswork on the part of the design engineers when adding new product features The partner divisions helped eliminate guesswork during the final coding phase by testing product quality and v
7. of utility procedures has many significant advantages Eliminating code duplication not only reduces overall code length but also reduces source code clutter in individual modules thereby helping to preserve transaction clarity Relying on many shared utility procedures eliminates the risk of errors caused by oversight and accident when rep licating source code throughout many individual modules Code maintenance and product enhancement are made simpler since a correction or enhancement can often be made ina utility module and effected immediately through out the entire application by replacement of the utility module alone Some utilities screen handling and data base operations can be used by similiar software sys tems saving time and effort when developing and testing other products Reliability is increased since a utility called by several different individual transactions is excer cised thoroughly during testing Some typical HP JIT utilities are compare_lines Onascreen with multiple data entry lines determine that the user hasn t entered any duplicates Inform the user about the success or failure of finding or not findinga requested data item dataset msg Post Deduct a Parent Part EC Level Used Deduct Quantity Point 10 Deducted UM Description 111 Vep POST DEDUCT Fig 4 The HP JIT post deduct transaction records production completed and simultaneously re lieves stock areas of inven
8. this technique can be used to measure sector imaging mode and M mode output power Because this method really measures the radiation pressure from an acoustic wave the M mode beam should be aimed straight at the target and the sector mode result must be corrected for the fact that for a portion of the scanning the acoustic beam is hitting the target obliquely Intensity Measurement with a Hydrophone Intensity is a measure of power concentration at a spot Since it has a direct relationship with tissue damaging mechanisms such as heating and cavitation this parameter has been of great concern to health physicists Intensity measurement in medical ultrasound is complicated be Target Fluid Transducer Weight Fig 1 Measuring radiation pressure with a force balance 32 HEWLETT PACKARD JOURNAL JUNE 1986 cause it is generally not constant throughout a time period and is not uniform spatially Thus ultrasonic intensities must be further defined The three commonly used inten sities are Spatial peak temporal average intensity SPTA Spatial peak pulse average intensity SPPA Spatial average temporal average intensity SATA To measure these intensities a passive listening device is positioned in the acoustic field to measure the pressure at various locations This listening device or hydrophone transforms the pressure waves received into voltage waveforms which can be displayed on a CRT When the hydrophone is cal
9. 2 coauthor of two articles on the simulation of VLSI design and is interested in design automation sili compilation and expert systems Paul was born in Canton China and grew up in Hong Kong A resident of Sunnyvale California he likes jog ging skiing and basketball With HP since 1985 Andy Berlin is an R amp D engineer at HP s Cupertino Divi sion He has contributed to the development of ICPL and has taught an HP soft ware engineering class He was born in New York City and holds a 1985 SBE de gree from the Massachu setts Institute of Technology Andy lives in Cuper tino California and enjoys playing piano Jeff Lewis was born in Oak land California and studied economics and computer science at the University of California at Berkeley pleting his studies for both the AB and BS degrees in 1983 While stil student he worked for IBM asa cir cuit designer Atter coming to HP in 1983 he contributed to the development of an IC design tool and later became the technical leader for the ICPL project He is a coauthor of one other article on ICPL ands interested in pro cedural layout and Lisp programming Jeff and his wife live in Mountain View California His outside Interests include running bicycling scuba diving and skiing 11 JIT Alvina Y Nishimoto G j Alvina Nishimoto was born in Honolulu Hawaii and specialized in computer utilizati
10. 4 The reason is that the structure causing the reflected echo is acting as both the listener and the source Since the structure is in motion the transmitted frequency it perceives is Doppler shifted according to equ ation 2 The echo that leaves the structure is shifted again since it now acts like a source which is in motion relative to the stationary receiver according to equation 4 Since only the component of the velocity directed toward or away from the transducer contributes to the Doppler shift fre quency one must scale the frequency shift by the cosine of the angle between the structure s velocity vector and the line connecting the structure and the transducer The Dop pler shift can then be expressed fa 2VEJ C 6 for V lt lt C 8 where V is the structure s velocity 6 is the angle between the ultrasonic beam and the structure s velocity vector f is the frequency of the incident ultrasonic pulse and C is the velocity of sound in the body Acontinuous wave blood velocity measuring instrument typically extracts the Doppler shift frequency by mixing the received signal with the local oscillator that produced the transmitted signal After filtering out the sum term one is left with the frequency difference between the transmit ted signal and the received signal which is the Doppler shift frequency It turns out that for velocities present in the body and for the frequencies typically used the Doppler shift fr
11. Although the phase and mag nitude of the Doppler echoes are random variables they do change in a reasonably orderly fashion In the case where the red blood cells are moving through the sample volume at a constant velocity toward the transducer the frequency of the echoes will increase most of the time CW Doppler Measurements The first reported application of the Doppler effect to medical ultrasound took the form of a continuous wave CW Doppler velocity measurement These systems typ ically use a transducer made with two piezoelectric crystals for transmitting and receiving the ultrasonic wave The transmitted wave is sinusoidal at a frequency usually be tween 2 MHz and 10 MHz Attenuation of ultrasound in the body tends to increase linearly with frequency which dictates lower frequencies to obtain greater penetration Higher frequencies on the other hand provide narrower beams for a given transducer size When the transmitted pulse enters the body reflections occur at the tissue inter faces in the path of the beam If the interfaces are stationary the reflected signal returns to the receiver at the same fre quency as the transmitted signal albeit attenuated by the intervening tissue However if the echo returns from struc tures such as a heart wall or a group of red blood cells that are moving the returning signal receives a Doppler shift What is interesting is that the Doppler shift is twice that predicted by equation 2 or
12. Hatle Doppler Ultrasound in Mitral Regurgitation and Aortic Stenosis ractical Cardiology Vol 9 10 September 1983 pp 73 84 J Rawles M D and Haites Doppler Ultrasound Mea surement of Cardiac Output Practical Cardiology Vol 9 no 8 July 1983 pp 143 156 9 L H Light and Cross Convenient Monitoring of Cardiac Output and Global Left Ventricular Function by Transcutaneous Aortovelography An Effective Alternative to Cardiac Output Measurements Cardiac Doppler Diagnosis edited by M P Spencer M D Martinus Nijhoff Publishers Boston 1983 pp 69 80 10 U Elkayam M D et al The Use of Doppler Flow Velocity Measurement to Assess the Hemodynamic Response to Vas odilators in Patients with Heart Failure Circulation Vol 67 no 2 1983 pp 377 383 11 DJ Sahn M D and L M Valdes Cruz M D Two Dimen sional Echo Doppler For Non Invasive Quantification of Cardiac Flow A Status Report Modern Concepts of Cardiovascular Dis ease Vol 51 no 10 October 1982 pp 123 128 12 J Giroud A S Pickoff and P L Ferrer Pulsed Doppler Echocardiographic Quantification of Left To Right Shunts in Chil dren With Isolated Ventricular Septal Defects Cardiac Doppler Diagnosis edited by M P Spencer M D Martinus Nijhoff Pub lishers Boston 1983 pp 227 234 JUNE 1986 HEWLETT PACKARD JOURNAL 25 Copr 1949 1998 Hewlett Packard Co Doppler E
13. Systole Diastole Systole Diastole Systole Diastole HHHH Diastolic Flow Left Atrium LA tral Valve MV mits Left Ventricular Inflow Tract LVIT Systolic Flow Lett Ventricular Outflow Tract LVOT Aortic Valve AOV Aorta AQ b atrium through the tricuspid valve to the right ventricle during dia stole Systolic flow occurs when bloodis ejected from the right ven tricle through the pulmonic valve LA to the main pulmonary artery b The left heart Freshly oxygenated LVIT blood from the lungs returns to the heart via the left atrium Flow then proceeds from the left atrium through the mitral valve to the left ventricle during diastole Systolic flow occurs when the left ventricle pumps blood through the aortic AOV valve back into the circulation LvoT system JUNE 1986 HEWLETT PACKARD JOURNAL 21 Copr 1949 1998 Hewlett Packard Co Clinical Applications Doppler echocardiographic systems are of little use un less they can determine the necessity for surgic tion drug treatment or other therapy The following is a al interven list of several disease states in which Doppler measure ments fulfill a critical diagnostic role Valvular stenosis and regurgitation o Mitral Aortic Tricuspid o Pulmonic Prosthetic valve function Congenital heart disease Ventricular septal defects VSD 1 Atrial septal defects ASD Patent ductus arteriosus PDA 1 Coarctation of the aorta Va
14. The use of ultrasound in the diagnosis of coronary artery disease has had litfle success except in pediatrics Coronary angiography still provides the critical information for assessment of coronary function Technical Limitations The use of the Doppler effect in ultrasound measurement of blood flow has some limitations There are two important aspects to the Doppler equation see article page 26 that must be considered where evaluation of cardiac disease is concerned The first is the angle between the flow velocity of interest and the incident ultrasound beam The most accurate velocities are measured when the angle is very small However when searching for certain cardiac anoma lies such as high velocity jets caused by stenotic regurgi tant or shunt lesions or defects in the heart the exact angle of flow is unknown and movement or rotation of the trans ducer is necessary until the location of the highest maxi mum velocity is obtained The other important aspect of the equation is the propor tional relationship between the frequency used to interro gate the blood flow and the resultant frequency shift In pulsed Doppler systems the maximum measurable shift is limited by the rate of the ultrasound pulses sent out For example assume that a pulsed Doppler system is sampling flow from a vessel or heart chamber at a depth of 12 cm Further assume that the speed of sound in body tissue is approximately 1540 m s and the frequency o
15. a program that describes how to build the layout and connec tivity of a part rather than drawing it graphically This allows a user s cell definitions to be compiled parsed by the Lisp reader and have embedded Lisp constructs like T D impian ittusion in ana p conse Fig 1 As process technology changes the placement and size of the rectangular elements form ing circuit element must be al tered to achieve similar perfor mance a NMOS C cell b Equi valent NMOS lll cell Similar CMOS H cell All cells were built using the same ICPL code and dif ferent fabrication process libraries any other Lisp dialect Simply being a language is not ICPL s strongest point Through the use of the symbolic programming capabilities of ICPL a designer can specify far more than a simple geometric layout of a circuit Cell libraries written in ICPL have a significant advantage over traditional graphically created libraries in that an ICPL cell can take parameters like any other procedure to alter its functionality Forexam ple a buffer circuit can take as a parameter the capacitive load it must be capable of driving and the speed at which it must run and then size itself to guarantee proper oper ation for a particular application Parameters can also be used to create many variations on a particular design For example a single register cell is capable of incrementally a
16. a knowledge assisted design KAD system We use the term knowledge assisted to represent the synergy of traditional CAD tools algorithmic knowledge and artificial intelligence Al software technology Primary motiva tions for such a system include Supplementing not replacing the designer Not all designers who wish to take advantage of VLSI capabilities can do so using traditional CAD tools and methodologies This is be cause of the vast amounts of knowledge needed to use these tools to perform VLSI design In addition to the VLSI specific knowledge the amount of knowledge needed just to apply the right tool to the right problem is considerable An environ ment that eases the use of new tools for the designer would be quite useful Forming the communication medium in which all parties in volved can share and retain information and constraints The VLSI design process often requires coordination and com munication among several different people The system de signer communicates specifications to the IC designer who in turn communicates with other designers production en gineers and mask designers to achieve chip layout and pack aging Information lost in the transition between designers can result in increased design difficulties Current design systems are lacking in the ability to capture much of this information Preventing the loss of expertise frequent occurrence is the departure or promotion of designers The exp
17. are inefficient electrical to mechanical energy transformers Thus the hydrophone calibration is normally done at intensity levels several or ders of magnitude lower than the levels put out by medical ultrasound equipment Therefore the linearity of the hy drophone and other measurement equipment must be calibrated to guarantee the accuracy of the intensity mea surement This is normally carried out by comparing the hydro phone received signal with the corresponding round trip signals when the driving transducer is driven at different intensity levels The hydrophone is positioned at the peak pressure amplitude position beyond the end of the near field The round trip measurement is made when a special target is positioned at the same place The intensity range of this calibration covers the intensity level where the hy drophone sensitivity calibration is done and the level where intensity measurements are performed Another method usable for this calibration is the com parison of signals received by the hydrophone against the driving signal throughout the intensity range In both methods it is critical to account for any nonlinear effects associated with water Water is a medium that starts to generate nonlinear effects at rather low intensity levels about 10 W cm The way to carry out this calibration with out being too sensitive to nonlinear effects is to compare the temporal average intensity instead of peak intensity This is
18. creates continued on page 9 REAO WRITE A RERO WRITE REFRESH READ REAO WRITE REFRESH FEEDTHROUGH A READ WRITE REFRESH READ WRITE A RERO WRITE NO REFRESH READ WRITE REFRESH READ WRITE READ WRITE REFRESH POWER WIOTH Fig 3 Configurable CMOS register cell Some of the fifty possible configurations with two buses are shown JUNE 1986 HEWLETT PACKARD JOURNAL 7 Copr 1949 1998 Hewlett Packard Co Knowledge Assisted Design and the Area Estimation Assistant VLSI complexity has increased to the point that it is not unusual to spend several engineer years doing a state of the art IC chip Software development has faced similar problems The software engineering problem was handled by raising the level of abstrac tion at which the programmer operates Aspects of the solution are the use of higher level languages for software developers and new programming methodologies and tools to handle the organizational complexity of large programming projects VLSI design complexity requires a similar solution One preliminary step is the use of module generators which allow the designer to build large IC parts such as random access memories without needing to pay attention to the detailed artwork implementation Knowledge Assisted Design Framework At HP s Cupertino Integrated Circuits Division efforts are under way towards the realization of
19. incorporate customer feedback team to lose their focus Prototyping also helped reduce the risk in the early stages JUNE 1986 HEWLETT PACKARD JOURNAL 13 Copr 1949 1998 Hewlett Packard Co of product engineering by unifving the development team s own understanding of the product making the features and functionality of the system more tangible This helped iden tify potential problems in system design early and mini mized costly engineering redesign in the later phases of the project Our strategy for implementing prototypes was to produce a prototype in three to four weeks We felt prototypes should be done quickly using whatever tools were avail able Their purpose was to allow us to iterate the solutions to vital questions The prototypes were not meant to be an early stage of the coding phase since the prototypes were to be thrown away after the features or problems they addressed were understood No attempt was ever made to use any actual prototype implementation later since we felt that this would inhibit the project team from solving the problems quickly If an actual prototype were to be used later as the basis for some effort in the coding stage of the project we felt that we would quickly become overly concerned with engineering the perfect prototype during the investiga tion and lose our focus on asking vital questions and solving problems that the prototypes were supposed to address In the JIT project the
20. is still significantly lower than that of the heart wall signal to noise ratio To detect the weak Doppler sig nals the system and transducer must be designed to be sensitive enough to detect low level signals and simultane ously handle the high level signals from slowly moving interfaces such as the heart wall Therefore when attempt ing to extract Doppler data it is sometimes necessary to drive the transducer harder than when extracting normal ultrasound imaging information However this translates to a higher acoustic energy into the patient s body which must be carefully controlled to cause no adverse effects Thus it was clear early in our system design that not only does the system have to be designed to have exit power and intensity below the generally accepted limits but also its output has to be measured carefully to ensure that they are within those limits All of HP s acoustic output measurement procedures conform to the definitions and methods suggested by guidelines set by the American Institute of Ultrasound in Medicine and the National Electronic Manufacturer s Associ ation This ensures that we meet labeling requirements and DOPPLER SIGNALS scattered that we use the same terms and values as government reg ulatory agencies and other ultrasound medical equipment manufacturers However HP procedures do not follow the regulations literally because features particular to our sys tem make literally following
21. of peripheral vascular disease carotid arteries veins etc and for the study of fetal heart rate Hewlett Packard s first involve ment in Doppler techniques was the HP 8021A Cardiotoco graph introduced in 1971 In the case of carotid artery examinations the systems were designed to map the blood flows so that a two dimensional presentation could be made that was very close to those obtained through X ray techniques Pulsed Doppler technology followed It was used in two fashions First it allowed the user to separate velocity in formation from several vessels in close proximity as in the neck Second it allowed the use of multiple timing gates which gave better information about the distribution of velocities within a vessel Many of these systems employed a spectral analysis technique called a time interval histo gram for real time analysis or processed the data off line with a software fast Fourier transform FFT program Two dimensional imaging systems were introduced which al lowed placement of a pulsed Doppler sample volume over a wide area This equipment was designed for cardiac work and used the time interval histogram spectral analysis ap proach because of its low cost and speed Although articles were published on cardiac studies based on the use of pulsed Doppler techniques with this equipment clinicians were slow to adopt Doppler ultra sound as an accepted aid in the diagnosis of cardiovascular disease because
22. of the limitations of the technique the Fig 1 Four transducer positions are used for obtaining Dop pler information using CW probe a The suprasternal notch for ascending and descending aortic flow and pulmonary artery flow b The left parasternal area for right ventricular inflow and outflow and pulmonary artery flow c The right parasternal area with the patient rotated in right lateral decubitus position for ascending aortic flow d The cardiac apex for left ventricular inflow and outflow ascending aortic flow and right ventricular inflow Copr 1949 1998 Hewlett Packard Co limited number of applications and its relative newness Recent Developments Within the past two years cardiac Doppler ultrasound technology has been recognized as an important tool in evaluation of cardiac blood flow rates Although it has been available for many years it has not been considered as having clinical utility until the work of Hatle and Holen which demonstrated that valve pressure gradients could be quantified using Doppler ultrasound techniques In 1982 radionuclide and contrast angiography were pre ferred tools for the diagnosis of global ventricular function identification of regurgitant valvular cardiac lesions iden tification of intracardiac shunts and assessment of coro nary artery disease Today ultrasound techniques are ex tremely successful in aiding the diagnosis of many cardiac abnormalities
