Home

pdf - Departamentul de Comunicatii

1. LII TIR RTE IN pH EL IL Fig 19 Waveform 3 a 16bit 44 1kHz PCM original signal Kaugresgs FER APE em NAT CONTE AU net Missi NUN Fig 20 Waveform 3 b 16 bit 44 1kHz ATRAC encoded signal Te 4 4 Evaluating the occurence of pre echo ATRAC utilises adaptive transform block length the main function of which is to control the occurence of pre echo Pre echo may take place during highly transient passages of the input signal due to the distribution of quantisation noise over the entire transform block 7 The audible manifestation of pre echo is quantisation noise at the onset of the transient signal involved This undesirable effect can be seen graphically using waveform analysis in the time domain The following experiment utilises time domain analysis to investigate the presence of pre echo on a highly transient signal castanet that has been encoded using ATRAC Additionally the same test has been applied to a shareware MPEG Layer III encoder at varying bit rates for comparative measures Test 3 Objective To demonstrate the occurence of pre echo as a result of encoding a highly transient signal using ATRAC and the Audio Catalyst MP3 encoder at the following bit rates 96kbit s 128kbit s and 192kbit s Contents of Waveform Audio Engineering Society AES CD Perceptual Audio Coders What to Listen For track 1 Castanets fig 21 Synopsis of results ATRAC with a bit rate of 292kbit
2. MH N2 83 A Cin Sun Carry Cout Voltage V Voltage V Voltage V Voltage Y Be 4Ons 8 ns 128ns 166ns 266ns 24005 288ns 32 ns Tine Fig 4 The logic state of the outputs Sum Cary and Cou for all possible logic combination of the inputs in the case of the 4 2 compressor presented in 6 EVI eg Ere pep pei sp oe ES p m nig ped e lis Ans B ns 128ns 168ns 266ns 240ns 28 ns 32 ns Time Fig 5 The logic state of the outputs Sum Cary and Cout for all possible logic combination of the inputs in the case of the 4 2 L compressor ais 0120 2 071 d 05m yl 128 y2 056 dy 072 SimPUST 5 i aF 1 1 1 1 Cout Se nt E ne a M RE MES Ee A Eas a E a see a gat ENEA p sede ne procnl E a E Ee A A E E E A OUEN ETES EN E LES i i h i h h 1 1 1 Cu zuo c4 cle mode a aay ea NS LIEGE TUUS wig S T P V Qu cp ur PILAM A SIS SP ED aseo Eh BIENS Dx fe pi RA dod Dr Es CSRS diet Ses el oua ate s Y i j r r i 1 Le De Ea aa fin ee PE Spat Ium oL ecuni Se ee eee feo IET pin Ee Pe ae ee eee Ee d as nc St yn eae E jw x C ERE et eter teh ret M sg aware tert Wee we men ete SE EE GA 0n 05 10 15 20 25 Time ne Fig 6 The logic state of the outputs Sum and Cout in the most
3. Fig 3 Primary Element Description 35 IV THE MULTIMEDIA APPLICATION MODEL CARACTERISTICS The general characteristics of this model consist in the possibility to reutilization of the elements developed previously the possibility for any user to introduce new functions possibility to development a old application higher level of portability In concordance with the actual needs my model present Optimizations from the structure point of view witch is open for an ulterior development I use for the primary level of structure modules For all modules is possible to define new sub modules A sub module accept functions methods and routines is possible to grow the number of that in the future The user can add new functions methods and routines Is possible to reuse some parts of an old application Object oriented implementation for structure of used objects in applications Aspect oriented implementation for the temporal aspect of multimedia objects and applications If the developers of multimedia environment use the proposal model is possible for the users to develop the functions libraries and to transfer these between applications The model propose a strong system for the quality of services A special module QoS is developed for this Multimedia Environment Software Developers Private Library Multimedia Environment Private Library Multimedia Application M Multimedia So
4. 969 978 June 1990 5 M V Wickerhauser Acoustic signal compression with wavelet packets in Wavelets A tutorial in theory and applications C K Chui editor Academic Press 1992 pp 679 700 6 R Boite M Kunt Traitement de la parole Presses Polytechniques Romandes Lausanne 1987 7 Eva Wesfreid M V Wickerhauser Etude des signaux vocaux par ondelettes de Malvar Quatorzi me colloque GRETSI Juan les Pins 1993 pp 379 382 8 T Asztalos A Isar Wavelets and Audio Data Compression IEEE International Conference on Signal Circuits and Systems SCS 99 5 7 July 1999 lasi Romania 9 A Cubitchi Metod de compresie a semnalului vocal bazat pe utilizarea func iilor wavelet Referat nr 3 n cadrul preg tirii pentru doctorat conduc tor tiin ific Prof dr ing Nafornita Timi oara 2001 46 Buletinul Stiintific al Universit tii Politehnica din Timisoara Seria ELECTRONICA si TELECOMUNICATII TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Romanian Bank Notes Recognition Using Cellular Neural Networks Corina Botoca Abstract It is presented a new method to avoid bank notes counterfeiting on color copiers This method use a sequence of analogic algorithms on cellular neural networks CNN to perform complex image processing Keys words cellular neural network analogic algorithm CNN template I INTRODUCTION A CNN is a nonlinear piecewi
5. Cit rile se fac simplu prin plasarea num rului corespunz tor 5 Nu sunt permise referin e bibliografice n notele de referin n subsol Se recomand scrierea tuturor autorilor i nu folosirea expresiei gi al ii dec t dac sunt peste 6 autori V SFATURI UTILE A Abrevieri i acronime Explicitati abrevierile i acronimele prima dat c nd ele apar n text Abrevieri precum IEEE IEE SI MKS CGS ac de i rms se consider cunoscute i nu mai trebuie explicitate Nu se recomand utilizarea abrevierilor n titluri dec t n cazul c nd sunt absolut inevitabile B Alte recomand ri Se recomand utilizarea unit ilor de m sur din sistemul interna ional Utilizarea unit ilor britanice poate fi f cut doar ca unit i secundare n parantez Se va evita combinarea unit ilor SI i CGS Nu se admit rezultate experimentale preliminare Numerotarea cu cifre romane a titlurilor de paragrafe este op ional n cazul utiliz rii acestora se vor numerota paragrafelor propriu zise si nu BIBLIOGRAFIA sau MULTUMIRI BIBLIOGRAFIE 1 A Ignea Preparation of papers for the International Sympozium Etc 98 Buletinul Universit ii Politehnica Tom 43 57 1998 Fascicola 1 1998 pp 81
6. l INTRODUCTION The skeleton procedure i e the reduction of the analyzed characters to one pixel thick lines and curves is one of the most important steps in the typed amp hand written images preprocessing This procedure has to be done in order to eliminate recognition errors caused by the dissimilar line thickness Stroke thickness is a specificity of each man s handwriting mainly caused by different pressure put on the writing instrument Even in the situations involving a single subject different writing thickness may occur in diverse situations due to the specificities of the used pen or the writing speed Line thickness can also arise from image acquisition and subsequent geometrical transformation especially scale 3 Basically the skeleton is obtained by using recursive thinning till analyzed characters are reduced to a set of one pixels thick lines and curves Subparts connectivity points and stroke s end pixels have to be preserved The straightforward approach to realize this task is employing the set of structuring elements described in 1 1 O Fa O m O 1 0 0 O m 1 Tis 1 1 1 The process is a little slow but gives adequate response in the case of simple test images as the 1 United Technologies Paris Branch ioan_snep uteh fr 2 3 Electronic and Telecommunication Faculty Communications Department Bd V Parvan Timisoara 1900 dlacrama go ro 37 one in Fig 1 1 Because
7. 0 5 cm ntre ele Pagina A4 orientat pe n l imea paginii are marginile de sus i jos de 1 78 cm iar cele din st nga i dreapta de 2 54 cm Prima pagin a articolului va avea marginea superioar de 5 cm Pentru editarea articolului se recomand utilizarea procesorului de text Microsoft Word for Windows cu caractere Times New Roman dactilografiate la un r nd Dimensiunile i stilul caracterelor sunt Titlul articolului 18 pt ngro at autorul 12 pt rezumatul 9 pt ngro at cuvintele cheie 9 pt ngro at titlu paragraf 10 pt majuscule titlu subparagraf 10 pt italic distan a de la num rul de ordine la titluri va fi de 0 25 cm textul normal 10 pt afilierea autorului 8 pt notele de subsol 8 pt legendele figurilor 8 pt i bibliografia 8 pt III FIGURI I TABELE Figurile i tabelele trebuiesc inserate n text aliniate la st nga Se recomand evitarea plas rii figurilor nainte de prima lor men ionare n text Se va folosi abrevierea Fig l chiar i la nceputul propozitiilor Ecua iile Facultatea de Electronic i Telecomunicatii Departamentul Comunica ii Bd V P rvan Timi oara 1900 e mail geo ee utt ro trebuie tip rite cu un r nd gol deasupra i dedesubt Numerotarea lor se face simplu n paranteze 1 2 3 Numerotarea va fi aliniat fa de marginea dreapt IV REFERIN E BIBLIOGRAFICE Referin ele bibliografice se numeroteaz consecutiv n forma 1 2 3
8. 1 2 A By B2 2 i e exactly the correct value from equation 11 The principle of the adder subtracter circuit is the following if A B is positive it is selected at the output and the sign bit of the result will be 0 if A B is negative it means that B A is positive and at the output the term B A negated is selected i e B A the sign bit of the result will be 1 In order to apply the method from 6 the equations 11 and 12 are rewritten as follows A B A4 Ao By Bz 1 1 14 KB A A A2 Bat Bo 1 0 15 In other words we maintain the situation of initial carry in 1 respectively that of initial carry in O at the two adders which work in parallel conditioned by the presence of a carry in equal with 1 at the last 4 2 compressor block in the case of performing a subtraction Obviously this carry in will be O in the case of addition The carry in is applied at the input Ci which is not used of the least significant 4 2 compressor Further on the length of the last block of the tree included in the ALU will be supplemented with 9 bits to the left for the concatenation of the positive exponents with the bias value of 127 included according to the modified algorithm which represent the biased integer part of the logarithm of the two operands The fractional parts of the logarithms are hidden being distributed between A4 and A respectively B4 and B of the two pairs
9. L Chua L Nemes Deblurring of images by Cellular Neural Networks with Applications to Microscopy The Third IEEE International Workshop on Cellular Neural Networks and Their Applications Roma Proceedings 199 50 3 L Nemes G T th T Roska A Radv nyi Analogic CNNAlgorithms for 3D Interpolation Aproximation and Object Rotation using Controlled Switched Templates MTA SZTAKI DNS 1994 4 Zar ndy T Roska Gy Lisyka J Hegyesi L K k Cs Rekeczky Design of Analogic Algorithms for Mammogram Analysis IEEE Third International Workshop on Cellular Neural Networks and Their Applications Roma 1994 Proceedings 5 Zar ndy F Werblin T Roska L Chua Novel Types of Analogic CNN Algorithms for Recognizing Bank notes IEEE Third International Workshop on Cellular Roma 1994 Proceedings 6 P Venianter F Werblin T Roska L Chua Analogic CNN Algorithms for Some Image Compression and Restoration Tasks UCB ERL Memo Berkeley 1994 Buletinul Stiintific al Universitatii Politehnica din Timisoara TEE ___ Seria ELECTRONICA si TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Evaluating Objective Audio Quality for Minidisc and Influence of Multiple Generations Audio Coding for Perceptual Quality Konrad V Pfaff Rezumat Datorit r sp ndirii echipamentelor ce folo sese tehnici de codare audio cu debit binar redus dbr in activitatea stud
10. R j 3 TEGO y exp 3 or if Z R jX B R R cos i 4 X TU INE In practice there are 3 cases function of caracter of unknown impedance Z can be pure resistor inductive impedance or capacitive impedance Thus if U R and Z are fixed B and g must be modified so the next relation will be true RU Z U 5 1 For Z x R from equations 5 and 2 results d T From 4 results R gt 0 and X 0 so it expected shows the block diagram of the adaptive filter which 2 For Z Ry joL equation 5 is shown in fig 2 implements this algorithm U can be write as a weight sum between a sin function and a cos function as phasor diagram and follows del 31 2 Fig 2 Phasor diagram for K R j L From 4 it follows R gt 0 and X gt 0 so it expected 3 For Z R 1 j C equation 5 is shown in fig 3 as phasor diagram and results de 1 2 n Lab PC board joC Fig 3 Phasor diagram for Z R 1 jaC From 4 it follows R gt 0 and X lt 0 so it expected In this paper the AC bridge with virtual arms is implemented using a data acquisition board of National Instruments Lab PC type and the LMS algorithm Fig 4 The block diagram of AC bridge AC Bridge II SYSTEM IMPLEMENTATION to DAC A Hardware Description from ADC The block diagram of realized AC bridge is presented in fig 4 The voltages U and U are generated using two 12 bit digital to