23. the scattering site The spectrum produced based on this win dowed waveform and a single velocity becomes a band of Doppler shift frequencies Furthermore the faster the par ticle travels the shorter the time window in the sample volume and the wider the band of frequencies produced The second effect of this Doppler shift frequency spread is that aliasing will begin to occur at a frequency slightly different from the Nyquist rate which is just the Doppler pulse repetition frequency divided by 2 Note what is in volved here aliasing from the higher clutter harmonics occurs before the nominal Nyquist frequency 30 HEWLETT PACKARD JOURNAL JUNE 1986 There is a second mechanism responsible for anomalous spectral spreading which is caused by the statistical nature of the signal scattered from red blood cells Since the amplitude of the echo varies with the specific orientation of each red blood cell and since new red blood cells enter the sample volume and others leave the sample volume during the time required to obtain the samples of the Dop pler shift waveform the echo amplitude is modulated in dependently of the Doppler shift This amplitude modula tion creates sidebands in the received signal which overlap the Doppler shift sidebands This has the effect of creating spurious Doppler shift information which tends to increase the apparent bandwidth of the Doppler spectrum In spite of the theoretical limitations ultrasonic Doppler ve
24. the acoustic field for a location that gives maximum voltage amplitude HP s Doppler system varies its repetition rate as well its driving voltage to meet the optimum Doppler measure ment requirement It also allows users to vary the repetition rate and driving voltage to meet their individual needs This adds to the complexity of the search for the maximum intensity How we manage to determine the location of maximum intensity depends on what type of intensity is to be mea sured We will limit our discussion to spatial peak temporal average intensity as an example of how intensities are mea sured The spatial peak temporal average intensity Ispra is defined as follows 1 st v t Kz 8 Ispra where T the pulse repetition period To measure Ispra requires the recording of a transient waveform v t and integration as shown in equation 6 The transient can be captured by a commercial transient re corder or can be captured by photographing its trace on a CRT screen and digitizing the photograph Then software can be used to carry out the integration Either way is a lengthy procedure and cannot easily be done in real time In the peak intensity search phase it is very desirable to have a method that will read out Ispra instantaneously A special instrument setup was put together for this purpose An HP 3400 RMS Voltmeter gives an instantaneous reading of rms voltage which can be directly related to Ispra follow ing
25. the product was enhanced by the use of rapidly developed prototypes as a basis for communication about product function and design The establishment of frequent mile stones each with a deliverable part of the product from prototype to completed module aided in sustaining the momentum of the project team and kept the effort success fully on track from start to finish Acknowledgments Nancy Federman was the R amp D project manager for HP JIT until the first release of the product We would like to thank her for her leadership and for developing some of the figures used in this article We would also like to thank product team members Steve Baker Marc Barman Jim Heeger Pamela Hinz Kristine Johnson Mike Kosolcha roen Mary Ann Poulos and Chris Witzel for their contribu tions in developing HP JIT Our thanks also to Computer Systems Division Cupertino and Roseville Disc Memory Division Boise San Diego Division San Diego and Van couver Division Vancouver for their active participation and feedback continued on next page JUNE 1986 HEWLETT PACKARD JOURNAL 17 Copr 1949 1998 Hewlett Packard Co References 1 Federman and R Steiner An Interactive Material Planning and Control System for Manufacturing Companies Hewlett Packard Journal Vol 32 no 4 April 1981 Novel Approach to Computer Application 2 L Winston System Design and Implementation ibid 3 D Kur
26. to the point of causing over flows Only in lower variance signals is the gain allowed to increase if the overflows are below the selected threshold Reference 1 L Schirm IV Get to Know the FFT and Take Advantage of Speedy LSI Building Blocks Electronic Design Vol 27 no 9 April 26 1979 Hewlett Packard 3000 Hanover Street Palo Alto California 94304 HEWLETT PACKARD JOURNAL Bulk Rate U S Postage Paid Hewlett Packard Company G200020707EEL5LACEC ADD JOHN HOPKINS UNI i KINS UNI APPLIED PHYSICS Yas JOHNS HOPKINS LAUREL MD 20707 CHANGE OF ADDRESS 5953 8548 1949 1998 Hewlett Packard Co
27. usually provide some sort of spectral display in which a Fourier transform is taken of the Doppler with the same periodicity The delay time determines the axial location of the sample volume and the length of the sample volume depends on the duration of the burst It should be noted that signals returned from cells throughout the entire volume arrive at essentially the same instant so that the electrical sample represents the sum of all scattering within that space The first pulse launched travels farthest and the pulse launched last travels the shortest distance but their reflections arrive together see Fig 1 Extension of the sample volume can be obtained by in crease in the transmitted burst length but we have chosen an alternative method where the volume is extended by accumulat ing additional samples taken at appropriate successive intervals The effect is to increase the volume as desired without having to change the transmitted burst any way see Fig 2 if the burst length is reduced to a single pulse and the period between bursts is reduced to the pulse period the ultrasound period and the sampling is also done with this periodicity the arrangement becomes a CW Doppler system The sample length then extends indefinitely This has the effect of a searchlight which affords a fast deep look but has no spatial resolution In this mode the transmitted amplitude is reduced to maintain an allowable power level
28. APE 0003 5 a 1949 1998 Hewlett Packard or 15 bits of significance This does occur Carrying only 12 bits internally means then that overflows can poten tially occur during the final three passes Fortunately these overflows can be minimized by decreasing the front end input gain Overflows are less likely to occur with broad band input signals in the presence of noise a realistic Doppler scenario Pure tones would be more likely to cause overflows in any one frequency bin At the completion of each FFT the spectral results are transferred to the waveform moment calculators This op eration is accomplished by effectively adding an eighth and ninth pass to the FFT operation This time is also used to transfer the input buffer s data into the FFT working memory for the next transform Thus we can determine the total FFT calculation time by multiplying the nine passes by the 64 butterflies per pass and by the eight 5 MHz system master clocks necessary per butterfly The result ing total time is 907 us Spectral Moment Calculations After the FFT operation a good amount of number crunch ing still remains The first block of circuitry converts the complex spectrum into magnitudes by squaring summing and taking the square roots of the real and imaginary com ponents for each of the 128 spectral bins Again this pro cess must be run at the 907 rate to be ready for the next
29. Doppler shift waveform are obtained by sampling the echo at a fixed time after the pulse leaves the transducer As the wall advances the phase of the sample changes corresponding to the amount the wall has moved between successive samples If the wall moves quickly enough to advance the phase of the sample by more than 180 degrees as in Fig 3b the sampled Doppler shift waveform is aliased to a lower frequency This is a severe limitation of pulsed Doppler instruments As the depth at which one wishes to interrogate blood flow increases the length of time between the transmitted pulse and the sam ple gate increases Each echo must be allowed enough time to return from the maximum depth of interest before the next burst is transmitted to prevent ambiguous range infor mation The burst repetition frequency in turn determines the maximum Doppler shift frequency that can be detected without aliasing a result of the Nyquist phenomenon As a result a trade off between maximum sample volume depth and maximum unaliased blood velocity exists that did not occur with the CW system Pulsed Doppler systems have some other interesting properties which are related to Heisenberg s Uncertainty Copr 1949 1998 Hewlett Packard Co Principle The problem with pulsed Doppler systems is that to localize the sample volume accurately in space one requires a shorter transmitted burst This corresponds to a wider transmitted bandwidth However as H
30. FFT output After magnitude conversion the data is passed through a ROM that contains several spectral compression maps Working ROM To Microprocessor Shift Register ranging from linear to logarithmic These maps are selected by the user depending on what information is sought from the spectrum Once the magnitude of each of the 128 bins is computed the spectral mean maximum and standard deviation need to be determined This calculation also must be performed in the same time 1 6 one FFT time 907 To compute the mean frequency of a spectrum several summations take place The statistical names of these sums are the zeroth moment and the first moment The zeroth moment is simply the sum of the 128 bin magnitudes The first moment is the sum of the 128 bin magnitudes multiplied by their corresponding bin number The mean then is defined as the first moment divided by the zeroth moment To determine the standard deviation a third summation the second moment is computed The second moment is merely the sum of the 128 bin magnitudes multiplied by the square of their corresponding bin numbers The stan dard deviation is defined to be the square root of the second moment divided by the zeroth moment minus the mean squared The equations are 127 Zeroth moment gt Magnitude bin bin o 127 First moment gt Magnitude bin x bin bin o Gray Scale Spectrum to Display Sray S
31. IF filter section Fig 2 consists of two parts the first being a general filter common to all three IF signal frequencies one for each transducer This filter has Butter worth bandpass characteristics which pass the complete range of intermediate frequencies from one to three megahertz In addition a notch filter at 3 81 MHz is used to attenuate any signal component from the local oscillator for the 2 5 MHz transducer Should such a frequency get Transducer Type IF Signal Selectable from the Scanner 1 125 225 Frequency MHz Frequency MHz Fig 2 IF filter section JUNE 1986 HEWLETT PACKARD JOURNAL 35 Copr 1949 1998 Hewlett Packard Co passed it would result in an unwanted tone the Doppler audio output and spectral display The latter part of the general filter is a set of processor selectable tuned tank circuits one for each IF input and transducer operating frequency Each tank circuit consists of a simple two pole LC filter centered at 1 25 2 0 or 2 5 MHz and having a bandwidth matched to the spectral content of the transmit ted signal Although simple in design each tank circuit filter in conjunction with the Butterworth filter provides a sensitivity within 0 5 dB of a perfect matched filter Quadrature Sampler Although the IF filters optimize the signal to noise ratio of the returning echo the Doppler signal has yet to be extracted A sampling process provides the me
32. It is possible to track this change in curvature by dynamically altering the delay settings during reception This feature known as dynamic receive focusing provides an extended depth of field T HE HP 77200B SCANNER is the beam forming elec Architecture A block diagram of the scanner is shown in Fig 3 A custom HP 5 microprocessor determines the scan mode by programming various cards in the system The timer provides the timing and clock signals needed to transmit pulses and receive reflections at the desired time for each line Transmitter and receiver time delays or coefficients are sent over the coefficient bus by the coefficient generator Transmitter time delays are produced by counters pro grammed by the coefficients The outputs of these counters are stepped up to form the high voltage pulses required to drive the transducer array Echoes are detected by the same array and passed on to the receiver processing chain Re ceiver delays are achieved by two different mechanisms both programmed by coefficients Fine delay adjustment is produced by altering the phase of the mixing frequency used to convert the received RF signal to the desired IF The receiver outputs are then entered into appropriate taps of a summing delay line The choice of a tap input corre sponds to coarse receive delay The output of the tapped delay line is summed IF ready for processing by the Dop pler instrument Receiver gain is modi
33. L display Copr 1949 1998 Hewlett Packard Co design rules If a parameter is not passed to the inverter code then the default value will be used Inside the deflayout we have two part calls to defiayouts that have been previously defined standard pullup and rectangular fet Since the four pa rameters to the inverter code actually refer to the transistor sizes of the constituent parts we must pass these parame ters down to the pullup and pulldown code actually re sponsible for sizing the transistors Finally notice that the last two lines of the align the part named pulldown of type rectangular fet so that the top center of the part aligns with the bottom center of the diffusion connection of the part pullup We have specified the alignment of the pulldown part such that no matter where the pullup part is placed or how large it becomes the pulldown part will always align with it to form a correct inverter One long term goal of our project is to foster the creation of a large library of highly configurable ICPL cells which will in effect be a collection of module generation proce dures that work together Several advanced users of ICPL have written module generators for many of the regular structures of VLSI design including random access memories read only memories and programmable logic arrays In advanced ICPL code even a simple cell such as an inverter is created by a detailed procedure which
34. Massachusetts is married and has four children His outside interests include sailing and writing computer games 26 Doppler Theory Paul A Magnin Paul Magnin earned a BSEE degree from Prince ton University in 1977 and PhD degree from Duke University in 1981 With HP since 1981 heisa special istin Doppler flow mapping and the author of 12 papers phased array imaging BA and Doppler effects He contributed to the development of the HP 77410A Doppler Processor andis currently an R amp D project manager Paul was born in Marion Indiana and nowlves in Andover Massachusetts with his wite and son lt Power intensity Measurements Born in Nanking China James Chen earned a BS degree at Chung Yuang University in Taiwan in 1970 He also received a PhD degree in physics trom Boston College 1976 did research on ultrasound material and nondestruc tive evaluation related to the nuclear industry before coming to HP in 1981 His primary contribution at HP has been to phased array transducer development He is the author or coauthor of several conference papers and his work on acoustic imaging devices is the subject of both a patent and a patent application James lives inChelmsford Massachusetts is married and has three children thor of several technical papers and is named nventor on two patent applications related to ultra sound Les was born in New Yo
35. TT PACKARD JOURNAL JUNE 1986 Hewlett Packard Company 1986 Printed in U S A Copr 1949 1998 Hewlett Packard Co In this Issue The integrated circuit artwork on our cover this month was drawn automat ically by what s sometimes called a silicon compiler As the authors of the article on page 4 explain The process of silicon compilation is analogous to the role of high level language compilers in software engineering The designer specifies the behavior and or structure of the desired circuit high level language Working from the high level specifications the silicon compiler generates the detailed circuitry programmatically This not only is faster than design by hand but also allows system designers not familiar with IC design techniques to create an IC chip The subject of the article is a high level language used in house by HP designers to specify a chip Called ICPL for Integrated Circuit Procedural Language it s a Lisp based language similar to MIT s DPL About as new as silicon compilation is just in time manufacturing the technique pioneered Japan and now spreading throughout the world An estimated 250 plants in the U S A have implemented it and most large manufacturing companies are testing the concept The article on page 11 describes the design of HP JIT an HP software product intended to be used by just in time manufacturers for material requirements planning and production control Worthy of note a
36. a set of resistors whose ratio determines two pole locations The Nyquist filter is designed as a modified But terworth filter Because it is a Butterworth filter it passes frequencies from dc to the Nyquist rate with unity gain The reason for its modification is to account for the filtering characteristics of the zero order hold at the quadrature samplers At the Nyquist frequency a zero order hold at tenuates the signal by 4 dB The modification of the Nyquist filter compensates for this attenuation by appropriately in creasing the gain Therefore the overall DDC frequency response has an extremely flat profile from dc to the Nyquist rate The output of the Nyquist filters is the recon structed quadrature Doppler flow signal Blood flow is observed by periodically sampling the ultrasonic backscattering from blood moving through a selected region The results are processed to obtain blood velocities The term velocities is appropriate since the sample contains cells moving at varying and various speeds and directions throughout the sample volume The angular location of the sample is selected by directing the transducer The lateral size of the sample depends on aperture and focus Pulsed Doppler techniques permit selection of the sample site and extent Larger sample sizes make searches easier smaller sample volumes permit pinpointing of flow phenomena The Doppler data is usually obtained and displayed together with the image Pu
37. a software program Circuit parts resemble subroutines take parameters and be called by other parts Thus smaller parts can be hierarchically built into larger parts freeing the designer to concentrate on higher level issues This provides a high degree of de sign leverage Parts generated programmatically are also less prone to error than parts drawn by hand One reason is that values such as the size of a component or the spacing between two components can be expressed as fabrication process variables rather than by hardcoded numbers De signers can also specify a part s location relative to other parts We chose the Lisp programming language as a basis for silicon compilation and procedural layout design tools for several reasons After more than 30 years of language and application development the Lisp environment is un equaled in providing a powerful set of editing and debug ging aids that translates into a friendly and extensible user interface Lisp is symbolic meaning that it naturally cap tures the way in which real world knowledge is rep resented for problem solving The interactive nature of Lisp allows for a rich debugging environment for development of both procedural layout design tools and the circuit de signs that use them Dynamic linking and loading provide a high degree of incrementalism for exploratory program ming and Lisp can be compiled for production quality programs Copr 1949 1998 He
38. al batch stock INSP wot w02 wo2 communication Many readers found it difficult to vis With these concerns in mind we decided to try a differ ualize the final product solely on the basis of written ent approach on the HP JIT project Rather than use the specifications and often did not review the documents traditional method of creating an external specification carefully document we decided that software prototyping was a External specification documents typically spend a great much better channel for feedback and would allow us to deal of time in the draft stage going through a lengthy solicit customer and marketing input early the develop draft review rewrite cycle before they can be shown toend ment cycle when it would be most effective in determining users field marketing and others outside of the develop the feature set of the product and providing a basis for the ment team in the lab The time spent drafting the external internal design Prototyping the software at this point in specifications is a period when much of the progress on the project provides the best basis for asking vital questions the project is not readily measurable This makes the pro about the proposed feature set of the product and for seek cess of creating the product difficult to manage during this ing answers to these questions iteratively by means of stage and in the worst case makes it possible forthe project successive prototypes that
39. ans by which the Doppler shift is detected and hence the means by which the blood velocity can be determined at a given depth within the body To describe how the sampling process works requires a basic understanding of how the desired Doppler signal is encoded in the returning echo In pulse Doppler mode the signal transmitted into the body by its repetitive nature contains energy only at the harmonics of the pulse repeti tion frequency PRF The returning echo on the other hand can be considered to originate from two independent types of sources The first echo comes from stationary tis sue and like the transmitted signal contains energy only at the PRF harmonics The second echo originates from blood and nonstationary tissue Since these targets are mov ing the returning energy is shifted from each of the PRF harmonics by an amount proportional to the target velocity as described by the Doppler equation It is this shift in frequency that we detect The sum of these two echoes is then sampled at each and every pulse repetition interval PRI 1 PRF at a specific time corresponding to the depth of the desired sample volume One might object to the sampling of a relatively high fre quency signal 1 25 MHz by such a slow rate as the PRF e g 5 kHz in that the Nyquist theorem is being violated However the theorem only refers to the ability to recon struct the original IF signal which is not important for pulse Doppler mode
40. because nonlinearity generates high frequency har monics which affect peak intensity significantly However these harmonics are high in frequency short time dura tion so they contribute little to the time average intensity Fig 2 compares linear and nonlinear waveforms Peak in tensities are very sensitive to these harmonics because the high frequency spikes occur normally at the end of the rarification phase of a pressure cycle significantly increas ing the received pressure amplitude at that phase Since this only occurs in a fraction of the cycle of the pressure wave the temporal average intensity is nearly unaffected 34 HEWLETT PACKARD JOURNAL JUNE 1986 by these high frequency spikes All the hydrophones we have evaluated appear to behave linearly through the inten sity range we normally use them for Intensity Measurement After complete calibration the system is ready to mea sure intensity This is done in a water tank where the trans ducer under test and the hydrophone are both immersed Since the objective of this measurement is to measure the highest intensity the ultrasonic transducer can output the operator has to locate a hydrophone position and a system setting that give the maximum intensity before making the intensity measurement For an ultrasound system that uses a fixed pulse repetition rate and fixed driving voltage the position of maximum intensity is easy to find The hydro phone is used to search through
41. cale ROM User Select From Microprocessor Reject ROM Shift Register Serial Ac Second Serial Accumulator 1 umula First Moment Bin Counter Accumulator Select Fig 2 Block diagram of digital processing chain 46 HEWLETT PACKARD JOURNAL JUNE 1986 Copr 1949 1998 Hewlett Packard Co 127 Secondmoment Magnitude bin xbin xbin bin 0 First moment Men Zeroth moment Second moment Standard deviation Zeroth moment Mean These moment calculations require 256 multiplies 128 squaring operations and 3 128 or 384 additions Also the growth of the summations is a problem The magnitudes of each bin are stored at this juncture as eight bit quantities In the first moment calculation if we multiply an eight bit quantity by a seven bit bin number we get a 15 bit product Summing over 128 bins yields 22 bit result The second moment is potentially worse since we multiply by a squared quantity seven bits squared gives 14 bits The second moment could then grow to 29 bits To resolve the long word lengths and to avoid using another flash multi plier in a time shared fashion serial calculations are used By serializing the data paths and using a hybrid serial mul tiplier half serial half parallel the amount of circuitry needed is greatly reduced This works because we have 907 ps to complete each summation We can calculate the maximum number of bits
42. ches to integrated circuit design view circuits as a collection of rectangular regions the geometrical patterns used to define the devices on an IC chip Circuit synthesis traditionally has been boiled down to the art of handcrafting this large collection of rectangles most recently accomplished through the use of computer based graphic editing This manual approach is highly tedi ous and error prone when dealing with hundreds of thousands of rectangles where one misplaced or under sized rectangle can cause catastrophic malfunction in the ci In addition the designer must constantly juggle many conflicting criteria such as total chip area power consumption and signal delay As the number of compo nents grows these interrelationships become exponen tially more complex This complexity causes many design ers to leverage their designs by using as many pieces of old designs as possible But since integrated circuit pro cesses are changing so rapidly very few old designs can be used in new ones The same type of adder for example must be redesigned for every fabrication process it is used in see Fig 1 Clearly more automated mechanisms for designing circuits are needed RECENT YEARS the design and fabrication of elec Silicon Compilation Silicon compilation is a burgeoning subdiscipline of computer aided design CAD that addresses the problem of managing VLSI complexity by automatically synthesiz ing complex des
43. cted by Doppler s theory to result ina much larger frequency shift Ballot was loaned a locomotive and a flatcar which was to carry a trumpet player able to play note with perfect absolute pitch A second musician also with perfect pitch stood in the train station and listened as the trumpet player passed The Stationary musician heard the the trumpet note one half tone higher as the train approached and one half tone lower as the train passed Although the experiment seemed to verify Doppler s theory Ballot published his account of the experiment in an article in which he voiced several objections to 5 theory Rather than establishing Doppler s theory Ballot s article seems to discredited Doppler for many years Reference 1 White Johann Christian Doppler and His Brief History Utrasound in Medicine and Biology Vol B no 6 1982 583 581 JUNE 1986 HEWLETT PACKARD JOURNAL 27 Copr 1949 1998 Hewlett Packard Co Listener was to process the received signal in quadrature so as to allow the differentiation of positive frequency shifts from negative frequency shifts Positive Doppler shift frequen cies correspond to structures moving toward the transducer upper sideband of the received signal and negative fre quencies correspond to structures moving away from the transducer lower sideband of the received signal These improved i
44. d silicon compilation are being integrated with the ICPL data structures see box on page 8 Although still in their infancy expert systems for CAD have begun to show their worth in solving problems that were intractable using conventional programming styles ICPL will provide the basis for powerful silicon compi lation tools in the future The productive Lisp environment the symbolic nature of ICPL and integration with expert systems will allow designers to design chips much more rapidly than ever before It is conceivable that an ICPL based silicon compiler could take an initial chip specifica tion generate artwork and continually iterate through the design refining the weakest areas in size speed power consumption etc until the design converges on the op timum Users will be able to get the power of VLSI design without needing to be familiar with the fine details of IC design techniques Acknowledgments The authors would like to thank and acknowledge a number of people for their contributions to ICPL These include William Barrett Bob Rogers and Mike Sorens for their early work in the design and implementation of ICPL Bill McCalla Joe Beyers Marco Negrete and the late Merrill Brooksby for their support Ira Goldstein Martin Griss Craig Zarmer Chris Perdue Alan Snyder and Gwyn Osnos for their help with the Prism system and with our early questions and certainly Liz Myers Dave Wells Jon Gibson Ed Weber Pete F
45. dding the circuitry necessary to read to or write from additional buses A single cell with variable characteristics eliminates the need for graphically generating the hundreds of traditional cells that correspond to every possible com bination of parameters The physical constraints imposed by a process technol ogy are captured symbolically in ICPL via Lisp functions that describe geometrical relations between parts For example min spacing poly diffusion will return a number that represents the minimum separa tion between a rectangle on the polysilicon layer of an IC chip and a rectangle on the chip s diffusion layer In tra ditional design methodologies a designer would have the poly diffusion separation rule memorized and would graphically separate the rectangles In ICPL a designer sim ply includes a Lisp form that tells the system to separate JUNE 1986 HEWLETT PACKARD JOURNAL 5 Copr 1949 1998 Hewlett Packard Co the two rectangles align gt gt some poly rectangle gt gt center left some poly rectangle pt to right gt gt center right some diffusion rectangle min spacing poly diffusion The Lisp form shown above would move some poly rectangle such that its left center is to the right of some diffusion rectangle by the minimum spacing between poly and diffusion ICPL captures the designer s reason for separating the two rectangles thus allowing the design to withstand minor cha
46. determined and expected to remain the same for a long time Quartz transducers can also be used to generate the sound field for calibration at the base frequency and at odd harmonics Thus a single transducer appropriately selected can be designed to carry out this broadband calibration at the required discrete frequencies This transducer should have a lower resonance frequency so that it can cover not only the frequency range but also the frequency interval Fig 2 Comparison of a linear waveform upper trace and waveform with nonlinear distortion lower trace The non linear distorted waveform generates high frequency spikes which affect the peak intensity significantly JUNE 1986 HEWLETT PACKARD JOURNAL 33 Copr 1949 1998 Hewlett Packard Co requirement For example 0 5 MHz transducer can be selected to do calibration at 0 5 1 5 2 5 3 5 MHz Sen sitivity calibration is basically the same as before However the quartz transducer must be calibrated for its acoustic output or mechanical conductance at all these harmonic frequencies Equation 3 can now be extended to this broad band calibration case be J Co ha o and 02 The sensitivity coefficient becomes V XoVor S f a 5 Linearity Calibration Quartz transducers are ideal for hydrophone calibration because of their frequency accuracy and stable mechanical conductance However they
47. e quencies from blood occur in the audible range Heart and vessel walls usually produce shifts in the 0 to 1200 Hz range normal blood flow causes shifts in the 0 to 5 kHz range and jets from septal defects and malfunctioning valves generate shifts in the 5 to 20 kHz range Early CW instru ments simply played the Doppler shift frequencies into a speaker The first refinement of these primitive instruments Johann Christian Doppler Johann Christian Doppler s paper the Colored Light of Double Stars and Some Other Heavenly Bodies was delivered to the Royal Bohemian Society of Learning in 1842 In spite of the current significance of his contribution to science he was so little regarded by his colleagues that today we know few details of his personality He was born in Salzburg Austria in 1805 and died at the early age of 49 of pulmonary disease He worked in relative isolation as a professor of elementary mathematics and practical geometry at the State Technical Academy in Prague Buys Ballot a contemporary of Doppler published his doctoral thesis in 1844 stating that he did not believe Doppler s theory could explain the color of double stars It was Ballot who felt theory should be put to the test and he conducted the now famous experimental verification of the Doppler effect He chose to use sound waves rather than light waves since the speed of sound being much slower than the speed of light was predi
48. e which can be measured by an uncalibrated hydrophone across a profile Thus the equation for p x y above allows us to calculate a pressure intensity at a point in the field Measuring the output of a hydrophone positioned at that point can be used to determine the calibration constant for that hydrophone The same quartz transducers previously mentioned are used regularly to generate the intensity field to calibrate our hydrophones Frequency Response Calibration and Compensation For the CW Doppler mode nearly single frequency pres sure waves are output from our system A hydrophone calibrated at the corresponding frequency is sufficient to measure the intensity However for most of the operating modes broadband acoustic signals are involved To mea sure the acoustic intensity accurately the measuring sys tem must be broadband calibrated This calibration is necessary to correct for a system frequency response that is not flat when calculating intensity There are two ways this frequency response curve can be used to correct the measured intensity One way is to store this curve in a computer and correct each hydrophone received signal for this frequency response in software The second way is to build hardware that compensates the frequency response of the hydrophone so that any signal received is frequency compensated The advantage of the first method is that it can compensate for a frequency re sponse of any shape The disadvan
49. e designer permit routing and automatic placement utilities Parameters are passed on a keyword value basis ting integration of defaulting to specified values if they are not supplied by the caller For example consider the following ICPL de scription of an NMOS inverter defiayout INVERTER pullup length min pullup length pullup width min gate size pulldown length min gate size pulldown width min gate size part standard pullup channel length gt gt pullup length pullup width part pulldown rectangular fet channel length gt gt pulldown length pulldown width channel width channel width top center gt gt bottom center diffusion connection pullup It shows one other powerful feature of ICPL other design ers can look at the code and understand why things were done in a certain way For example an explicit align state ment will show that there is some sort of spatial relation ship between two parts Another thing to notice about the above inverter example is the procedural style of the deflayout A deflayout is analo gous to a Lisp defun or Pascal procedure definition because it can take paramete 8 and gives a sequence of steps to The inverter takes four parameters pullup length pullup width pulldown length and pulldown width which all have default values taken directly from the fabrication process perform if meena Fig 2 ICP
50. e dynamic focus ing capability of the instrument Thus the receive focusing consisted of several radial focal zones while the transmit focus was fixed at one depth For accurate extraction of Doppler shift information it is desirable to calculate a set of mixer coefficients to focus the beam precisely at the sample volume This can give improved signal to noise ratios since Doppler operation then does not rely on the approximate focusing of a mixer focal zone As the operator moves the Doppler sample gate around the sector searching for blood flow the scanner dynamically recalculates transmit and receive Doppler coefficients to focus the beam at the sample gate location precisely This was accomplished by replacing the coeffi Wave Velocity v cient EPROMs with a 68000 microprocessor and a RAM When a new transducer is installed the processor board recognizes this change and the MC 5 microprocessor com municates the transducer identity to the 68000 New coef ficients fora full sector are calculated requiring about three seconds Since coefficients are stored for only one trans ducer at any one time the RAM memory space needed is much smaller than the earlier EPROM space Transmit coef ficients for four different imaging depths are calculated and stored This allows the optimal transmit coefficients to be selected dynamically each time the sector depth is changed without any coefficient recalculation When Doppler mode is enab
51. e next pass also in all succeeding passes the gain is decreased by k 2 This is in keeping with the fact that overflows in later passes are not as damaging as in the earlier passes The latches are again cleared and the process is repeated Finally the overflows are driven to the last pass by suc cessively less aggressive AGC control AGC gain at which point the gain must now be toggled to maintain the over flows within acceptable bounds It must be set low enough so as to contaminate too many bins in the output but high enough to see some acceptable number of overflows This number is the AGC setpoint But what keeps the AGC from bringing up the noise floor in the absence of signal The answer is nothing since that would be the role of any AGC in an attempt to find a signal However if another mechanism is added to the AGC al gorithm this noise pumping can be avoided Since the variance the square of the standard deviation is calculated in the Doppler processing chain this information can be used in conjunction with the overflows to control the gain The signal is considered to be noise if the variance is above a predetermined level If noise as determined by the vari ance is detected then the incremental increases in gain for the last pass are suppressed even though the overflows are below the level where gain increases would normally be asked for In this way the background noise level is pre vented from being increased
52. e sampler the implications are twofold first the Doppler shift range can only extend from 0 Hz to the Nyquist rate and second positive Doppler shifts cannot be resolved from negative shifts In other words forward flow can not be distinguished from reverse flow when a single sampler is used This fact is best illus trated in the time domain as shown in Fig 5 The echoes t 1 0 PRF 3PRF 2PRF SPRF 3PRF 2 2 2 Nyquist 4 Frequency Sinc f PRF Sin t PRF t 0 PRF PRF 2 PRF 3 PRF b 2 Valid Spectral Output of Zero Order Hold Doppler b Data i Unwanted Byproducts of Sampling t PRF 2 PRF Nyquist Fig 4 Spectral output of ideal sampler b Sinc function Using a zero order hold sampler has the effect of multiply ing a by b Copr 1949 1998 Hewlett Packard Co from forward moving and reverse moving targets shown for five PRF intervals The filtered output of the first sampler is the same for both directions Therefore the DDC is implemented with quadrature samplers to resolve forward and reverse flow A quadrature sampling scheme uses a second sampler quadrature sam pler that samples a short time after the first sampler in phase sampler This time delay is equal to one fourth the period of the IF Again looking at Fig 5 it is shown how the output signals of the quadrature sampler are differ ent for fo
53. ead al filter to complete the transformation imates quadrature samples supplied b y pre scale spectral frequ d al sign F frequency on the Y axis and ma am for dig spectral mean maximum and stan The FFT performs main to the frequency domain It tak transformat dor Doppler data jesigned arou rature samples obtained from the dept this oks like with the g 2 shows process d a 12 bit flash multiplier levice is capable of performing a 12x12 multi yields a 12 bit complex 128 bit product 140 ns 128 point radix input data rate is tired FFT requires 448 butterflies which in turn become rate f splay 4 integer multiplies performed in seven bout 5 kHz while new spectrur r is used t lisecond an input buff processing from the sampling proc spectrum to be calculated on de es each The operation for processing rnal to an FFT is referred to as a butterfly 1 resembles Since the FFT numbers on each pass many new input samples have been taken In fact most of one n bit growth from a seven pass the samples the input buffer will overlap with previous FFT ince we start with eight bit samples and have FFTs This overlap serves to smooth the gray scale spectral seven passes one would expect these sums to grow to 8 7 FROZEN T
54. ed by the lens of the probe Once the maximum Ispra is calcu lated it is scaled down to clinically acceptable levels by programming a reduced transmit drive voltage Refer to the article on page 31 for a discussion of how acoustic power is measured Transmit and TGC Controls To make system operation as simple as possible it is desirable to eliminate the need for the operator to make gain control changes when scanning modes change Typi cally the operator first optimizes the control settings with only imaging mode on Imaging usually requires moderate to high levels of transmit power and moderate TGC receiver gain settings Thus the full dynamic range of the front end is rarely used in imaging Higher TGC settings cause the image to bloom and become noisy Echoes received from body tissue typically are 40 dB higher than those from blood Therefore the optimal sys tem configuration for imaging does not usually coincide with optimal Doppler operation Increasing transmit or TGC gain to allow for good Doppler data would degrade 44 HEWLETT PACKARD JOURNAL JUNE 1986 the sector image unacceptably Too large a Doppler signal could cause saturation producing unwanted tones in the audio and spectral display outputs Thus the Doppler signal requires a greater dynamic range To resolve the differing sensitivities a separate three position Doppler transmit control was added One position yields the maximum clin ically acceptable output powe
55. eisenberg pre dicts when our ability to localize the sample volume in space increases our ability to measure the velocity accu rately decreases This can be seen if one examines the pulsed Doppler system in the frequency domain If the transmitted pulse shape is assumed to have a Gaussian amplitude envelope the time domain signal can be written gt i pit Gaus 152 sin 2af t 1 9 where o standard deviation of the Gaussian envelope to time of flight to and from the sample volume of interest and Gaus 2 exp x Since the pulse is repeated at the pulse repetition frequency PRF the in sonifying signal can be described as Reflected Pulse z i g pit 5 d t PRIxn 4 5 10 where PRI pulse repetition interval 1 PRF The magnitude of the Fourier transform of the time do main signal in equation 10 can be written Pf K _d f nPRI x Gaus otV2 64 11 where is gain constant 6 x is the Dirac delta function and 5 is an asymmetric delta function where the im pulse function assumes the sign of the argument Equation 11 is drawn in Fig 4a for f 8 This is the spectrum returned from stationary objects and is referred to as the clutter spectrum Optimistic souls may assume that the spectrum of the signal with Doppler information included Fig 3 a Sampled voltage for Doppler s