11. The experiments were carried out over the standard NIST database for hand printed amp handwritten characters The processing with each set of matrixes was performed in identical software environments 4 UG h AN i a b Fig 3 1 a Image processed with L structuring element b Image processed with H structuring element Each of the four skeleton techniques were analyzed and it resulted that A The standard thinning and the M homotropic structuring element technique are not able to perform satisfactory with the hand printed amp handwritten documents images B The L homotropic structuring element technique behaves better but the resulting skeleton is still coarse and it has a thorny aspect The coarse parts and the thorns are sometimes source of inadequate final classifications C The proposed H homotropic structuring element technique is more precise while it is at least as fast as the L technique The skeletons have almost 4 less pixels due to the absence of the coarse parts and thorns It is also important to stress here that the final recognition results on the experimental test set improved with 0 62 because of using H instead L homotropic technique The explanation of this resides mainly in the less coarse strokes that leave less space for miss interpretations V CONCLUSIONS The skeletonization is an important step in preprocessing documents images If the skeleton is
12. e 1 pass 2 pass Fig 2 Diagram of cascading Many changes in operational practices can be expected as a result of the proliferation of bitrate reduction in the broadcast transmission network Tandem coding will need to be minimized or a high enough bitrate chosen at the beginning of the broadcast chain Where no other processing is required cut and paste editing may be carried out in the encoded domain instead of decoding to linear PCM and re encoding the signal Although a codec may be found transparent for one encoding decoding pass serious impairements can become evident after just a few tandem passes through the same codec These impairements are often signal dependent and so a simple spot check is not enough to verify signal quality of a tandem chain Audio coding algoritms based on auditory masking attempt to shape the quantization noise spectrum to fall below the input signa spectrum or according to psychoacoustic models causing the noise to be masked by the signal If a sufficient margin is not included in the noise to mask ratio subsequent equalization may alter the masking characterisitics of the signal rendering this noise audible In studio applications multiple decoded signals will be mixed and then re encoded for transmission This may occur up to several times in the course of developing a broadcast program It is conceivable that mixing two transparently decoded audio signals which have different
13. pp 294 305 7 Yan H amp all A Low Power 16x16 b Parallel Multiplier Utilizing Pass Transistor Logic IEEE Journal of Solid State Circuits Vol 34 No 10 October 1999 pp 1395 1399 8 Lai F A 10 ns Hybrid Number System Data Execution Unit for Digital Signal Processing Systems IEEE Journal of Solid State Circuits Vol 26 No 4 Apr 1991 pp 590 599 9 Jurca L A Logarithmic Unit for Fast Performing of Multiplication and Division Operations in Single Precision Format Buletinul UPT Tom 45 59 Fasc 2 2000 pp 29 32 4 2 L compressor 16 Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Physical Causes of Integrated Circuits Behaviour Degradation under lonising irradiation Avram lonel Adrian Resume Modern research in integrated circuits IC fiability domain show that varied IC fabrication technologies schemotechnics and constructive particularity and exploitation condition leads to different behaviour of IC under ionising conditions Introduction For a precise design of IC under ionising irradiation conditions are necessary detailed information not only about irradiant physics process in semiconductor materials but also are required information above IC characteristics modification under ionising irradiation and information about i
14. witch is organized sequentially parallel or in a mix mode The elements from the object structure can take a value witch can define an instance of the object It is unnecessary for the object to have defined values for all of its components We use the heritage concept from object oriented programming Fora more real description of the real world it can be established a semantic interpretation for each component of a multimedia object The semantic description became usefully in the case of searching different objects in a multimedia application or when the number of objects is very large Is necessary to have a structure of the elements based on semantic concepts Next I will present the parameters that are considered very important to me for describing a multimedia object and simple elements I have organized object descriptions in a structure of classes that has close properties to the classes that we meet in the case of object oriented programming A model of multimedia description is presented in Fig 2 For an efficient description of multimedia objects I proposed a describing parameters structure 1 General parameters that are describing the object like object ID storage physical location of the object a flag witch indicate 1f the object 1s simple or complex and information s about element structure of the object 2 Semantic parameters witch describe the object thru key words 3 Type objects paramet
15. 28 30 and 31 416 vg ug TAW 201 am TANS 12287 02080 re Vig 24 16KHZ sine 0dB 19 5KHz sine 70dB 16bit 44 kHz PCM original signal frequency domain e 3 vm a ne em Vig 25 16kHz sine 0dH 19 SKHz sine 70d 3 16bit A4 KHz ATRAC encoded signal 2 Generation frequency domain wes 166 wm r6 LE Gar EE 1 SME ae a ce ia 4 Fig 26 16kHz sine 0dB 19 5kHz sine 70dB 16bit 44 1kHz ATRAC encoded signal 2 Generation time domain Vig 27 1GKHZ sine 0dB 19 5kHz sine 70dB 16bit 44 kHz ATRAC encoded signal 6 Generation frequency domain D K z 4 orgie Fig 2 EGKHz sine 0di3 19 5k 142 sine 70413 16bit 44 Kli ATRAC encoded signal 6 Generation time domain sii Len ea Doe s Fig 29 IGkHz sine dB 19 5kHz sine 70dl4 l bit 44 I I ATRAC encoded signal 24 Generation frequency dom un analogue transfer Li Fig 30 loki Iz sine 0dB 19 5kHz sine 7041 16bit 44 iki lz ATRAC encoded signal 2 Generation ttime domain analogue transfer Fig 21 16kHiz sine 0dH 19 5kHz sine 704R 16bit 44 1k 47 ATRAC encoded signal 9 Generation time domain digital transter SON Y V CONCLUSIONS The most effective way of avoiding the accumulation of errors is to stay within the coded data format as long as possible Thus no further quantization processes introducin additional quantization noise are