56. er echoes are received by the scanner at an RF frequency of 2 5 3 5 or 5 0 MHz depending on the trans ducer used These frequencies are mixed down to 1 3 1 9 and 2 5 MHz respectively and tapped off the scanner s video processor card to be sent to the Doppler subsystem The scanner timer generates a 60 MHz master clock and start of line sync pulse for the HP 77410A This clock is used to phase lock the 77410 sampling of the IF signal to the scanner Dynamic Coefficient Generation The most significant feature incorporated into the new scanner is the dynamic calculation of delay coefficients The coefficients determine the angle and depth of focus of the acoustic beam for both transmit and receive This is accomplished by adjusting the time delays of transmitted pulses and detected echoes Earlier versions of the coeffi cient memory card stored the coefficients in a large set of EPROMs electrically programmed read only memories This limited amount of read only memory allowed the scanner to support only four different transducers Instead of storing a set of coefficients with fixed foci dynamic generation of coefficients allows focusing at any point in the scan sector a major requirement for Doppler operation For each acoustic beam angle a set of transmitter tap selector and mixer coefficients was originally stored Ad ditional sets of mixer coefficients allowed fine adjustments of focusing while receiving providing th
57. erience of those designers is normally lost or becomes very inaccessible to the designers that follow Incorporating expertise from multiple experts Designers often have to consult with many other designers to achieve the best possible results However the required expertise is frequently not available when needed As an alternative the KAD system could collect the knowledge of many expert sources allowing the designer to use the system as a readily accessible group of consultants in the design and critique of the chip In one scenario the designer presents to the system a block diagram for a data path showing interconnections between regis ters adders and multipliers Certain information may be missing or incompletely specified The system in such cases can assist by filling in much of the omitted information For example the chain length for an adder can be deduced from the number of inputs and outputs This in turn provides the means to estimate the maximum time delay the approximate area that the adder will take up and the power it will consume The environment should know what CAD tools are applicable for each design 8 HEWLETT PACKARD JOURNAL JUNE 1986 problem perform data translation and interpret results of running simulation and analysis tools We intend to provide a KAD framework for tool writers to build knowledge based tools upon and to provide an intelligent control mechanism for incorporating those tools We have c
58. erification A graphical editing capability is also tightly integrated into the ICPL design environment Developing designs can be viewed on a color display with mouse actuated access to the ICPL data structures Using the mouse designers have the ability to trace electrical networks move parts around on an experimental basis and obtain descriptions of the design hierarchy see Fig 2 Layouts entered graphi cally can also be converted to ICPL code Although graphi cal entry is a powerful tool for experimenting with circuit layouts it is not recommended for the final library cell because it does not capture the intent of the designer Some users find it helpful to begin a layout graphically and then modify the ICPL code to include the placement constraints that represent their intent Geometric structures and their combinations are rep resented in ICPL as Lisp objects Arbitrary properties can To acquire the knowledge for this system we have incorpo rated in the project team the skills of an IC design expert with whom we have conducted a long series of interviews Eventually we hope the system will capture his expertise in making use of all the above information sources partitioning problems combin ing cells adjusting aspect ratios consulting different but related designs making trade offs and setting up tolerance ranges tor block headroom based upon his confidence level at different steps There are also interesting techniq
59. erifying that the software indeed solved the problems it was de signed to handle and correctly provided the features specified The credibility of HP JIT as a product was in creased by the combination of being looked upon as a just in time manufacturing leader in the U S A and the partner divisions acting as reference sites showcasing HP Application Architecture HP JIT was designed to be coded in Pascal and to use the CT customizable technology software tools developed by HP s Administrative Productivity Operation CT soft ware tools free the application designers from much of the detail work involved in opening data bases and files han dling terminals and updating screens HP JIT went farther however and placed many of the common calls to th underlying CT tools inside of utility subroutines in the product that could be called from any of its modules see Fig 6 Utilities were written for data base searches screen handling and transaction logging as well as other more specific JIT functions The initial impetus behind the de velopment of these utilities was to avoid duplication of Pascal source code but each utility was designed for maximum flexibilty so as to be usable by transactions and modules yet to be developed Through the use of flexible common utilities duplication of Pascal source code throughout the HP JIT software was nearly eliminated Building an application such as HP JIT on top of a layer
60. essor for the HP 77200A Scanner worked on Doppler soft ware and participated in clinical trial evaluations and enhancements of the HP 77200A Betore coming to HP he designed radar signal processors Syd lives in Somerville Massachusetts and is an avid landscape photo grapher He has beena gold medal winner inthe Photographic Society of America annual competi tion He alsa plays ice hockey and is a Boston Bruins hockey fan Color Flow Map Sub system He studied computer systems engineering atthe University of Massachusetts receiving his BS degree in 1977 worked on digital design for missile bome computers before coming to HP He is coinventor on two patents related to scan conversion andis the coauthor of a past HP Journal article Barry is married and lives in Nashua New Hampshire He and his wife are expecting their first child later this year His hobbies include softball golf writing and personal computers With HP since 1979 Dave Hempstead has developed software for the HP 77400A Display Subsystem and for the HP 77410A Doppler Processor He was born in Morristown New Jersey and is a graduate of Rensselaer Polytechnic Institute BSEE 1979 the coauthor of an HP Journal article on the HP 77020A Ultrasound Imaging System and is in terested in real time graphics software and Motorola 68000 Processor programming Dave and his wite are residents of Methuen Mas sachuse
61. ett the company A California native she was bom in Healdsburg and lives in San Jose Her hob bies include water skiing flying and sewing Raj Bhargava An R amp D project manager at HP s Manufacturing Pro ductivity Division Raj Bhar gava has been with the company since 1981 He was born in London and eamed a BS degree in is mechanical engineering from Banaras Hindu Uni versity in 1978 He con tinued his education at Pennsylvania State Univer sity MSME 1979 and at the University of Michigan 1981 He also held research assistant posi tions at both Pennsylvania and Michigan AtHP he has led the initial implementation of the HP Desk 18 HEWLETT PACKARD JOURNAL JUNE 1986 Copr 1949 1998 Hewlett Packard Co networktor the ware development team for HP Management 3000 He is currently responsible for Califomia with h 20 Doppler Cardiac Diagnosis Raymond G O Connell Bom in Annapolis Mary land Ray O Connell is graduate of Worcester Polytechnic institute BSEE 1966 and Northeastern University MBA 1976 He joined HP in 1966 and has contributed to the develop ment of HP medical prod 4 ucts including the HP 7822A Arrythmia Monitor and the HP 78301A Bright Display He was project manager for the HP 77400A Display Subsystem and the HP 77410A Doppler Processor and now manages the develop color flow mapping hardware Ray lives in Andover
62. f the beam is 2 5 MHz Given the depth the maximum pulse rate is 6 4 kHz This means that the maximum measurable frequency shift is 3 2 kHz if the Nyquist criteria is observed This shift corresponds to a blood flow velocity of about one meter second assuming an angle of zero degrees between the transducer and the flow direction However the veloc ities associated with many valvular defects are much higher 3 to 5 m s in some cases For this reason continu ous wave CW Doppler ultrasound is used The trade off here is between the measurement of flow at a selected depth available from pulsed Doppler measurements and the maximum velocities obtained from CW Doppler tech niques Use of both techniques during an examination has become an accepted practice because of the importance of determining severity as well as location of the disease Fig 1 shows common cardiac windows used in obtaining signals from the numerous areas of interest of the heart Fig 2 illustrates how the two sides of the heart function normally Systole Diastole Systole Diastole Systole Diastole PA Diastolic Flow Right Atrium RA Tricuspid Valve TV F Right Ventricular Inflow Tract RVIT Systolic Flow Fig 2 Normal cardiac flow a The right heart Blood enters the RVOT Right kapine gno Tract RVOT heart through the vena cavae RVIT Pulmonic which empty into the right atrium a PORET Flow then proceeds trom the right
63. f the circuit and the output is the geometric layout of the design in ICPL Datapath users are provided with two libraries the MADL library which contains the behavioral descriptions for simulation and the ICPL library for artwork generation A designer only needs to specify the behavioral description for the design and the order in which the library cells should be placed to generate the full data path The physical layout of each cell is such that cells only need to be placed adjacent to one another for all bus connections to be made automatically The cells can also be sized for different out put drive capabilities by passing in the value of the capaci tive load on the driven net Since ICPL represents net based connectivity the calculation of total capacitance of the net is easily done by traversing the net and summing the capaci tance of each cell The design generated by Datapath is readily testable since all the library cells have scan path capability allowing all the nodes in the design to be excited and observed easily 10 HEWLETT PACKARD JOURNAL JUNE 1986 Advanced Work ICPL has been extended in many ways to take advantage of the powerful environment available in the HP 9000 Series 300 Computer family Extensive use has been made of the windowing system and the Starbase Graphics Li brary and ICPL has been rewritten in Common Lisp to conform to that emerging standard In addition expert sys tem tools for layout assistance an
64. face for forecast and backlog informa tion and a five year planning horizon Fig 2 shows a master production schedule produced using HP JIT JIT Material Requirements Planning determines the Fig 1 Seven major software modules make up the HP JIT package for just in time manufacturing JUNE 1986 HEWLETT PACKARD JOURNAL 11 Copr 1949 1998 Hewlett Packard Co quantities of parts required to support the master produc tion schedule daily production schedule and planned extra use of component parts JIT MRP reports can be run selectively over a range of part numbers controllers buyers or dates or according to selection criteria of the user s choice in either daily weekly or monthly incre ments or some combination of these The features of this module include selective MRP by date range component part range controller buyer range or other user defined attribute single level explosion option mix planning op tion effectivity dates and action preshortage reports Fig 3 on page 14 shows a JIT MRP screen for entering select criteria for an MRP run Manufacturing Control Manufacturing control is carried out on the factory floor via the post deduct transaction and the Production Report ing and Post Deduct module which records actual produc tion completed and at the same time relieves stock areas of inventory consumed in assembly Post deduct is a one step approach as opposed to the traditional issue and rece
65. factur ing management H P JIT IS A SOFTWARE SYSTEM that assists in the About HP JIT HP JIT is designed to be a stand alone product or to run in a combined environment with HP Materials Manage ment 3000 The combined system allows a user to avoid redundant parts and stock area data when both the tradi tional work order material requirements planning and just in time philosophies are used in the same facility The majority of the HP JIT software runs on the HP 3000 Computer using Hewlett Packard s Application Customizer and Monitor technology The Application Customizer and Application Monitor are themselves software tools that are used by application designers to engineer generalized soft ware systems and then by users to adapt the systems to their individual needs HP JIT also uses the HP 150 Personal Computer as a management workstation The HP 150 Com puter can serve as the interface to the HP 3000 Computer for accessing the material and production functions or as a stand alone personal computer for access to local decision support tools HP JIT is meant to complement the characteristics of a just in time method of production including Fixed production routings on the factory floor Single level bills of material flat structure Planning by production rate to accommodate steady pro duction targets over a period of time Short production cycles featuring quick assembly HP JIT has seven major software modules
66. ffect History and Theory by Paul A Magnin HE DOPPLER EFFECT sometimes thought of as a universal wave phenomenon is in fact fundamen tally different for light or electromagnetic waves and sound waves For sound waves one can derive the exact Doppler equation by imagining a stationary source of sound of some frequency and a listener at some distance from the source If the listener is stationary Fig 1a the listener experiences waves seconds where equals the speed of sound and is the wavelength How ever if the listener moves toward the source at a velocity v Fig 1b the listener will experience an additional waves in time T The frequency the listener hears is just the number of waves per unit time or vr 1 The frequency at the source which the listener hears while stationary is just f C A The difference between the two perceived frequencies is referred to as the Doppler shift frequency It can be expressed as fs f 1 v C f wf C 2 If the listener moves away from the source the listener experiences fewer waves in time T and the frequency heard is lower than the source frequency Equations 1 and 2 still hold but the motion away from the source corresponds to a negative velocity What is interesting is the difference between this case and the case where the source is in motion relative to a stationary listener Fig 2 When the source
67. fied by the time gain compensation TGC generators The generator outputs in crease receiver gain as a function of time to compensate for the weaker echoes received from deeper in the body The HP 77200A Scanner underwent considerable changes to allow acquisition of Doppler data and to interface to the Doppler instrument HP 77410A Major changes were made to the processor timer coefficient generator and TGC functions The operating software was restructured with the help of new hardware capabilities A description of the scanner problems and solutions involved in achiev ing a Doppler capability is given in this article For a more detailed description of the operation of the rest of the im aging system see references 1 and 2 Scanner System interface Commands defining the modes of operation are received Transducer Object 0 Time Delay Array Control ea ie la a Wee AN Ultrasound Pulse ek Direction _ 3 Ben Azimuth Receiver t 1 Output w X Axis AA Echo Wavetronts Range Fig 1 Block diagram of basic phased array ultrasound imaging system JUNE 1986 HEWLETT PACKARD JOURNAL 41 Copr 1949 1998 Hewlett Packard Co by the scanner processor from the HP 77120A Controller Subsystem over an HP IB IEEE 488 IEC 625 link Image data is digitized in the scanner and passed through the HP 77410A to the HP 77400B Display Subsystem DSS There it can be postprocessed and displayed on the monitor Doppl
68. fted frequencies Thus forward and reverse flow are phase encoded in the quadrature channels By using a decoding scheme which will be described shortly forward and reverse flows can now be separated from one another Equally important by adding the second sampler the effective Doppler bandwidth is doubled because now Doppler shifts from Nyquist to Nyquist frequencies be distinguished The most critical feature in the design of the samplers is the large dynamic range required approximately 100 dB This requires that the samplers be capable of dis criminating a low level signal in the presence of another signal one hundred thousand times larger in magnitude This huge dynamic range places a constraint on the amount of aperture jitter the samplers can have Aperture jitter is the variability in the actual time of sampling A dynamic range of 100 dB leads to a tolerable aperture jitter of about four picoseconds light travels only one millimeter in this time To obtain this kind of performance at an acceptable cost a discrete track and hold circuit was designed The two identical bandpass filters which strip out any high frequency components greater than 40 kHz and any low frequency wall noise lower than 600 Hz These filters are very carefully matched so that any phase or magnitude errors between them kept to an absolute minimum The two quadrature signals are then sent to the Doppler detector card which further f
69. g sources and or listeners is described with its application for medical analysis 27 Johann Christian Doppler 31 Power and Intensity Measurements for Ultrasonic Doppler Imaging Systems by James Chen Carefully controlling the acoustic energy transmitted into the human body requires accurate analysis methods 3 5 Extraction of Blood Flow Information Using Doppler Shifted Ultrasound by Leslie l Halberg and Karl Thiele Frequency shifts in ultrasonic echoes are detected by means of specially designed filters and a quadrature sampler 37 Continuous Wave Doppler Board 39 Observation of Blood Flow and Doppler Sample Volume 4 1 Modifying an Ultrasound Imaging Scanner for Doppler Measurements by Sydney M Karp Changes in timing more precise focusing processing enhancements power limiting software had to be developed 4 5 Digital Processing Chain for a Doppler Ultrasound Subsystem by Barry F Hunt Steven Leavitt and David Hempstead Time domain quadrature samples con verted into a gray scale spectral frequency display using a fast Fourier transform moment calcu lations and digital filtering nager KennethA Shaw Assistant Editor Nancy Teater Director Photographer Avid A Danielson Support Supervisor Susan E Wright European Production Supervisor Mit Editor Richard Dolan Associate Editor B Iustratoe Nancy Contreras Administrative Services Typography Anne 5 LoPrest 2 HEWLE
70. grated with HP s internal design sys tems permitting efficient interchange between ICPL and various circuit simulators graphical editors and design rule checking programs Designers who use ICPL properly can be assured that their designs will weather many process and methodology changes before they become obsolete Applications of ICPL Several complex module generators have been developed using ICPL A module generator automatically creates the layout for a large piece of a chip such as the RAM or ROM section Simpler library cells have also been designed that allow a large combination of parameters to have a large effect on the configuration of the cell Only one generic cell needs to be designed and stored in the cell library as a template from which many different configurable cells can be produced This feature can be shown in the example of the configur able register cell Fig 3 This register cell can read from write to not connect to or both read from and write to either of two buses There is an optional refresh capability JUNE 1986 HEWLETT PACKARD JOURNAL 9 Copr 1949 1998 Hewlett Packard Co and the width of the power buses be adjusted for different current densities and spacing constraints Fifty different configurations of the register can be generated by passing in different sets of parameters in the part call to the register This register cell shows that parameterization in ICPL al lows a cell libra
71. h level low frequencies The wall filters on the DDC are designed with four user selectable cutoffs 200 Hz 400 Hz 600 Hz and 800 Hz For the 2 5 MHz trans ducer 400 Hz corresponds to a target moving at 12 cm s By comparison blood velocities typically exceed 100 cm s in the aorta of a normal healthy patient It is desired that these filters reject the low frequency data without affecting the Doppler blood flow data This is accomplished by using a filter whose characteristics pro vide a sharp cutoff Since cutoff variability is a must this implied the need for an active filter where pole locations could be repositioned to maintain the sharp cutoff indepen Time 1 200 us dent of one another A major problem in using such a filter is its nonlinear phase response in the vicinity of the cutoff A slight variation in component tolerances between the in phase and quadrature wall filters would induce a consid erable phase difference between the two channels Since the ability to separate flow direction depends on the phase difference between channels being exactly 90 degrees any additional phase difference will show up as mirroring where positive flow will be seen in the reverse flow chan nel and vice versa Therefore the wall filters on the DDC are designed using components with tight tolerances Nyquist Filter In the quadrature channels all the useful Doppler infor mation is contained the frequency band from dc
72. hallenge in implementing this rota tional filter was how to delay the Q channel by 90 degrees over a relatively large frequency range 200 to 20 000 Hz This suggested the need for an all pass filter or phase shift er which adds phase to the signal while passing all frequen cies Since it is extremely difficult to shift a single channel by an absolute 90 degrees the rotational filter actually phase delays both in phase and quadrature signals such that the phase difference between the two channels is 90 degrees over the desired frequency range The operation of the FFT card for the spectral display is described in the article on page 45 References 1 R N McKnight Mixing Scheme to Focus Transducer Array Dynamically Hewlett Packard Journal Vol 34 no 12 December 1983 16 17 2 R D Gatzke J T Fearnside and S M Karp Electronic Scanner for a Phased Array Ultrasound Transducer ibid pp 13 20 Copr 1949 1998 Hewlett Packard Co Modifying an Ultrasound Imaging Scanner for Doppler Measurements by Sydney M Karp tronics box in the HP 77020A Phased Array Medical Ultrasound Imaging System To understand how beam forming is done consider the hypothetical phased array system shown in Fig 1 This system consists of n parallel channels each with its own transmitter and re ceiver Each transmitter outputs an ultrasound pulse of short duration into the human body The pulse is partially ref