16. 5 and used to recognize the Canadian dollars The method was developed by the author in Matlab considering the particularities of Romanian banknotes IL BASIC CONCEPTS The most important aspect in recognizing a bank note is to select the significant features of the tested image These features can be chromatic topologic or related to the objects size For instance the characteristics selected to recognize the 10 000 bank note are presented in Fig 1 The analogic CNN algorithm has to test if the selected feature is present or not in the image The CNN algorithm is similar to a sequential classic subroutine Its output image is binary The image pixels have a value of 1 true in locations where the feature was found and 1 false in the rest of the image RANCA NA IONALA A ROMANIEI A A E sy Objects group selected for ecognition Fig 1 Results of different algorithms can be combined applying spatial logic CNN operators All these operations can be processed on chip without read write except original input and final output Another important aspect of the problem is the processing speed The modern copy machines must have a very good resolution and a copying speed greater than four or five copies second So the time Comunica ii Bd V P rvan Timisoara 1900 e mail corinab hermes ee utt ro 47 necessary to verify the document must not exceed 200 ms Due to their parallel processing and possible hard
17. Analysis Design and Implementation of an AC Bridge for Impedance Measurements IEEE Trans Instrum Meas vol 43 pp 894 899 1994 3 L Angrisani A Baccigalupi and A Pietrosanto A digital Signal Processing Instrument for Impedance Measurement IEEE Trans Instrum Meas vol 45 pp 930 934 1996 4 B Widrow S Stearns Adaptive Signal Processing Prentice Hall New Jersey 1985 5 Lab PC User Manual National Instruments 1994 Buletinul tiin ific al Universit ii Politehnica din Timisoara Seria ELECTRONICA si TELECOMUNICATII TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Considerations Concerning the Practical Behaviour of a Digitising Oscilloscope in Different Signal Recording Modes Daniel Belega and Septimiu Mischie Abstract In this paper is analysed the practical behaviour of the digitising oscilloscope DSO HM407 in different signal recording modes Refresh RFR Average AVM and Envelope ENV In RFR and AVM modes were experimentally evaluated the dynamic performance of the HM407 by the method proposed 1 It is presented some important conclusions concerning the dynamic performance of the HM407 obtained in these modes For different input signal types were shown the signals acquired in ENV mode The signal acquired in this mode permits a qualitatively appreciation of the input signal amplitude and frequency variations Index terms Testing of waveform digitizer soft
18. Compression Algorithms Proceedings of the First COST G6 Workshop on Digital Audio Effects pp 99 110 Barcelona November 1998 Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 2 2001 Instruc iuni de redactare a articolelor pentru Buletinul tiin ific al Facult ii de Electronic i Telecomunica ii Gheorghe I Popescu Rezumat Aceste instructiuni sunt concepute s va prezinte modul de redactare al articolelor pentru Buletinul stiintific al Facult tii de Electronic si Telecomunicatii Prezentul material este conceput ca model pentru modul cum trebuie s arate articolele gata de publicare Rezumatul trebuie s con in descrierea problemei metodele i solu iile propuse i rezultatele experimentale ob inute n cel mult 12 r nduri Nu se admit referin e bibliografice n cadrul rezumatului Cuvinte cheie redactare buletin tiin ific I INTRODUCERE Formatul buletinului va fi A4 Articolele inclusiv tabelele i figurile nu trebuie s dep easc 6 pagini Num rul de pagini trebuie s fie obligatoriu par II INSTRUC IUNI DE REDACTARE Articolul trebuie s fie transmis n forma standard descris n acest material Tip rirea se va face cu o imprimant de bun calitate pe o singur fata a paginii Textul se va plasa pe dou coloane de 8 cm cu spa iu de
19. M the normalized mantissa represented on 23 bits and E the biased exponent represented on 8 bits In the Logarithm Number System LNS a binary number z is represented z 1 2 4 where Sz is the sign bit and Nz is a fixed point number having n bits out of which i bits for the integer part lz and f bits for the fractional part Fz We have Facultatea de Electronic i Telecomunicatii Departamentul n i f and Nz lz Fz 5 The only deficiency of the LNS processors is that they so far permit only single precision computation and zero is not representable in the LNS For system compatibility a simple but slightly inefficient encoding is used Zero is represented by a distinct bit Z in the LNS When the bit Z is set to 1 the number is zero regardless of all other bits in the value Thus i 8 and f 23 which including the bit Z and the sign bit corresponds to 33 bits 3 Considering the normalized mantissa 1 0 M in the domain 1 2 the biased integer part of the logarithm of the number is given by the very value of the biased exponent and the fractional part by the logarithm of the mantissa This concatenation is possible because the binary logarithm of the normalized mantissa is always a positive fractional number The calculation of logarithm and antilogarithm use the partition of the argument and the memorising of only certain values for reducing the amount of memory and in addition we apply a corr
20. The adders are 2 level hybrid adders with four blocks of 8 b carry look ahead adders CLA plus the most significant block of 4 b CLA in the 1 level and a carry select mechanism in the 2 level In the most unfavourable case the delay due to the carry propagation through the circuit presented in figure 5 was found 4 ns by simulation with MSim program in 0 5um CMOS technology The maximum propagation time through the Wallace tree which contains three levels of 4 2 L compressors is 3 x 1 1 ns at which 0 7 ns are added for the generation of the partial products which leads to a total of 4 ns Sop Fig 5 Block diagram of the ALU the adder subtracter circuit The sign bits of the operands are processed separately in the ALU The sign bit of the result can be obtained by the implementation of the logical function Sc Sa Sg 16 IV ERROR ANALYSIS After analysing with the help of the Matlab program the errors introduced by using the algorithm described in 1 for the generation of binary logarithm and antilogarithm we had the confirmation of the value of 3 x 10 initially mentioned in 1 as the maximum conversion error However this error is of about 2 5 times bigger than the one in single precision format which is of 1 19 x 10 For the further minimisation of the conversion error we suggested a solution of our own which will add correction values on certain address intervals of the ROMA and ROMC in the