73. hift caused by an echo froma slowly moving wall b Ifthe wall moves fast enough to shift the reflected wall by more than 180 degrees between sample points the frequency shift is aliased to lower frequency JUNE 1986 HEWLETT PACKARD JOURNAL 29 Copr 1949 1998 Hewlett Packard Co is simply the superposition of a clutter spectrum with another identical spectrum perhaps with lower mag nitude shifted by the Doppler shift frequency Such a spec trum is drawn in Fig 4b This spectrum shows only asingle Doppler shifted signal at each of the clutter harmonics This however is emphatically not the case An idealized spectrum of what one would actually see is shown in Fig 4c Each clutter harmonic produces its own Doppler shift frequency which is not the same differ ence frequency as each of the other clutter harmonics recall the f term in equation 6 This leads to the somewhat disconcerting fact that for a single particle moving at a single and constant velocity a spread of Doppler shift fre quencies will be received As the velocities increase the frequency shifts from the clutter harmonics increase The difference between each Doppler component and its respective clutter harmonic is proportional both to the frequency of the clutter harmonic and to the velocity of the scatterers A spread of shift fre quencies is measured whose bandwidth is proportional to the bandwidth of the transmitted pulse and to the vel
74. hosen as a pilot project a subpart in this scheme to use as a learning vehicle for understanding what is needed in the framework Furthermore this work will bring out in more detail the issues involved in build ing expert systems within such framework The pilot project is a knowledge assisted tool for aiding the planning phase of chip design Such a tool tracks design deci sions attempts to fill in missing details propagates hints and constraints and allows the user to explore functional and struc tural chip design alternatives In the decomposition phase of chip design the choice of different design alternatives results in differing impacts upon block and chip area power consumption timing delay and testability Our initial pilot project will provide feedback on one of these figures of merit block and chip area estimation for the design Thus there are two parts in this tool A front end planning assistant facilitates exploring of design alternatives and a back area estimation expert provides analysis information for de cision making Planning Assistant The front end segment deals with representation and manage ment of functional and structural hierarchies The problem in volved here is to record distinguish relate and apply knowledge of blocks to each new task During a session the planning assis tant accumulates design vocabularies in terms of block names part types and part properties in an attempt to ident
75. iac function not only for discovery of atient s initial condition but to track the patient s four chamber apical view occupies the up ise to treatment black and white display Below the image a single it is The traditional method for measuring cardiac c trace is displayed to allow correlation bottom half is oc the thermodilation technique A thermally sensitive throughout an entire cardiac cycle The in the upper left side of the cupied by Doppler spectral data obtained with sample atheter is placed to th 21 nov PROC a 3 HEWLETT 3 h Copr 1949 1998 Hewlett Packard Co heart A ere the solution comes around the poin below the temperature of the blood is injec the second time cardiac output volume flow can be cal vein through a port in the cathe ated Because of the inherent risks associated with flows into the right atrium A therm tion and fluid injection a noninvasive method of the catheter monitors the initial blc ture ardiac put would be much preferred stimated with Doppler ultrasound es by measuring the mean velocity of flow through temperature after injection of the cooled solution By Cardiac output t grating the concentration curve and extrapolating S MM S 2 5 2 _ FROZEN MODE GDC PROC 27A 12 JUL 84 09 29 19 HEWLETT PACKARD b Copr 1949 1998 Hewlett Packard C
76. ibrated the measured voltage waveform can be used to derive the intensities Various types of materials can be used to fabricate hydro phones Each has its advantages and disadvantages the discussion of which is beyond the scope of this article Our attention will focus on calibration and intensity mea surement using PVDF polyvinylidene diflouride hydro phones Hydrophone Calibration In our case a hydrophone is used as a pressure to voltage transformer What is the pressure voltage transfer ratio How flat is the response as a function of and How linear is the response as a function of intensity are key questions whose answers are linked to the accuracy of intensity measurement The piezoelectric properties of PVDF are stable and rather linear over a large dynamic range The hydrophones must be able to withstand the acoustic field put out by medical ultrasound equipment and they should have a flat frequency response However our ex perience with commercially available hydrophones indi cates that sensitivity can drift unexpectedly and that the frequency response is not flat and can vary with time This suggests that hydrophone calibration must be included as an integral part of intensity measurement The way to calibrate a hydrophone is to position the hydrophone in a known pressure field and read the hydro phone output voltage For example a hydrophone gives a voltage reading v when it is posit
77. ications liquid crystal display control and creation of custom plug in ROMs are some of the topics Also featured will be an article about the latest capabilities of HP UX HP s version of the UNIX operating system now available on HP 9000 Series 300 Computers We ll also have the design story of the HP 49715 a protocol analyzer for local area network installation and troubleshooting JUNE 1986 HEWLETT PACKARD JOURNAL Copr 1949 1998 Hewlett Packard Co Integrated Circuit Procedural Language ICPL is a Lisp embedded procedural layout language for VLSI design Circuit design in ICPL involves writing and working with programs that resemble procedures take parameters and can use the full symbolic programming power of Lisp This allows circuit designers to write high level software that procedurally builds ICs by Jeffrey A Lewis Andrew A Berlin Allan J Kuchinsky and Paul K Yip tronic circuits on microchips has achieved the level of very large scale integration VLSI It isn t uncommon today for a single chip to contain hundreds of thousands of transistor devices and we are beginning to see designs where the component count exceeds one million devices With this dramatic increase in chip density comes an equally pronounced increase in the complexity of the cir cuit designs resulting in a design bottleneck in the trans formation of functional specifications into manufactured products Conventional approa
78. igns from high level architectural descrip tions Variously referred to as systems for silicon assem bly module generation macrocell assembly or automated layout silicon compilation tools come in many variations What they have in common is that they raise the level of abstraction at which the chip designer operates so that the designer is more insulated from the complexity and vari 4 HEWLETT PACKARD JOURNAL JUNE 1986 ation in lower levels The designer is freed from performing the tedious and error prone tasks required by detailed man ual layout of rectangles Exploration of alternative architec tures is simplified making it easier to create more efficient and cost effective designs The process of silicon compilation is analogous to the role of high level language compilers in software engineer ing The designer specifies the behavior and or structure of the desired circuit in a high level language Working from the high level specifications the silicon compiler gen erates the detailed circuitry programmatically This is not only faster than design by hand but also allows system designers not familiar with IC design techniques to create an IC chip Silicon compilation ultimately makes chip de sign accessible to every circuit design engineer One popular mechanism by which silicon compilation tools generate circuit artwork programmatically is known as procedural layout In the procedural layout paradigm a chip design resembles
79. ilters them and processes them for the spectrum analysis Rich Jundanian Development Engineer Andover Division Transmit Control Bandpass F 600 He to 3 oe roma Fig 1 Block diagram of continu ous wave CW Doppler board JUNE 1986 HEWLETT PACKARD JOURNAL 37 Copr 1949 1998 Hewlett Packard Co basis of this design is very high speed low jitter analog switch Wall Filters Of the 100 dB dynamic range of the output signal from the samplers the upper 40 dB portion of the range consists mainly of unwanted wall signals A wall signal refers to the echo received from stationary or slowly moving specu lar reflectors such as vessel walls and intervening tissue between the probe and the flow volume interrogated Such an echo is typically 100 times as large as the echo received from blood and is distinguished by having a low frequency Doppler shift Thus a high pass wall filter is used to at tenuate these low frequency wall signals to the point where they no longer obscure the desired blood flow data For a given velocity the Doppler shift obtained from slowly moving tissue will vary according to the Doppler equation with a varying transmit frequency f In addition the velocities of slowly moving tissue can vary significantly from patient to patient For this reason it was decided to allow a limited selection of cutoff frequencies to reject these hig
80. ion that occur in acquired heart disease Many people are born with abnormal blood flow patterns The most common of these are openings in the septum of the heart that did not close in the normal development of the fetus The most dramatic of these is the transposition of the great vessels in which the pulmonary and systemic systems are completely separated Pulsed Doppler tech niques are useful in locating and determining the direction and extent of abnormal flow valvular abnormalities shunt lesions such as patent ductus arteriosus and ventricular and septal defects An extension to the ability to measure cardiac output is the quantification of left to right side shunts whether they be in the septum or elsewhere Two recent papers evaluat ing this technique stated Our work in animals and in the clinical setting has attempted to demonstrate sources of error in using echo Doppler for noninvasive flow quantification The multiple sampling site technique while time consuming suggests that echo Doppler can be used to quantify pulmonary and systemic flows and calculate flow ratios noninvasively We have demonstrated in this study that quantitative pulsed Doppler echocardiography can be successfully applied in the pediatric patient with isolated VSD to deter mine the magnitude of left to right shunt noninvasively Summary This has been short review of cardiac Doppler measure ment technology Use of Doppler ultrasound techni
81. ioned in a pressure 2 The ratio v p x y z is the calibration coefficient of that hydrophone Two methods are generally used to establish a known pressure intensity p x y z One is based on self reciprocity the other on radiation pressure The basic principle behind the first method is that if the input and output impedance conditions are properly met the acoustic field in a particular location can be derived from an appropriately designed round trip driving receiving voltage measurement The other method is based on the concept that the total acoustic power is an integration of the acoustic intensities measured over a field profile and the same total power can also be determined by a radiation pressure measurement using a force balance This discus sion will be limited to this second method The total power from a transducer is an integration of the intensities measured at various locations in an acoustic field profile For a two dimensional field Pr Sf PO 2 Copr 1949 1998 Hewlett Packard Co where 2 is the acoustic impedance This can further be reduced to ee 3 where 2 PRY and p X WP Z C which is the absolute peak pres sure amplitude in a pressure field The total power can be measured with a force balance as has been described The acoustic impedance Z is known The constant C is an integration of the square of the normalized pressure profil
82. ity design intent and apply its knowledge towards the goal It allows a designer to build design hierarchy in the designer s own style and terms using little or no detailed information When performing a task it tries to supply the missing detail from facts described by the user from related works done by some other designers and from knowledge it has based upon its understanding of similar designs When these mechanisms fail the system prompts the user for just enough information to continue the task The system thus aids in filling in forgotten details and identifying inconsistencies and unnecessary duplication This should en courage the exploration of as many alternatives as possible to find the best design decision The planning assistant can also record useful information such as hints and constraints to be used in later development phases This creates an on line notebook capability for the designer and improves information flow between design phases For example during the planning stage the system designer may have made a design decision based on the assumption that the RAM module will have an aspect ratio of 1 1 This vital information is stored with the RAM block as the only acceptable aspect ratio When the design is transferred to the IC designer to implement the physical layout this constraint is also communicated to the IC designer The planning assistant can carry large amounts of such information previously informally passed
83. ity numbers are very useful for de termining the volume of blood flowing in a vessel at a particular interval during the heart cycle and providesome measure of turbulence In addition the peak velocity can give some indication of the amount of occlusion a vessel since the narrower an opening the higher the velocity of blood flow given the same volume throughput FFT Automatic Gain Control The FFT fast Fourier transform within the Doppler dig ital processing chain uses a 128 point radix 2 implementa tion This breaks down to 64 two point butterflies on each of seven passes to complete an FFT After each butterfly the potential for a bit growth in the data width becomes a reality resulting from the fact that additions subtractions of two n bit numbers can yield an n 1 bit number The input data width starts at eight bits and after seven passes can grow to 15 bits for a pure tone input in the absence of noise As such provision must be made to accommodate the growth Several techniques are possible One can automatically scale the data after every pass which can incur a dynamic range problem One can assess whether growth has oc curred and if so then scale that pass which incurs addi tional processing time between passes One can provide a 15 bit data path through the FFT which incurs unnecessary processing costs particularly the multiplier or one can provide an AGC on the input signal such that overflows that serio
84. ive of a work order driven production reporting sys tem Defining a deduct point in HP JIT provides visibility of the consumption and movement of materials on the fac tory floor As a part is built and passes through a defined deduct point HP JIT records the creation of the part incre menting its inventory quantity and decrementing the in ventory quantities for its components On line review of actual versus planned production is provided in HP JIT features of the Production Reporting and Post Deduct module include post deduct of up to 25 options per parent ata time post deduct by engineering change level post de duct at subassembly or end product levels summarization capability that allows tuning of the HP JIT system by the user to optimize performance and actual production re porting by the hour or by the day Reviews of planned versus actual production use graphics Fig 4 on page 15 illustrates an HP JIT post deduct transaction Stock Areas and Deduct Lists module further defines the manufacturing process Corresponding to each deduct point is a deduct list which associates the component parts and their quantities consumed at that deduct point with the stock areas from which those components were picked in the assembly process The features of this module in clude multiple stock areas per multiple deduct point types intermediate end of line subcontract multiple de duct points per line multiple prod
85. l specifica tion phases of a large software project The traditional method produced investigation and ex ternal specification documents for review but these docu ments were generally felt to be an ineffective method of Hewlett Packard YOUR COMPANY NAME Master Production Schedule Product Number 93827A 12 85 Number of working days 21 Past Due Production 10 Production Rate Day prev plan o curr plan 15 Backlog Orders ie eee ee 18 200 Order Forecast lt gt se ss tes 50 Carry Forward Orders o Ce 5 coe sa En 250 Production month 325 Beginntig RGE TOTAL Shippable Units 325 12 HEWLETT PACKARD JOURNAL JUNE 1986 01 86 23 02 86 03 86 19 23 0 o 80 10 10 80 180 130 100 50 90 130 0 0 230 220 230 230 190 1840 75 75 45 Fig 2 An HP JIT master produc 305 265 1885 tion schedule Copr 1949 1998 Hewlett Packard Co Comparing Manufacturing Methods The difference between the batch stock flow of a traditional manufacturing system and the demand pull stock flow of a just in time system is illustrated by the following example Fig 1 shows that product A is composed of two parts B and C Part C is purchased while part B is fabricated from purchased X and Y Fig 2 shows traditional batch stock flow Parts X Y and are received inspected and placed in the wa
86. lected by structures in the body and these reflections are detected by the receivers The receivers then send the resulting signal to a delay mechanism and a summing junc tion By appropriate choices of delay settings beam steering can be performed If the transmit delays are set so that the delay for element i is equal to i 1 T then an acoustic wave is launched from element 1 at time 0 followed by a wave from element 2 at time and so Referring to Fig 2a these waves add to form a composite plane wave travel ing in a direction at an angle to the normal of the array By varying T the direction of the beam can be changed The composite wave can be formed into a nonplanar wave by choosing a nonlinear sequence of delays appropriate sequence of delays can result in a curved wave front that converges is focused at a single point as shown in Fig 2b Reception of a wave operates in a reverse manner Reflec tions from a target are returned to the transducer elements as curved wave fronts similar to those shown in Fig 2b The response of each element is amplified delayed and added to other element responses by a summing junction see Fig 1 Constructive addition of the element responses occurs when the delays are selected to compensate for the fact that a single wavefront strikes different elements at different times As the incident waves are reflected from targets deeper in the body their radius of curvature in creases
87. led the coefficients for a single Doppler line are calculated focused at the sample gate Each time the sample gate is moved by the operator commands are received by the scanner describing the new gate position Then the coefficients for the Doppler line are recalculated Since only one line and one focal zone are involved the recalculation takes only 15 ms Timer Changes The second major hardware change to the new scanner involved synchronizing all clocks in the instrument Earlier scanners had separate crystal oscillators driving the trans mitters mixers video analog to digital converter ADC and microprocessor hardware Doppler operation requires that all these processes be synchronized otherwise differ ent phase relationships at successive Doppler sample times will create phase changes in the Doppler data This would manifest itself as erroneous Doppler flow signals The timer card was chosen to be the source of all scanner clocks Master frequencies of 120 MHz or 180 MHz can be generated on this card depending on transducer frequency These master frequencies were chosen so they could be divided down to the mixer and transmitter frequencies required by the different transducers The video ADC clock and all other clocks are also derived from the master clock New Doppler modes involve sequences of both imaging and Doppler lines In duplex Doppler mode the scanner alternates between firing imaging and Doppler lines This allows
88. ll Jr the toolset that was described by Dr Richard Popp in his article Physician s View of Echocardiog raphy Although the technology is not new to medicine its acceptance in the United States for cardiac diagnosis is relatively recent This article will discuss a few of the clin ical applications of Doppler ultrasound and compare alter native procedures D OPPLER ULTRASOUND represents an extension of History of Ultrasound For over twenty years ultrasound has been used to aid in the diagnosis of certain cardiac diseases The first use was the time motion study called M mode today The technique involves transmitting a beam of ultrasound and plotting the intensities of the returning echoes across a strip of paper As more and more lines are plotted the locus of the motion of the echoes is plotted in time This technique allows the diagnosis of stenotic valves valve leaflet defects and pericardial effusion Two dimensional real time imaging systems with Dop pler capabilities were developed to enhance M mode The two dimensional imaging systems caught on quickly and led to the demise of stand alone Doppler equipment Ul trasound imaging allowed the visualization of the heart over an entire cardiac cycle in real time Reference 2 is a good review paper for the state of ultrasound medical tech niques 1982 Early Uses The first Doppler systems available were continuous wave systems that were used in the study