21. a table with n locations using 10 Next it follows a periodic process that contains mainly the following steps the two weights w and w has the initial values at 1 and the voltage of both DAC s was initially OV 1 Acquire one sample from E using ADC 2 Senda 12 bit word to DAC to obtain U 3 Compute U with 6 4 Senda 12 bit word to DAC to obtain U 5 Compute w and w with 11 This periodic process is finished when E is smaller than a given threshold or the LMS algorithm has converged In step 3 the values for cos function has been obtained by reading from the sin table with a n 4 offset The program in C language gives values for a sin function in a floating point format Then these values hase been converted in a 12 bit two s complement format for DAC s command Writing the least significant 8 bits and then the most significant 4 bits loads the each two DAC s The period of this process T is very important and must be precisely generate because 1 The sample time T for acquisition of E is equal with period 7 2 The period T of both waveforms U and U is multiple of T namely T nT D 12 Thus 7 has been achieved using a 8253 Counter Timer which is on LabPC board III EXPERIMENTAL RESULTS A Values of parameters By programming the 8253 Counter Timer 7 had a value of 100 us Thus follows that F 1 7 10 kHz and because for n number of points per period a value of 100 is assigned
22. baz o nou abordare n definirea acestui tip de circuit ce a condus la l rgirea familiei din care face parte Simul rile efectuate au dovedit avantajele n ce prive te viteza de propagare a transportului prin noua structur realizat n tehnologie CMOS 0 25 micrometri comparativ cu alte solu ii prezentate n literatura de specialitate Cuvinte cheie compresor 4 2 format LNS format virgul flotant simpl precizie l INTRODUCTION The increased level of integration brought by modern VLSI and ULSI technology has rendered possible the integration of many components that were considered complex and were implemented only partially or not at all in the past The multiplication operation is certainly present in many parts of a floating point FP processor most notably in signal processing graphics and scientific computation Parallel multipliers have even migrated into the fixed point processor of digital computers for de purpose of speeding up and facilitating address calculation and more recently into the hybrid FP LNS Logarithm Number System processors 8 9 for de purpose of allowing implementation of algorithms needed for format conversion between FP and LNS The most appropriate multiplier is obtained when the partial products from the multiplication array are grouped into a Wallace tree 1 Initially the component blocks of the tree were CSA Carry Save Addition blocks which were made up of independen
23. become four as shown in 2 2 1 1 11 1 11 0 o 1 o ip wb pep d x us pp 1 01 02 00 0 1111 1 1 1 1 In order to improve speed and accuracy an auxiliary set of four matrixes were borrowed from the L technique 1 1 0 0 1 1 1 1 OF JO 1 1 23 Or Log ERES Ig CE 1 Thus the H structuring elements set is 1 1 1j 1 1 1 1 0 2 4 This homomorphic structuring elements set behaves better with the hand printed amp handwritten documents images as it will be shown in section IV The resulting skeletons are less coarse and the number of needed processing cycles is equal or less the one in L technique III SOFTWARE IMPLEMENTATION The program performing the processing tasks has been developed in Java JDK 1 3 It is a Window application and it is able to e Display the documents image e Execute the recursive thinning skeleton stage is reached e Display a contour showing the number of cycles performed e Save the final image resulted after the process The class structure contains a main class W extending the javax swing JFrame and implementing the Runnable interface and a until 38 processing Sb class extending java awt image ImageFilter E subtiere E SCH Contor cicluri 0 Fig 2 1 The Skeleton program frame in the starting state The main class is the program main thread and contains code to implement the window frame This window has a file menu with open filt
24. but depending of expected values of w or w can be decrease to about 1 5V Value of the coefficient z from LMS algorithm must have a value in interval 0 1 Amax where Amax is the largest eigenvalue of the autocorelation matrix of input sequence the sum between a sin function and a cos function in adaptive filter Because values of input sequence are numbers in range 2048 2047 1 must have a very little value Experimental was obtained that a value of about 10 leads to a fast convergence of LMS algorithm B Obtained results In experimental operations we have been measured resistors capacitive impedances and inductive impedances In table 1 are presented values which has obtained after the measure with implemented AC bridge and for comparison values which has obtained by measuring with another instrument Values of amplitude of U reference resistor R and values obtained for w and w are presented too It can see that generally errors are small specialy for resistors capacitors and inductances For serial resitors of capacitive impedances and inductive impedances the errors are higher because theirs values are very little comparative with R and from relation 9 w is very small Consequently the first term in 6 will be of order of few LSB of DAC In fig 6 is shown variation of error voltage E The amplitude is in LSB It can see that after about 700 samples or 7 periods this voltage is approximately 0 It res
25. components and to show how this varies with spectral frequency and amplitude Contents of Waveforms To determine the effect of ATRAC encoding on the masked component of a rudimentary complex waveform signals consisting of a masker component D and a maskee E were applied Based on the theory of one critical band being aproximately 1 4 of an octave waveforms were created consisting of two partials separated in frequency by one critical band using the following formula fnix f f where f 5 4 f 1 Synopsis of results At lower frequencies the performance of the ATRAC encoder shows no measurable increase in spectral quantisation noise see figures 7 to 10 At higher frequencies however there is a significant increase in spectral quantisation noise rising by an average of 10dB at 10kHz through to the upper frequency limit as shown in figure 12 Fig 8 400Hz OdB 520Hz 50dB 16bit 44 1kHz PCM ATRAC encoded signal 54 18S 2818 4218 L Dap RD o Fig 9 By le 0AB 98COHz SOdB 16bit 44 KHz PCM signal KT 2 t at un di am Main d n vg 7895 122762 toto re Fig 10 RUOUHZ 0dB 9800117 50dB 16bit 44 1 kHz PCM ATRAC eneoded signal Ht Le 44 L L H 4 bech a ds je L t 2419 seh PATS 12262 206080 be Fig 11 16000HZ 0di3 19500Hz SOdB 16bit 44 1KHz PCM signal Tee boyy pop sue RACE LR at rr pe Gert jeep pa mea antes a m im us v gm 818 2
26. demonstrate the allocation of bits according to the frequency range of the spectral components to illustrate potential situations where the allocation of all available bits takes place and the resultant effect on bandwidth to compare the level of pre echo present in ATRAC recordings of highly transient signals with those of other high quality low bit rate digital audio coding Systems to illustrate potential problems in the time domain which occur as a result of ATRAC encoding over multiple generations 53 4 2 Experimental Procedures Source waveforms were digitally created using SoundForge 5 0 waveform generator apart from those Signals used for transient tests which were taken from a test CD AES All waveforms were 16bit 44 1K Hz sample rate The constituent spectral content of each digitally generated waveform was selected based on a theoretical understanding of how the object under test operates Sine waves were used for tests based on theories of perceptual masking whilst square waves were used as a demonstration of a situation where all available bits might be used hence resulting in the maximum amount of quantization noise that is likely to occur as a result of data reduction Digital transfer for recording of all test items was employed as shown in fig 4 via the Sony Philips Digital Interface S PDIF optical input of the Sharp MD722 eliminating the posibility of analogue circuitry affecting t
27. en gardant sous contr le les distorsions On pr sente aussi quelques r sultats de simulation Cuvinte cheie parole compression paquets de cosinus I INTRODUCTION La compression des donn es est utilis e souvent pour le stockage et pour la transmission des donn es Il y a une grande diversit d algorithmes de compression Il y a deux grandes cat gories 1 les algorithmes pour la compression sans pertes et les algorithmes pour la compression a pertes Tous ces algorithmes ont le m me but r duire la redondance de la suite de donn es consid r e mais ceux appartenant a la premi re classe r alisent une reconstuction parfaite et ceux qui appartient a la deuxi me classe ne r alisent pas une telle reconstruction Cette deuxi me classe contient divers algorithmes pour la compression de la parole de la musique ou des images Dans ce travail on pr sente une m thode pour la compression d compression de la parole appartenant cette deuxi me classe qui exploite une transform e orthogonale tr s connue appel e Transform e en paquets de cosinus TPC 2 3 4 5 Cette transform e a t s lectionn e pour Goldstern Timisoara 41 sa vitesse de calcul comparable a la vitesse de la Transform e de Fourier rapide et pour son ad quation au mod le de la parole La structure de ce travail est la suivante Dans le paragraphe num ro 2 est pr sent le systeme de compression propos Le sujet du tr
28. frequency lines might get set to zero deleted The most common case is that some high frequency content is lost If the loss of the bandwidth is not constant but changing frame to frame the effect becomes more objectionable than in the case of a constant bandwidth reduction Preechoes are a very comrnon and famous possible artifact for high frequency resolution perceptual coding systems To understand preechoes let us have a look at the decoder of a perceptual coding systera fig 3 The reconstructed frequency lines are combined in the synthesis filterbank This filterbank consists of a modulation matrix and a synthesis window The quantization error in the coder can be seen as a signal added to the original frequency linc The length in time of such a signal is equal to the length of the synthesis window Thus reconstruction errors are spread over the full window length If the music signal contains a sudden increase in signal energy like a castanet attack the quantisation error is increased too As explained above this quantization error noise signal is spread over the length of the synthesis window If the attack occured well within the analysis window this error will precede the actual cause for its existence in time If this prenoise extends beyond the premasking backwards masking period it becomes audible and is called preecho There are a number of techniques to avoid audible preechoes including variable bit rate coding
29. il 2 BULETINUL TIIN IFIC al Universit ii POLITEHNICA din Timi oara Rom nia Seria ELECTRONIC I TELECOMUNICA II SCIENTIFIC BULLETIN of the POLITEHNICA University of Timi oara Romania Transactions on ELECTRONICS AND COMMUNICATIONS Tomul 46 60 Fascicola 1 2001 ISSN 1583 3380 ele RS EDITURA POLITEHNICA Redactor sef Editor in chief Prof dr ing Ioan Nafornit Colegiul de redactie Editorial Board Prof dr ing Virgil Tiponut Prof dr ing Alexandru Isar Conf dr ing Dan Lacr m Conf dr ing Dorina Isar Prof dr ing Traian Jurc Conf dr ing Aldo De Sabata As ing Kovaci Maria secretar de redactie Colectivul de recenzare Advisory Board Prof dr Prof dr Prof dr Prof dr Prof dr Prof dr Prof dr ing Ioan Nafornit UP Timisoara ing Monica Borda UT Cluj Napoca ing Br ndusa Pantelimon UP Bucuresti ing Ciochin Silviu UP Bucuresti ing Dumitru Stanomir UP Bucuresti ing Vladimir Cretu UP Timisoara ing Virgil Tiponut UP Timisoara Buletinul Stiintific al Universit ii Politehnica din Timisoara Seria ELECTRONIC SI TELECOMUNICATII TRANSACTIONS ON ELECTRONICS AND COMMUNICATIONS Tom 46 60 Fascicola 1 2001 SUMAR ELECTRONICA APLICATA Lucian A Jurca Using Masked SE Method de Fast ah and Division Logarithmic Unit Design a tocado Seno Aba Lise hdc Wis dea mes serere ony Me ibat Lucian A Jurca
30. iradiation damage n metal oxide semiconductor Proc IEEE 1966 V 54 p 894 895 7 Stanlez A G A model for shifts in the gate turn on voltage of insulated gate field effect devices induced by ionizing irradiation IEEE Trans Electron Devices 1967 V 4 p 134 138 Ha Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Analog Input LabVIEW Applications for Measuring Temperature Mihaela Lascu Abstract Signal Conditioning eXtensions for Instrumentation SCXI is a highly expandable signal conditioning system This paper describes the basic concepts of signal conditioning the setup procedure for SCXI hardware the hardware operating modes the procedure for software installation and configuration the special programming considerations for SCXI in LabVIEW and some common SCXI and acquisition applications Keywords SCXI transducer thermocouple temperature data acquisition I INTRODUCTION Transducers can generate electrical signals to measure physical phenomena such as temperature force sound or light To measure signals from transducers you must convert them into a form that a DAQ device can accept For example the output voltage of most thermocouples is very small and susceptible to noise Therefore you may need to amplify and or filter the thermocouple output bef