89. lly different in the case of Doppler shifts of light waves The difference stems from Einstein s Theory of Relativity which states that the speed of light is a constant in all reference frames Since light needs no material medium to support its propagation its speed relative to the source or observer is always the same Therefore it is only the relative motion between the ob server and the source that determines the Doppler shift frequency The Doppler shift frequency for light predicted by Einstein s theory is fa devi wwe 1 7 where cis the velocity of light and 1 is the relative velocity of the source and observer and is positive if the source and observer are moving away from each other It is fairly easy to see that for the case where the source and listener velocities are small compared to the velocities of sound and light equations 6 and 7 are approximately the same Ultrasonic Doppler Blood Flow Measurements It was recognized that ultrasonic waves similar to those used in submarine sonar systems could be used to locate structures the body shortly after World War Il It became apparent that one could receive not only large echoes from Source Fig 1 a A stationary listener hears cycles from sta tionary sound source in time b If the listener moves toward the stationary source at velocity the listener hears an addi tional cycles in time
90. locity meters have met with tremendous clinical success My OA Re 18 18 38 PRI PRI PAI PRI PRI PRI PRI PRI Pit P t ala Al 14 Pai pri PRi fos for tan ta tar fen Fig 4 a Clutter spectrum returned from stationary objects 6 Smaller shifted spectrum superimposed the clutter spectrum Actual Doppler spectrum see text Copr 1949 1998 Hewlett Packard Co New systems incorporate both continuous and pulsed wave techniques to avoid the frequency aliasing ambiguities of the latter and the range resolution ambiguities of the former Physicians tend to overlook the problem of spectral spread ing and for the most part it has not proved to be a major drawback of the instrument One remaining problem with commercial Doppler systems seems to be their sensitivity Because of the weakness of the echoes from red blood cells even the best commercially available systems can rarely acquire good Doppler signals at depths beyond 14 cm Dop pler systems also tend to be fairly cumbersome to use The operator must move the sample volume to the position of interest and listen or look at the Doppler spectrum for a number of heart cycles to determine if abnormal flow is present Searching for a jet from a ventricular septal defect or a regurgitating valve may require stepping the sample volume over a large a
91. lsed Doppler mode creates a sample volume in the following manner A burst of several pulses is launched periodically The return from scatterers in the blood is sampled some time later w b 37 b s o 28 j 2 p gt 28 Fig 1 The leading edge t 0 of the pulse train a returns from the far end of the sample volume at time 2 d s v The trailing edge t T of the pulse train returns from the near end of the sample volume at time 20 For s v 7 2 these returns are simultaneous and therefore are auto matically summed by the sampler c The sampling period at least equal to 2dlv Observation of Blood Flow and Doppler Sample Volume AGC and Analog to Digital Conversion After removal of unwanted signal components it is necessary to provide gain This gain control must be vari able because the blood flow echoes extracted from different vessels and chambers vary in signal strength This is complished through the use of a pair of analog multiplier chips one each for the in phase and quadrature channels The gain level for these multipliers is set by an eight bit digital to analog converter DAC under processor control which allows for a swing of 48 dB A complex algorithm based on the fast Fourier transform FFT determines when and by how much the gain should change This algorithm also has a time constant that controls the rate at which gain is changed Doppler systems
92. lvular Stenosis The degree of valvular stenosis has traditionally been diagnosed in the catheterization laboratory by m pasuring the pressure drop across the valve Pressure measurements are obtained by inserting pressure catheters into the right brachial vein and the left femoral artery the tricuspid valve and the pulmonic valve can be measured Pressures across directly with the venous catheter by weaving it through each valve and measuring pressures as the catheter is pulled back through the valve In a similar fashion the pressure across the aortic valve SOMM S 2 5 2 PROC 17A 18 FEB 84 be measured by pushing the arterial catheter around the aortic arch and through the aortic valve into the left ventricle Pressure across the mitral valve cannot be measured in this fashion because the catheter must be forced against the flow of blood and thus cannot be easily positioned behind the valve To obtain mitral valve pressure drop pulmonary wedge pressure is measured In this procedure a venous balloon catheter is floated or sailed up the pulmonary artery until it wedges as the pulmonary artery size decreases This pres sure value is used for the left atrial pressure behind the mitral valve since the pressure drop across the lungs is very small To guide the catheters and avoid the risk of puncturing vessels as the catheters are inserted fluoro theter position constantly Cardiac catheterization has bee
93. moves toward the stationary listener the wavelengths are in effect shortened because between the time the crest of one wave is emitted and the succeeding crest is emitted the source has moved closer to the listener If the source has a frequency f and a velocity toward the listener of v then in the time it takes to emit one wave the source travels a distance v f The wavelength is compressed from the stationary source case of to Api C f vfs The frequency that the listener hears is fo C C v 3 and the Doppler shift frequency is fa fs f va 4 If the source moves away from the listener the velocity term becomes negative and likewise the Doppler shift be comes negative From equations 2 and 4 one can see that the motion of the source has a different Doppler shift fre quency than the same motion of the listener That is for sound it is not just the relative motion between source and listener that is important but which one is in motion Furthermore as the velocity of the source approaches the 26 HEWLETT PACKARD JOURNAL JUNE 1986 velocity of sound toward the listener the Doppler shift frequency becomes infinite Equations 2 and 4 can be com bined to give the general Doppler equation for sound waves fa f r va C v 5 This usually is simplified by assuming that v is small compared to C Therefore fa v C for lt lt 6 The situation is fundamenta
94. mum velocity present is needed This calculation puts the heaviest load on the software To calculate a maximum velocity the soft ware must search through the 128 spectral bins and attempt to determine the noise floor of the data Then the software again searches through all 128 bins this time looking for the maximum frequency bin containing an amplitude well above the noise floor After completing the statistical computations a smooth ing low pass filter is used to eliminate noise spikes in the output waveforms As a final step the digital values for mean variance mode or maximum are written to 10 bit digital to analog converters DACs The analog waveforms output by the DACs represent the mean frequency and standard deviation of the Doppler spectrum They are mapped to the mean and standard de viation of the blood flow velocities through the Doppler relationship velocity AfC 2f cos 0 where f is the frequency of the acoustic energy C is the velocity of sound in the body and the variables Af and 0 are calculated by the microprocessor from the pulse repeti tion frequency sample rate and the direction of the angle cursor on the sector The angle cursor is a special user ma nipulated orientation graphic superimposed on top of the two dimensional display image The clinician points this cursor in the direction of expected flow and all velocity numbers are corrected by this flow angle cosine term These corrected veloc
95. n considered routine proce scope is used to monitor the c dure for determining the need for surgical replacement of cardiac valves Using Doppler equipment a single noninvasive measure ment can obtain the press drop across a stenotic valve These dramatic results were the work of Holen et al who showed that Bernoulli s equation could be simplified to the following where is the pressure proximal to the valve P is the pressure distal to the valve and v is the blood velocity at the valve exit point highest velocity measured 16 FROZEN 33 35 HEWLETT PACKARD 100000 AER ees 0 4 Copr 1949 1998 Hewlett Packard ace not forall conditio for orifice diameters above 3 5 Cardiac Output m s A good summary of both the 8 Most ninvas cedures cannot accurately assess applications of Doppler ultrasound the function of the Cardiac lure can cause stroke tloading yet this overload can also occur found in reference 4 In addition nur failure Diuretics can remove all symptoms of been done to validate the accuracy of car sure calculations with positive results Fig 3 illustrates what Dor pulsed mode In this figure the two dimensiona l cardiac perfor without improv nance For these and other reasons it is important to be pler display able to measure card
96. nals allow flow direction to be determined The rotational filter accepts as inputs the in phase and quadrature signals from the multipliers decodes the sig nals and outputs forward and reverse flow values The first step consists of delaying the quadrature signal by 90 degrees relative to the in phase signal These new signals in phase and quadrature 90 are such that the forward flow information is in phase 0 phase difference for the two channels and the reverse flow information is 180 de grees out of phase When the two channels are summed the forward flow values add whereas reverse flow data is canceled On the other hand by taking the difference of the two channels forward flow is eliminated and only reverse flow is left Mathematically this can be seen by referring to the equa 40 HEWLETT PACKARD JOURNAL JUNE 1986 tions for the outputs of the quadrature sampler equations 2 and 3 Rewriting equations 2 and 3 to account for the 90 degree phase delay 01 907 wrwat COS wryt 4 Taking the sum and difference of equations 2 and 4 Sum I Q 90 2 cos wiwat 5 Difference I Q 90 2 cos wayt 6 Thus the sum channel contains the forward flow and the difference channel contains reverse flow These signals are then sent to an audio driver card which drives a set of stereo speakers on the sides of the ultrasound system cart and also an optional headset The biggest design c
97. nges in the design rules of a process If a cell is laid out graphically and a process change occurs affecting the minimum separation the designer must graphically move the rectangles to achieve the new minimum separation In most cases there is not enough room to move a rectangle to achieve a design rule change without hitting a new re ctangle and consequently having to move it and its neigh bors These changes can propagate outward until every thing in the cell must be moved Because of their resistance to change graphic layouts tend to be discarded when an IC fabrication process changes thereby eliminating the ability to reuse previous design work Although it may take a little longer to design a cell programmatically using ICPL such an effort is greatly rewarded by the ease with which sign changes can be achieved and by the ability to use a successful design for other projects ICPL incorporates a design management system to pr vide an environment in which rectangles can be placed wires can be easily run and a library of hierarchically organized cells can be maintained Connectivity informa ng ie stated regarding what he rerface that 18 locked Me bis Falon ching H oeo THING of Type REPLICATION JFtntance Timez of ore REPLICATE 95 rasene 1 eran es erent vr Fy tion is maintained as it is declared by th
98. nstruments played positive frequencies in one channel of a stereo pair of speakers and negative frequen cies in the other Since echoes returning from red blood cells come from the entire volume contained in the trans ducer beam and not a single point there is in fact a spec trum of Doppler shift frequencies returning to the trans ducer As it turns out the human ear is remarkably adept at recognizing frequencies and bandwidths of signals in the presence of wideband noise These systems were clin ically very successful once operators were trained enough to recognize the normal Doppler shift sounds from abnor mal flow patterns The next enhancement to the CW Doppler system was the addition of a crude spectrum analyzer The first spectral displays were time interval histograms made by measuring the times between zero crossings of the Doppler shift waveform The zero crossing times or time intervals are then lumped into bins and displayed as a histogram that gets updated many times in a heart cycle These histograms served as a crude version of a real time spectrogram Before long fast Fourier transform FFT and chirped z transform circuits were feasible that could calculate 128 point spectra in a few milliseconds and these replaced the less accurate time interval histograms Pulsed Doppler Measurements The major limitation of the continuous wave instruments is that they are sensitive to motion in particular blood flow along the en
99. nted for For example one can estimate the area for a given block by using area information that exists for a similar block in a different process technology if the two technologies can be mapped from one to another by a numerical transform Finally when these mechanisms fail 10 return a result the system has two further options If the block can be further decomposed into simpler blocks the system will do so and at tempt to apply the area estimation mechanisms on the compo nent blocks recursively If the block cannot be further decom posed i e it is leaf cell then the system will either calculate the area based upon gate count or prompt the user fora value the layout based upon specified parameters that account for connectivity requirements power bus sizing output load and aspect ratio Several features have been added to ICPL to aid the cre ation of just such a large procedural cell library An inte grated avoidance capability allows wires to be routed and parts to be placed without regard to specific design rules Connectivity information is maintained in the same data structures as the layout This allows detection of such de sign errors as shorting two global nets together and neglect ing to wire a particular cell Integrated connectivity means that parts have net attributes just like size and location attributes and it means that all of the parts tied to a net can be easily found for capacitance extraction or design v
100. o vessel of known area and multiplying this stroke volume by heart rate The primary limiting factor has been shown to be the inability to obtain an accurate flow cross sectional area Other limiting factors include accurate determination of the Doppler angle to flow velocity and the assumption that laminar flow exists This means that the absolute value for cardiac output cannot be obtained near a defective valve In adults this limitation seriously decreases the use fulness of Doppler cardiac output measurements for initial diagnosis Doppler cardiac output is still a viable alterna tive in tracking a patient s response to treatment however because there are sites like the descending aorta where consistent measurements can be made reliably in almost all patients Some published comments about the use of Doppler techniques for evaluating cardiac output are Because it is noninvasive Doppler ultrasound measure ment of cardiac output is ideal for repeated use in outpa tients for monitoring progress and Doppler ultrasound has been used for assessing the hemodynamic response to vaso dilators in patients with heart failure These findings suggest that aortic blood flow measure ments by Doppler velocimetry are useful clinically for the noninvasive evaluation of acute drug therapy in patients with heart 19 Congenital Heart Disease Congenital heart disease encompasses many of the same kinds of dysfunct
101. ocity of the scatterer What is then measured as the Doppler shift frequency for a single scattering site with a constant and uniform velocity is the summation of the Doppler har monics weighted by the scaled version of the Gaussian bandwidth of the transducer and electronics or V2 D Kx Gaus 92 f K a tf nK PRI where 2V cos 6 C f4 f This leads to two interest ing observations First if a narrow band of Doppler frequen cies is desired to represent only a narrow range of scatterer velocities then it will be necessary to narrow the band width of the transmitted pulse This can be done either by increasing the number of transmitted cycles or by filtering the input to the Doppler detector In either case the range resolution is degraded Another way to look at this position velocity uncertainty trade off is to picture each scattering site remaining in the ultrasonic sample volume for a finite length of time The sample volume is most sensitive to the scattering site near its center and the sensitivity tapers off near the edges As a result the Doppler shift waveform has a low amplitude as the scattering site enters the sample volume the amplitude increases as the site moves to the center of the volume and as the particle passes through to the other side of the volume the amplitude again drops to zero This behavior in effect windows the time domain waveform by convolving the sample volume with the impulse from
102. on at Stanford Uni versity Both her BS and MS degrees in industrial en gineering were completed in 1978 With HP since 1978 she is a software en gineer and has worked on the development enhancement and support of HP Materials Management 3000 She is currently a member of the team that developed HP JIT lives in Sunnyvale California with her husband and enjoys watching and participating in ballet and dance Robert A Passell Bom in Milwaukee Wiscon sin Bob Passell attended the University of Wisconsin at Madison from which he received a BA degree in English in 1975 and an MBA degree 1978 He was also a budget and management analyst for 47 the University and studied computer science until joining HP in 1981 He was firsta programmer anaiyst and then a software en gineer and has contributed to the development of HP Maintenance Management 3000 and HP JIT interested in computer based management information systems tor manufacturing environ ments A resident of San Jose California Bob is married and has two children He likes spending time with his family motorcycle touring fishing hunting and cultivating roses Teri L Lombardi Teri Lombardi is a 1982 graduate of California Polytechnic State Univer sity at San Luis Obispo from which she received a BS degree in computer sci ence After coming toHP in 1983 she worked on HP Materials Management 3000 and HP JIT and has since l
103. on only by rough sketches and verbal communication Area Estimation Expert The back end area estimation expert provides feedback that can be used recursively in design decision making We chose Copr 1949 1998 Hewlett Packard Co area estimation over several other candidate applications be cause it is a hard physical constraint for which people usually fely on the most experienced designers to provide answers We also believe that area estimation should be incorporated into the early design planning phases because if not done right it causes expensive global changes in floor planning and is very time con suming to fix after detailed layout has been done Block area can be estimated using information from a few different knowledge sources First there are hardcoded numbers for information that has not yet been formalized into any other categories Next there are block sizes derived from block boun daries of actual cell libraries the system has access to Third there are routines that describe area characteristics for highly flexible parameterized cells such as module generators These functions take the same set of parameters but instead of generat ing the entire physical layout they only return a value for the estimated area and aspect ratio for the block Fourth there is the option if similar designs exist of using those designs as a basis for information provided the relevant differences between designs are accou
104. or IDAT FFT To Audio Card Fig 1 Doppler detector card DDC block diagram Fig 1 extracts the Doppler shifted echoes from the body These echoes are received by one of two probes the 64 ele ment phased array transducer which for the confines of this article is used for pulse Doppler mode only or a sepa rate probe for continuous wave CW Doppler mode In the first case an intermediate frequency signal is supplied by the scanner In the other case a pair of baseband quadrature signals is supplied by a separate CW card In both cases Doppler information is extracted from the input with much signal processing to allow outputs of forward and reverse flow signals for stereo audio speakers and a digitized com plex Doppler flow signal for digital processing IF Filters The front end filter on the DDC is referred to as the IF filter The intermediate frequency nature of the signal is a result of a heterodyning technique used by the scanner to steer and focus the ultrasound beam Sensitivity is crit ical to Doppler information extraction which requires the front end IF filters to be matched to the input signal A matched filter has a frequency response equal to the fre quency content of the input signal Such a filter maximizes the ratio of peak signal power to rms noise power This enables very low level blood echoes to be discriminated from the noise floor of the front end The