31. la sortie du d tecteur de seuil devient gal W Mais ce nombre repr sente une valeur minimale Donc en utilisant cette fonctionnelle de co t le nombre de coefficients non nuls la sortie du d tecteur de seuil est minimis C est le motif pour lequel on peut affirmer que le facteur de compression est maximis quand cette fonctionnelle de co t est utilis e pour la s lection du meilleur paquet de cosinus En augmentant la valeur du seuil t le nombre NW devient plus petit et la valeur du facteur de compression devient plus grande Malheureusement la valeur du snr devient plus petite aussi Donc l augmentation de la valeur du seuil t doit tre contr l e pour garder la transparence de la compression C est le motif pour lequel le d tecteur de seuil de la figure 1 doit tre adaptatif Un autre param tre de la TPC qui doit tre consid r pour l optimisation de la compression est son nombre d it rations IV LE DETECTEUR DE SEUIL En analysant le syst me de la figure 1 on peut constater que la distorsion introduite par la compression a la valeur D E r shi 8 o E repr sente l op rateur de moyenne statistique Parce que la TPC et son inverse la TPCI sont des transform es orthogonales la derni re relation devient D E y n ufr 9 43 La valeur du seuil t est choisie pour satisfaire la condition Dem del 10 o E repr sente l nergie du signal d entr e del On peut prouver la prop
32. le bloc num ro k et 32 B Vy kal 25 Le nombre N peut tre calcul en utilisant la relation 26 o Cr repr sente le nombre de bits demand s pour la repr sentation de la position de chaque chantillon de valeur non nulle du bloc num ro k Parce qu en chaque bloc il y a un nombre maximal de 32 telles positions les valeurs de sont inf rieurs 5 Donc une borne superieure pour Np est 32 5 Y N k 32 N k 1 27 45 VI LES AUTRES BLOCS DU SYSTEME DE COMPRESSION L utilisation du codeur C de la figure 1 augmente le facteur de compression Ce syst me r alise une compression sans pertes C est le motif pour lequel son utilisation n affecte pas la valeur du snr L impl mentation de ce bloc fait appel une technique traditionnelle de codage comme par exemple le codage de Huffman ou le codage arithm tique 1 Le signal v n obtenu la sortie du codeur repr sente le r sultat de la proc dure de compression C est le signal qui est stock ou transmis Les autres deux blocs du syst me de la figure 1 sont utilis s dans la phase de reconstruction Le bloc D r alise le d codage du signal dal On obtient sa sortie les signaux uy n et la s quence z y k 132 En utilisant la relation 21 on calcule les valeurs 7 La denormalisation d crite par la relation wl De ypeeop A 28 k H est r alis e Par le concatenage des blocs w n on obtient le signal w n Le dernie
33. local increase in the bit rate locally to reduce the amplitude of preecho but overall this type of artifact belongs to the mos difficult to avoid 4 audio Analysis Quantization 7 Encoding F stream ni Qi i VR eg out iei Perceprual M model ne o g audi ms Dana Cae D s Fig 3 Block diagram of a perceptual encoding decoding systera Especially at low bit rates and lower sampling frequencies there is a mismatch between time resolution of the coder and the requirements to follow the time structure of some signals This effect is most noticeable for speech signals and for listening via headphones The effect is sometimes called double speak because a single voice tends to sound like the same recording was done twice with the same speaker and overlayed How to measure codec quality To measure codec quality of high quality audio coders there are basically three measurement methods Listening tests simple objective measurement methods and perceptual measurement techniques The ITU R with input from a number of broadcasters and the MPEG audio group has developed a very elaborate set of rules for listening tests The aim of these tests is to stress the encoders under worst case conditions to find the material which is most difficult to encode and to exhibit the performance of the encoders under test for this material This follows the observation that in a lot of cases coding artifacts become audible only after an exte
34. of selected objects from Fig 1 In this situation the CNN will stop the banknote copying IV THE RECONSTRUCTION TEMPLATE As example of an elementary operation it will be presented the reconstruction template because most of the subroutines uses it This operator extracts from a set of objects the one with a desired feature signature It uses two images The image containing many objects is applied as input to the CNN The RANCA NATIONAL A ANDMANIED TE RAMA NAPONA A A MDIM NIEI TE QD Ai Fig 4 The steps of color filtering a input image b prefiltering result c Objects with a greater value than the minimum value color of interest d Objects with a greater value than the maximum value color of interest e XOR between image c and d f output image RAMA AORA A ANDMANIET wy 4 la 7 NC to 10000 d Fig 5 Steps of size classification a input image b image with n pixels levels deleted c image with m n pixels levels deleted d reconstruction of the image b e reconstruction of the image c f XOR application to d and e gt 6000 x ca Gb CET NA a i z u 10000 e 10000 e 2 20000 Fig 6 49 image containing the signature is applied as CNN state When the transition settles down at the output of the CNN will be obtained the reconstructed object with a particular signature as it can be seen in Fig 7 The reconstruction template is the follow
35. of the typographical quality requirements all images presented in this paper are negatives of the real ones a b Fig 1 1 a Test image rectangle b Skeleton Unfortunately when real life images i e hand printed amp handwritten documents images are processed the resulting skeleton is very coarse see Fig 1 2 Thus a supplemental curve relaxation procedure became compulsory 4 4 Wan XR urs a b Fig 1 2 a Handwritten document image b Skeleton with recursive standard thinning Better techniques rely on a sequence of successive thinning with homotropic structuring elements L 1 2 or M 1 3 able to deliver one pixel thick connected lines and curves skeletons approximations in a less time consuming process The symbol is used in order to mark the position where the white or black color of the pixel is of no consequence for the local thinning decision and therefore no test is to be made about them in the software implementation 1 1 Se 1 1 CG sm 11 1 2 o o oj 0 a op oer f 1 1 1 NOR 1 1 1 3 The other structuring elements reported in the image processing literature C D E are crafted for some slightly different targets and behave poorly with both typed and handwritten documents images The recursive thinning results with the two homotropic structuring elements on the same image are presented in Fig 1 3 The software implementation of the recursive thinn