105. ques as a screening procedure and as an alternative to the catheteri zation lab in cases of suspected cardiovascular disease has become a widely accepted practice because of its low risk relatively low cost and ease in obtaining critical clinical data Every time a Doppler procedure can be performed in place of a catheterization procedure the patient not only has reduced risk but also considerably less pain and ex pense approximately one fifth of the catheterization cost References 1 RL Popp M D A Physician s View of Echocardiographic Imaging Hewlett Packard Journal Vol 34 no 10 October 1983 pp 13 16 2 RL Popp M D et al Optimal Resources for Ultrasonic Examination of the Heart Circulation Vol 65 2 1982 pp 423 431 3 Holen et al Determination of Pressure Gradient in Mitral Stenosis with a Non invasive Ultrasound Doppler Technique Acta Medica Scandinavica Vol 199 1976 pp 455 460 4 L Hatle and Angelsen Doppler Ultrasound in Cardiology Lea and Febiger Philadelphia 1982 5 J A Requarth et al In Vitro Verification of Doppler Prediction of Transvalve Pressure Gradient and Orifice Area in Stenosis The American Journal of Cardiology Vol 53 May 1 1984 pp 1369 1373 6 Adhar M D and N C Nanda M D Doppler Echocar diography Part fay Valvular Heart Disease Echocardiog raphy Vol 1 no 2 1984 pp 219 241 7 L
106. r The other two positions scale transmitted power down by 3 and 7 dB To resolve TGC related issues registers are added to the TGC card to offset the TGC gain selected by the operator automatically Separate registers are included for imaging and Doppler offsets The Doppler offset is always set higher by the software to allow Doppler mode to use all of the front end s dynamic range Whenever a new transmit maximum is set as a result of a mode change the TGC offsets are modified to compensate for any changes in trans mit gain Hence the round trip gain remains the same and the image maintains its current intensity Thus optimized images do not change when control modifications are made Acknowledgments Many people contributed to the development of the HP 77200B Scanner Paul Hempstead designed the interrupt driven scanner software Tom Hunt Jacques Desjardins and Jim Mniece designed the coefficient memory TGC and timer cards respectively Ron Gatzke was responsible for project management of the scanner development References 1 Hewlett Packard Journal December 1983 pp 3 28 2 Hewlett Packard Journal October 1983 entire issue 3 T L Szabo and Seavey Radiated Power Characteristics of Diagnostic Ultrasound Transducers ibid pp 26 29 Copr 1949 1998 Hewlett Packard Co Digital Processing Chain for a Doppler Ultrasound Subsystem by Barry F Hunt Steven C Leavitt and David C Hempst
107. re the methods used by the software designers to develop a package with no examples or standards to use for guidance Working with HP manufacturing divisions starting up just in time facilities the software designers tried out ideas using throwaway prototypes before finally writing specifica tions and code The remaining six articles in this issue discuss a Doppler enhancement for HP s medical ultrasound imaging system first featured in these pages in October and December of 1983 In ultrasound imaging high frequency sound waves are bounced off structures inside the body letting physicians see the body s inner workings without surgery or health risk to the patient The Doppler system augments the normal ultrasound image by adding information about the velocity of blood flow and organ movements On page 20 Ray O Connell discusses the role of Doppler ultrasound in cardiac diagnosis On page 26 Paul Magnin presents Doppler history and theory and tells us a little about Johann Doppler page 27 The articles on pages 35 41 and 45 describe the design of the HP Doppler system which can be added easily to any HP ultrasound imaging system On page 31 James Chen tells how ultrasound power and intensity are measured to ensure patient safety Dolan Editor What s Ahead Next month s issue will describe the design of HP s portable computer family The Portable and Portable Plus Memory management low power operation I O and data commun
108. re were six simulators and pro totypes as follows Screen simulator I 3 weeks Test high level product definition 8 HP 150 prototype of HP 3000 link 4 weeks Test tech nology of HP 150 link HP JIT functional prototype 3 weeks Test design of data base and critical transactions performance Screen simulator II 3 weeks Test functional design Screen simulator 3 weeks Test user interface Screen simulator IV 3 weeks Refinement At the end of this process a concise document was written to confirm the consensus of the customer marketing and lab input we had received on the prototype models of the product Instead of the traditional approach where a docu ment is produced for review and revision before the prod uct feature set is determined the document produced for our project was effectively a summary of what the project team had agreed upon based on the results of an iterative prototyping process To summarize in the traditional method documents were an ineffective method of communication People often did not review documents carefully It is difficult to vis ualize a software system solely through the use of a docu ment There could be long periods of time during the draft ing of the document when no progress on the project was visible A lengthy review was required to get agreement on product specifications and there was little involvement outside of the development lab before the documen
109. rea This can become not only time consuming but tedious The next step in the evolution of these Doppler blood flow instruments is to create a flow image by simultaneously processing the Doppler shifts from every point in the image and coding the blood flow velocities in color This will provide a visual impression of the flow at every point in the image in real time References 1 Magnin Coherent Speckle in Ultrasound Images Hew lett Packard Journal Vol 34 no 10 October 1983 pp 39 40 2 8 Satomura and 2 Study Examining the Heart with Ultrasonics Japanese Circulation Journal Vol 20 1956 p 227 3 D W Baker and D W Watkins Phase Coherent Pulsed pler System Cardiovascular Measurement Proceedings 20th ACEMB Vol 27 1967 p 2 Bibliography P Atkinson and J P Woodcock Doppler Ultrasound and its Use in Clinical Measurement Academic Press New York 1982 D W Baker Forster and R E Daigle Doppler Principles and Techniques Ultrasound Its Application in Medicine and Biology edited Fry Vol 1 chapter 3 Elsevier Press 1978 Power and Intensity Measurements for Ultrasonic Doppler Imaging Systems by James Chen echoes from red blood cells These echoes are two to three orders of magnitude weaker than the echoes reflected from heart wall and body tissue Even though those echoes have very narrow bandwidth the signal to noise ratio
110. rehouse An X and Y are released to work order WO1 which uses these parts to Receive Post Deduct Fig 3 Demand pull stock flow Just in time manufacturing may also eliminate incoming inspection so that parts flow directly from receiving to production create subassembly B which is returned to inventory Subassem bly B and part C are then released to work order WO2 which uses these parts to create the finished product A whichis placed in finished goods inventory Fig 3 shows demand pull stock flow Parts X Y and are received inspected and delivered to their stock locations An X and a Y are used in the creation of subassembly B which then travels down the production line to the point where part C is added resulting in the creation of product A Raw material and finished goods inventory adjustments are made at this time by means of the post deduct transaction Besides eliminating the work orders and subassembly storage of traditional batch stock flow demand pull stock makes it possible to introduce further improvements that are now consid Fabricated Parts ered characteristic of the just in time manufacturing method i Purchased Parts X Y C These are the elimination of incoming inspection and the flow of parts directly trom receiving to production Fig 1 Assembly sequence for a product Warehouse Receive Receive Issue Receive Issue Receive Fig 2 Tradition
111. ricke Shaw Yang Duksoon Kay and Rick Brown who have beautifully demonstrated the feasibility of using Lisp based design tools in building pro duction chips References 1 J Batali and A Hartheimer Design Procedure Language Manual Al Memo 598 VLSI Memo 80 31 Massachusetts Insti tute of Technology Artificial Intelligence Laboratory September 1980 2 Infante Bales and Lock MADL A Language for Describing Mixed Behavior and Structure Proceedings of the IFIP Sixth International Symposium on Computer Hardware De scription Languages and their Applications May 1983 Copr 1949 1998 Hewlett Packard Co New Methods for Software Development System for Just in Time Manufacturing New approaches in prototyping next bench involvement performance modeling and project management created a high quality software product in the absence of standards or existing systems by Raj Bhargava L Lombardi Alvina Y Nishimoto and Robert A Passell planning and control of material and production for manufacturing facilities operating with a new manufacturing technique developed in Japan and known as just in time Just in time manufacturing reduces com plexity on the factory floor by using fixed production rout ings and a pull system for material handling In a pull sys tem raw materials are delivered to the factory floor as they are consumed in production This simplifies manu
112. rk City and received aBS degree in chemical and biomedical engineer ing from Columbia University in 1977 He earned his MSEE degree from the University of Cincinnati 1979 Les and his wile and two children live in Malden Massachusetts and he s a youth group ad visor He plays softball and is a photographer and avid handball player He s currently landscaping his back yard and says he has plenty of spare rock f anyone is interested Karl Thiele Karl Thiele was born in Los Angeles California and tended Rensselaer Polytechnic institute com pleting work for his BS gree in biomedical en gineering in 1982 and for his MSEE degree in 1983 After coming to HP the same year he helped velop the detector subsystem for the 77410 Doppler Processor and is now working on the theoretical aspects of color flow mapping His other prolessional experience involved work at on RISC architecture He is interested in communica tion theory and analog circuit design resident ot Stoneham Massachusetts Karl enjoys night life in Boston and investing in the stock market He s in volved in many sports including racquetball softball volleyball weight lifting and triathlon events 41 Imaging Scanner Modification Sydney M Karp Born in Queens New York Syd Karp is graduate of Rensselaer Polytechnic In stitute BSEE 1976 With HP since 1979 he has con tributed to the design of the transmitter and proc
113. rward and reverse flow For forward flow the quadrature signal leads the in phase signal by 90 degrees whereas for reverse flow the quadrature signal lags the in phase signal by the same amount This quadrature scheme works under the critical premise that the Doppler shifts are small compared to the reference frequency in this case the IF We know this to be true since Doppler shifts are audible 0 to 20 kHz whereas the reference fre quencies vary from 1 2 to 2 5 MHz Mathematically the filtered outputs of the two samplers can be approximated by In phase wat avt 2 An important part of the Doppler system is the continuous wave CW Doppler board Fig 1 This board incorporates a 10 MHz crystal oscillator which is divided down into two 2 5 MHz quad rature clock signals These signals 90 degrees out of phase each drive the high level LO input of a mixer One of these signals is also fed into the transmitter section of the board The transmitter buffers the signal and sends out 300 matched signal to the split crystal CW probe emitting a continuous series of ultrasonic pulses The received echo is amplified and buffered to drive the RF input of the mixers The result of the mixing gives two quad rature baseband signals These baseband signals fed into Continuous Wave Doppler Board Quadrature sin wfwat sin Oret 3 where wwa and the forward and reverse Doppler shi
114. ry to be more space efficient and provides designers with greater flexibility in selecting the precise cell configurations they want An example of a much larger ICPL module generator is the Ramgen project at Cupertino Integrated Circuits Division Ramgen is a program that captures optimizes and generates a general purpose static RAM see cover The modules generated are designed to be equivalent to off the shelf RAMs that can be easily inserted into a stan dard cell or custom chip Organization and size of the RAM are user programmable by specifying the total RAM size desired the word size and the aspect ratio This will create a RAM that will fit not only a variety of functional require ments but also a variety of physical ones These features allow a chip designer to design a RAM at a high level with the fast design and verify turnaround time of approxi mately ten CPU minutes on an HP 9000 Model 236 Com puter instead of five months handcrafting time for a 256 bit to 16K bit RAM Beyond the time improvement the user is also guaranteed a working Ramgen has been used for over a year in the design of nine chips One of the most comprehensive module generators under development at HP is the Datapath generator Datapath enables a complete data path system to be generated from a functional description Input to the Datapath module generator is a MADL Multilevel Architecture Description Language behavioral description o
115. s bits to determine the line type the line number to be fired and all the control bits necessary for that line Once scanning begins the software need only read each table entry perform a few tests and load the control words into a control register Upon a mode change the entries in the line table are recalculated Power Limiting Software The addition of Doppler capabilities to the HP 77020A Imaging System intensified the need to limit the trans ducer s acoustic power output by carefully controlling its Doppler Subsystem 60 and Start of Line Signals Fig 3 Block diagram of ultra sound imaging scanner with Dop pler mode JUNE 1986 HEWLETT PACKARD JOURNAL 43 Copr 1949 1998 Hewlett Packard Co drive voltage Firing lines at high repetition rates through the singular angle of the Doppler sample volume greatly increases the spatial peak power averaged over time 1 Software was developed to limit the output power to clin ically acceptable levels A register is added to the power supply card to scale all transmitter settings as necessary Software calculates the maximum possible Ispra by exam ining the scan sequence The number of pulses fired through the sample volume is divided by the period of the entire sequence The transducer acoustic efficiency and focal quality are also factored in Focal quality is a function of depth depending on where the elevation focus is fix
116. s without aliasing Large software overhead times added to the acoustic line time could not be allowed In addition the Doppler sample rate cannot tolerate any interruptions for overhead processing between frames To make the most efficient use of the time available and to ensure that enough time is put aside for line control bit generation a priority interrupt scheme is implemented The highest level interrupt is programmed to be generated by the timer card This interrupt directs the MC 5 processor to assemble the control bits for the next line The remaining time for each acoustic line is available for background pro cessing Lower level interrupts can interrupt the back ground processing to service a new HP IB command or to respond to the 68000 coefficient processor In earlier scanners a great deal of processing time was spent determining the next line type in a sequence Sector Doppler and M mode line types need to be fired in different sequences depending on parameters such as sector size or nonimaging modes being enabled Once the line type was determined its line number and control bits had to be found Sector line numbers had to be counted as scanning proceeded across the sector To reduce the processing time during scanning a large RAM line table is added to the processor card Before scan ning the software fills this table in accordance with the mode selected The table contains an entire frame of entries Each entry contain
117. see Fig 1 each of which addresses a different requirement of the just in time manufacturing method Parts and Bills of Material Rate Based Master Production Scheduling JIT Material Requirements Planning Production Reporting and Post Deduct Stock Areas and Deduct Lists Inventory Management Material Cost Reporting Parts and Bills of Material maintains basic information on every part used in production recording product struc ture standard costs and data on product options and en gineering changes Its features include single level manu facturing bills of material effectivity dates for engineering changes pseudoparent parts to represent subassemblies consumed in production multiple product options and ABC part classification Rate Based Master Production Scheduling matches proposed production and shipment schedules against back log and forecast orders to manage finished goods inventory more effectively Planning is accomplished without work orders with monthly production of each end product based upon a planned rate of output for that product per day Schedules for different production lines producing the same product can be adjusted separately The features of this module include a predefined VisiCalc spreadsheet what if capability for production planning graphics for reporting of monthly plans automatic data transfer be tween the HP 150 Personal Computer and the HP 3000 Computer an inter
118. ss providing focus and great satisfac tion to the team as we met each milestone We emphasized leveraging existing resources and avoid ed duplication of effort At each step we tried to direct our efforts to activities that could be used in the final product package For example the external specifications became the starting point for the user manual We always went through iterations before making decisions and starting to build However once a component of the product had been built or a specification defined we resisted augmentation and rework Early in the project all dependencies were defined and contingency plans were developed We recog nized that some dependencies were absolutely essential and no backup was available However we made it an objective to minimize dependencies on the critical path The three most important overall goals for the HP JIT project were To invent a high quality software product where there was none before either internally or externally To meet all scheduled milestones the most important being the MR manufacturing release date which was stated at the beginning of the project To increase the productivity of engineers in designing this type of software product The design of the product was influenced from its earliest stages by groups that would have a major influence on its ultimate success JIT practioners and HP product market ing The process of designing functionality and quality into
119. t review cycle was completed This brought considerable risk to the process of developing the product and made it difficult to measure whether the project was on track or not In contrast the HP JIT project used prototypes to provide a tangible display of product functions and features during the design phase of the project This proved to be an ex tremely effective method for communication and elicited more thorough feedback from potential users who could visualize the final product more fully based on the initial prototypes The many iterations of the product prototypes each one progressively refining the feature set of the prod uct provided the product with an opportunity to establish many short milestones and made progress during the de velopment phase more visible and measurable The in volvement of end users and marketing during the early phase of product design decreased the risk of redesign and reengineering later on in the project Next Bench Involvement HP is widely recognized as an industry leader in imple menting the just in time production method A primary objective of the HP JIT project was to work closely with Select Criteria for MRP Run SELECT MRP Ending date for MRP daily report mmdd Run Y N From To Controller ID Selection Range From To Other Criteria Selection Range From amy To Fig 3
120. t than when it is not A balance can be used to read out this difference which is directly related to the total acoustic power Ideally if pure water is used as the propagation medium one watt of acoustic power generates a weight loss of 68 mg which can be derived as follows where Py the total power in watts the mass in kilograms the ultrasonic velocity in meters per second and a the acceleration in meters per second squared P C ma Therefore and for Py 1 watt 1 9 8 1 494 6 83 107 kg 68 3 mg The accuracy of a force balance can be affected by several factors such as absorption by the target standing wave formation and convection current reduction To ensure their accuracy our force balances are calibrated with the use of standard quartz transducers The quartz transducers are calibrated for total acoustic power at the U S A National Bureau of Standards to within 3 accuracy To calibrate the force balance the standard transducer is positioned at the force balance window Since it emits a pressure wave of known acoustic power the weight loss reading from the balance divided by the known total acoustic power gives the calibration coefficient in mg watt The force balance provides a convenient means to mea sure total acoustic power from an ultrasonic device How ever prudence must be exercised during the testing to en sure that the measurements are meaningful For example
121. tage is that it is time con suming to carry out Since most hydrophones have a rather simple frequency response hardware can easily be im plemented to level the frequency response to within the accuracy needed for intensity measurement The method used to find the frequency response of the measuring system is to repeat the sensitivity calibration at the required frequencies This is very time consuming to carry out based on the method previously described How ever the sensivity calibration procedure can be modified to reduce the data collection and data reduction time sig nificantly Broadband Calibration Method The previously described calibration methods can be simplified based on two facts First the constant of relative intensity profile integration C should vary only with fre quency transducer aperture and axial distance Second a low frequency source transducer can be used to generate the acoustic field at the resonance frequency and at higher harmonics The acoustic field at distance z in front of a transducer is uniquely defined by the driving frequency and aperture size It can easily be derived that the relative intensity integral is a constant for a quartz transducer be cause of its frequency stability and fixed aperture In other words that constant should be just as stable as the radiation conductance which is a function of electrical mechanical energy transfer ratio Furthermore both constants be pre
122. the generation of an imaging sector and Doppler outputs simultaneously Thus the position of the Doppler sample volume within the body can be easily determined The disadvantage is that the Doppler sample rate is fairly low about 2 kHz This limits the maximum resolvable flow velocity Fig 2 a Phased array radiating a plane wave at an angle 8 b By on TTT TTT proper selection of the delay for Element Delays _ each transmitting element a curved Element Delays wavefront can be generated that is a b convergent desired location P 42 HEWLETT PACKARD JOURNAL JUNE 1986 Copr 1949 1998 Hewlett Packard Co In Doppler only mode the image is first frozen and all lines are then devoted to Doppler shift detection Doppler sample rates are as high as acoustic limitations allow Each Doppler line can be terminated as soon as the transmit pulse has propagated to the sample volume and been re flected back to the transducer This allows the determina tion of higher flow velocities than possible using duplex mode In triggered Doppler mode the scanner fires only Doppler lines until a trigger is received at a selected point on the associated ECG trace The scanner then interrupts the Doppler outputs to update the frozen image sector This mode gives the user a compromise combination of position determination with high velocity flow resolvability
123. the guidelines very difficult In some cases this would yield misleading data There are two types of exposure parameters that must be measured They are the total acoustic power and the intensity values measured at a particular point in the acous tic field The former is a measure of the total acoustic energy delivered by a transducer per unit time The latter is the amount of energy delivered to a unit area per unit time Total power depends on how much electric energy is input to the transducer and how the transducer transforms it into acoustic mechanical energy The intensity depends on the electric energy into the transducer the total acoustic energy out and how the energy is focused to a particular spot The total power is measured with a force balance and intensities are measured with a hydrophone Total Power Measurement with a Force Balance Fig 1 schematically describes the concept of a radiation pressure measurement Basically when a sound wave propagates through the liquid it generates a dc pressure at the target as a result of the fluid density variation during pressure wave propagation This is the radiation pressure JUNE 1986 HEWLETT PACKARD JOURNAL 31 Copr 1949 1998 Hewlett Packard Co associated with the pressure wave If a target is large enough it receives an uplifting force which comes from the entire radiation pressure field This force causes the target to weigh less when the pressure wave is presen
124. the prod uct One post deduct transaction can result in up to 400 separate data base updates According to the model the maximum production rate or the rate that HP JIT can han dle with real time updates of component part consumption is given by Maximum Production Rate 3600 0 75 0 25 1 5D L E L where D is the total number of deduct points P is the total number of parents L is the number of production lines and E is the total number of deduct list elements This model showed that for a low volume customer even the best case would cause this transaction to take 2 7 hours to process one hour of customer data This meant that even for a low volume customer the post deduct transaction would never catch up with incoming data that it was in tended to process To solve this problem we designed this transaction to summarize the processing of the data allowing the cus tomer to specify the summarization interval The model allowed us to provide a formula to help customers decide what summarization interval would fit their type of data and rate of production The formula is Summarization Interval 1 25DP L 0 75P E 900R 1 0 5DR L where D P L and E are as above and R is the total produc tion rate per second Quality Procedures Quality was built into HP JIT using a variety of tech niques In the lab formal design reviews and code inspec tions were used throughout the development process and this contrib
125. the simple relationship Ispra Vims K Z 7 With this rms meter the position of maximum intensity at a particular system setting can be found by moving the hydrophone in the acoustic field and reading the rms volt age output from the hydrophone system settings that can possibly give maximum Ispra are checked with the rms meter connected to the hydrophone This allows us to scan along the acoustic field easily until we locate the point for maximum Ispra Copr 1949 1998 Hewlett Packard Co Extraction of Blood Flow Information Using Doppler Shifted Ultrasound by Leslie Halberg and E Thiele tion of Doppler frequency shifts from the transmit ted ultrasound frequency allows clinicians to mea sure blood flow velocity from many vessels and heart cham bers in the human body HP s Doppler instrument the HP 77410A is inserted functionally between the scanning and display subsystems of the HP 77020A Phased Array Medical Ultrasound Imaging System The HP 77410A is made up of four state of the art printed circuit cards and a mother board These cards are the Doppler detector card DDC the FFT fast Fourier transform card the data output card and the processor card U LTRASOUND IMAGING in conjunction with detec Doppler Detector Card The HP 77410A s front end the Doppler detector card IF trom Scanner Pulse Doppler Mode Sample Gate CW Doppler from CW Card From Microprocess
126. tire length of the ultrasonic beam This 28 HEWLETT PACKARD JOURNAL JUNE 1986 Fig 2 Effect of a moving source wavelength and hence fre quency of the emitted sound wave The outermost circular wavefront was emitted when the source was at position 1 and the innermost circular wave front was emitted at position 2 The wave fronts are circular because the sound propagates in all directions at the same velocity after it is emit ted by the source Thus a station ary listener on the left hears fre quency lower than the source fre quency a listener on the right hears a higher frequency than the source can lead to ambiguities since the position of any aberrant flow cannot be localized in range To overcome this prob lem the first pulsed Doppler system was developed Pulsed Doppler systems transmit a short sinusoidal burst instead of a continuous sinusoidal wave By gating the received signals to correspond to the pulse s time of flight to the point of interest one can interrogate a small sample volume instead of the entire length of a beam Since these pulsed Doppler instruments require sampling the returning echoes at a fixed time after the burst is transmitted it is possible for the higher frequency Doppler shifts to alias to lower frequencies The pulsed Doppler system and the aliasing can be more easily understood if one considers the echoes returning from a moving wall as shown in Fig 3a Samples of the
127. to the Nyquist rate in each channel Frequencies above the Nyquist rate are meaningless and therefore must be re moved Thus a low pass Nyquist filter is needed that will pass the desired Doppler signals and attenuate the un wanted byproducts of sampling which are shown in Fig 4c The biggest problem in implementing such a filter is that the Nyquist frequency is equal to half the pulse repe tition rate which is dependent on the depth of the flow being interrogated Therefore a low pass filter with a vari able cutoff is required Ideal for such an application is a highly integrated switched capacitor filter commercially available at a very nominal cost A microprocessor programmed clock that is phase locked to the PRF controls the time constant of two noninverting integrators that filter the in phase and quadra ture signals from the wall filters The relationship between the clock and the cutoff frequencies is foto PRF 2 fetock 100 Output of In Phase Sampler Fig 5 Top Echoes for five PRF intervals containing both forward and reverse motion Doppler shifts Center Output of single in phase sampler for both echoes 38 HEWLETT PACKARD JOURNAL JUNE 1986 Bottom Output of second quad rature sampler allows differentia tion of forward and reverse flows by phase encoding Copr 1949 1998 Hewlett Packard Co The cutoff characteristics can be easily controlled by selecting
128. to the partner divisions throughout the design and development stages kept the testing effort progressing actively and kept the partners from feeling overwhelmed by bulky new ver sions of the product with large doses of new functionality Each of the partner divisions was assigned a member of the HP JIT lab team for support providing quick feedback and fostering closer ties throughout the entire development team Project Management One of the essential ingredients that made HP JIT a suc cess was the innovative project management We recog nized the need to make quick decisions as a team At times this meant taking action even when we did not possess complete or adequate information However we were will ing to take risks and accept the fact that we would learn along the way We expected that as we gained more experi ence our decisions would have to change We did not resist change but managed it so we could meet our milestones and objectives Transaction Code Utility Code Transaction Code Utility Code b Fig 6 Traditional software coding b HP JIT coding built the product on a layer of utility procedures We planned milestones in 4 to 6 week intervals These milestones were measureable and very visible to people involved in the project The lab team then viewed these as hard milestones and every effort was made to meet them With these frequent milestones it became much easier to measure our progre
129. tory consumed in assembly JUNE 1986 HEWLETT PACKARD JOURNAL 15 Copr 1949 1998 Hewlett Packard Co find_datasetrec multi_find_dataset Locate an entry or entries that match a set of criteria parameters in the specified data base Doall screen handling for the HP JIT maintainance transactions Check user access to screen Determine that key input fields on screen are valid for the transaction Opena specified MPE message file Create it if it doesn t exist process_maint valid_in_security valid_input open_message_file lock_datasets Lock the specified data items or sets specified by key parameters input_required Verify that a list of fields has been input Performance Modeling In the HP JIT project we decided to try a new approach to the performance testing of our application projects The old method collected a lot of data with little idea of the questions being answered In addition performance tests tended to be done after the product was coded Any bad news was received too late when it was costly to redesign and recode The HP JIT project established performance objectives in customer terms before any coding or design had been done After we first modeled performance we continued to refine the model based on actual performance test results Our performance model uncovered a major performance problem with our most critical transaction post deduct before we completed designing or began coding
130. tts and have anew baby He likes sports car rallies softball and writing real time 3 0 graphics software He also writes software for the PC in his church Steve Leavitt is a digital sig nal processing specialist who has been with HP s An dover Division since 1978 He was in Portland Maine served in the U S Air Force and graduated fromthe University of Maine with a BSEE degree in 1971 He worked on digital signal processing for the HP 77020A Ultrasound imaging System including the FFT Doppler pro cessor the color flow processor and R Theta scan conversion He is named coinventor on two patents related to the R Theta algorithm developed for the HP 77020A and is coauthor of a symposium paper on the same subject Before coming to HP he de signed communication systems Steve and his wite live in Hampstead New Hampshire his leisure time he enjoys mountaineering cross country ski ing snowshoeing and bicycling Copr 1949 1998 Hewlett Packard Co JUNE 1986 HEWLETT PACKARD JOURNAL 19 The Role of Doppler Ultrasound Cardiac Diagnosis In ultrasound imaging a pulse of acoustic energy is transmitted into the human body and the strengths of the returning echoes from various organs and tissues are used to form an image on a display screen Further information about blood flow and movement can be gained by measuring the shifts in the frequency of the echoes by Raymond G O Conne
131. tz Automating Application System Operation and Control ibid 4 J L Malin and L Bunton HP Maintenance Management A New Approach to Software Customer Solutions Hewlett Packard Journal Vol 36 no 3 March 1985 Authors June 1986 4 Allan J Kuchinsky With HP since 1978 Allan Kuchinsky manages the Lisp tools group at HP s Cupertino IC Division His contributions include work __ on HP Production Manage ment 3000 on ICPL He is acoauthor of three tech nical papers and is member of the IEEE and the American Association for Artificial Intelligence Allan was born in Brooklyn New York and studied psychology at Brooklyn College BA 1971 He taught science at a New York City secondary school and was a high schoo biology teacher in Mexico City He has also worked as a musician Be fore coming to HP he completed work for an MS degree in computer science in 1978 at the Univer sity of Arizona Now a resident of San Francisco he s married and has one son He enjoys playing his guitar listening to reggae music and spending time at music and comedy clubs He s also fond ot reading science fiction and murder mysteries An alumnus of the Univer sity of Ilinois Yip has a 1982 BSEE degree anda 1984 MSEE degree Hehas been with HP since 1984 and has enhanced and supported an HP multilevel simulator as well as work ing on ICPL and silicon compilation He is
132. uction lines per product and split merged and mixed mode production lines Inventory Management maintains the current status of each stock area and updates its status through the post de duct scrap or extra use transactions The features of this module include on line inventory balances multiple stock area status codes available unavailable scrap inspection recording of scrap and extra use two stage cycle counting and inventory count and adjust capabilities Material Cost Reporting summarizes accounting infor mation on materials produced during an accounting period This module features production data summarized by indi vidual product exception data recorded on an individual transaction basis for visibility and material variances cal culated based on scrap and extra use recorded Development Cycle The HP JIT development project was significantly differ ent from software deveploment projects that preceded it New methods for engineering this type of software were introduced to meet the objective of producing a high qual ity product ina shorter period of time than had been typical Fig 5 on page 16 shows two time lines contrasting the traditional and HP JIT development cycles The remaining sections of this article detail some of the methods used in the HP JIT project to change the typical project cycle Prototypes The HP JIT project provided an opportunity to try a dif ferent approach in the investigation and externa
133. ues discovered in those interview sessions that will contribute to the front end planning assistant as well particularly in the way this designer makes use of uncertainties at each phase and handles the iterative design of making block definition decomposition and recomposition Area estimation is more than just adding up numbers There are assumptions upon which the estimation is made Availability of library cells and module generators orientation aspect ratio and relative placement of blocks and intended signal paths are just a few examples of these assumptions These planning deci sions dictate restrictions the subsequent chip recomposition must follow and provide useful technical hints and tricks to help develop layout details Benjamin Y M Pan Michael How Development Engineers Allan J Kuchinsky Project Manager Cupertino Integrated Circuits Division be associated with a part thus enabling Lisp programs such as circuit extractors and cell compilers to store their data on the parts themselves Through properties users can ex tend ICPL to support virtually any configuration they choose For example resistance and capacitance values can be determined and tacked onto parts for use with a user developed circuit simulator Furthermore all ICPL data structures are based in Lisp thus enabling the ICPL programmer to use the full power of the Lisp programming environment ICPL is tightly inte
134. usly degrade the spectral output are prevented This last approach is implemented in HP s Doppler FFT processing chain Using the specifics of the design potential overflows can occur during the last three passes of the FFT With an understanding of the tree structure for FFT processing it is fairly apparent that overflows occurring during the ear lier passes get propagated as processing errors to many more frequency bins than do overflows in later passes JUNE 1986 HEWLETT PACKARD JOURNAL 47 Copr 1949 1998 Hewlett Packard Co Stated another way the overflow problem is less damaging in the later passes This is the key to the AGC strategy The AGC circuitry employs an overflow counter that counts the number of additions subtractions that result in a sign bit error two s complement arithmetic is used For example if two positive negative numbers get added with a negative positive resulting sign bit then an over flow has occurred The overflows are counted for each of the last three passes and stored in latches for processor readback The individual counts for each of the last four passes are latched The central processor reads the overflow latches periodically the AGC time constant and changes the input front end gain appropriately For overflows in an earlier pass the gain is decreased by k and then the latches are cleared When interrogated again if overflows have been suppressed in that pass and are showing up in th
135. uted to improving the overall functionality and correctness of the code for each module The use of utility routines to perform shared functions for separate applica tion modules contributed to quality by eliminating manual duplication of code Since these utilities are accessed by a variety of calling routines any defects or deficiencies were detected early and the utilities eventually proved to be very robust HP JIT also made use of its partner divisions within HP as alpha and beta test sites during the design and development phase of the product These partners were actively implementing just in time manufacturing tech 18 Months to 2 Years Lab Mktg gt User ES 10 9 Months to 1 Year 16 HEWLETT PACKARD JOURNAL JUNE 1986 Alpha Alpha Test Lab Test of Beta Beta Test Marketing Test Document Support Fig 5 a Traditional software de velopment cycle b The HP JIT Functionality development cycle was shorter than the traditional cycle Software prototyping replaced the usual in vestigation and external specifica tion phases Copr 1949 1998 Hewlett Packard Co niques and thus were qualified to assist in the design of the product and test its functionality Each partner was allowed to narrow its focus to a particular testing objective This made the process of testing more manageable and the testing more thorough Mini releases of HP JIT
136. we can work with in a serial fashion by dividing the time available in clocks by the number of multiplies or additions since they are done serially as well That number comes out to be 4550 128 or 35 bits Since our largest moment grows to only 29 bits we have sufficient time for performing all moment calcula tions in a serial fashion Frequency to Velocity Conversion Once the hardware completes calculating the moments the processor is interrupted In response to this interrupt the moment accumulators are read along with the 128 bins of spectral data The reason for this hardware software com bination is twofold First there is the requirement for a 29 bit by 15 bit divide While this process can be im plemented in hardware a software implementation is pre ferred because of its lower cost Second since we need only two divides for each mean variance pair there is suf ficient time available The mean calculation simply requires a single division while the standard deviation needs one division one mul tiplication a subtraction and a square root The numbers used in these calculations range in size from 15 bits for the zeroth moment up through 29 bits for the second moment The calculation of the mode takes somewhat longer in software All 128 bins of spectral data must be checked for maximum amplitude and a simple bin to bin averaging scheme is used to minimize the effect of noise spikes Sometimes an estimate of the maxi
137. wlett Packard Co Furthermore the Lisp environment is the natural founda tion for communication between CAD applications and powerful Lisp based expert systems tools Any tool that is embedded in Lisp can be invoked by any other to solve a partial problem or can be used as the top level design sys tem Thus CAD functions can be used within rules and knowledge base queries can be called from within CAD functions see box on page 8 In addition to expert system capability other advanced programming techniques such as object oriented and data driven programming are readily accessible Finally the Lisp environment provides extensibility of both the tool set and the knowledge base to drive it Extend ing programs in conventional programming languages re quires recompiling a fairly large world or providing a pro gram that covers all possibilities by run time tests This is not necessary in Lisp where functions are dynamically recompilable and relinkable HP s Integrated Circuit Procedural Language ICPL is a Lisp based language that can be used by circuit designers within HP to write programs that describe an integrated circuit s layout It is based on the Design Procedure Lan guage DPL developed at MIT s Artificial Intelligence Lab oratory and has been modified both in functionality and structure to run on HP hardware and address the needs of HP designers Fundamentally an ICPL designer writes

1986 , Volume , Issue June-1986

Contents

Download Pdf Manuals

Related Search

Related Contents