36. of time for the horizontal carry propagation ALU operated with data of any polarity which complicated its control logic and led to a further delay Later on the same author presented a new architecture in which the product EyxAEy was calculated not with binary multipliers but with PLA circuits 3 which permitted a saving of area on the chip however maintaining the same computation speed Considering these disadvantages this paper proposes a new method of logarithmic unit design which does not give up the multipliers and leads to a substantial increase of the calculation speed of at least 1 6 times resulting an occupied area which is smaller than in 1 and comparable to 3 II THE MASKED LOGARITHM METHOD In order to eliminate the disadvantages mentioned above we offered a solution of our own based on a new method of logarithmic unit design which we called the masked logarithm method To elaborate this method we considered the following furthers aims the new structure should be organised in no more than 6 pipeline stages the number of logical gates used on certain pipeline stages should not increase significantly the size of the latches between the certain pipeline stages should not increase significantly the propagation time on the critical stage of the pipeline should be as close as possible to the physically achievable limit namely to the access time to the ROM integrated on the chip the propagat
37. permits only a qualitatively appreciation of the input signal amplitude and frequency variations Sinewave T T Magnitude V 0 2 0 time ms a Square Magnitude V 0 1 02 0 3 04 0 5 time ms b Fig 6 The signals acquired as a function of time in the case of a a sinewave input signal b square input signal V CONCLUSION This paper analyses the practical behaviour of an actual DSO HM407 in some signal recording modes RFR AVM and ENV Based on the results of the dynamic testing of HM407 in RFR and AVM modes were reached some important conclusions e in both modes when the number of samples acquired is a power of 2 for testing all the digitizer levels is necessary to overcome the situations in which the number of the input sinewave cycles is equal or very close to a number which is an integral power of 2 e in AVM mode the best performance are obtained in the case of using the maximum available value for M The ENV mode permits a qualitatively appreciation of the input signal amplitude and frequency variations APPENDIX A DESCRIPTION OF THE METHOD PROPOSED IN 1 Assume a pure sinewave acquired by the digitizer under test at a sampling frequency f the number of samples acquired is M The data record obtained is y m m 0 1 M 1 For estimating the dynamic performance of the digitizer the method proposed in 1
38. signal x t est une constante ind pendante de la m re d ondelettes s lectionn e on peut affirmer que le nombre N est plus petit si la magnitude des coefficients dy est plus grande Donc pour la compression de la parole le meilleur choix pour la m re des ondelettes est la classe de Malvar qui conduit aux paquets de cosinus 4 B AUTRES PROPRIETES DE LA TPC La TPC est une transform e adaptative Elle peut tre optimis e en utilisant une proc dure de recherche de la meilleure base 2 C est une proc dure tr s puissante qui peut augmenter la qualit d une certaine m thode de traitement du signal Il y a plusieurs fonctionnelles de co t qui peuvent tre minimis es pour le choix de la meilleure base La fonctionnelle la plus utilis e est l entropie des coefficients diy 2 Mais la minimisation de cette fonctionnelle ne conduit pas a la maximisation du facteur de compression La fonctionnelle de co t dont la minimisation conduit a la minimisation du nombre N et donc la maximisation du facteur de compression est celle qui minimise le nombre des coefficients sup rieurs un certain seuil t 2 N En effet en utilisant cette fonctionnelle de co t pour le choix de la meilleure base le meilleur paquet de cosinus on obtient un certain nombre N En rejetant tous les coefficients inf rieurs au seuil t a laide du d tecteur de seuil de la figure 1 le nombre des coefficients non nuls
39. temperature sensor input limits which are different from your acquisition input limits To read from a temperature sensor based on an IC sensor or a thermistor set the input limit range from 2 to 2 V In the following we represent the Continuous Thermocouple Measurement VI which is used to read multiple temperatures from each channel Enter your Device number Thermocouple Type Cold Junction 20 Channel CJC Sensor Temperature Range in Deg C and the Channels you want to read glherm Fig 4 Connector Pane for Continuous Thermocouple Measurement Channel Latest TEMPERATURE Legend D E 2 a 2 HALTE James aj dm Temperature Range in Deg C i 3 low imit Cold Junction Channel CJC Sensor 000 jo AIC Sensor E ei Fig 5 Front Panel for Continuous Thermocouple Measurement TEMPERATURE p Po ae d CH R GEES CIC Sensei bal p B Er PRE ball Channel EIL SIM ml l Act T ture Fi Fr oom iit Plane Lgs Tess fe vu E un h HTML FieFah Trage Diesioy Fig 6 Block Diagram for Continuous Thermocouple Measurement This example is build only with a data acquisition board Lab PC without using a SCXI modul The following example presents a simulat
40. the following equation Ey ny A x y2 Eyin 1ytAEyiny x y2 1 Eymt Aua x Y2 A Y n 9 aALog 1 y y Fig 2 Negative slope segment achieved through linear interpolation between two consecutive memorised values E The implementation of this equation leads to an arrangement of the partial products as they are presented in figure 3 A generic presentation with qy for the complemented py bits of yo respectively with pa for the bits of AEy was used 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 0 0 0 0 0 Pa papapdpapapapapaPapapa 0 0 0 0 0 qy qydydy ydyqdy qydydyQy Ay 0 0 0 0 qyaqyqyaygynyaygy Gydydydy 0 0 0 qygyqyayaygyRyaygy dyaydy 0 qyqyqyqyqygygygy ayaydy dy qyqdydydydydydydyqdyjAyqdyqdy Fig 3 A section through multiplication area after the implementation of equation 9 Thus we can obtain the same result of the logarithm computation if we implement the right part of equation 9 Starting from the memory location corresponding to the address 907 of ROMA instead of memorising the value Ey Ey n 1 is memorised i e exactly what should have been found at the next address In each location a supplementary bit will be memorised called control bit which takes the value O for addresses 0 906 and 1 for addresses 907 2047 If this bit is 1 then the generation of partial products will be done with y2 having the bits reversed and another pseudo partial product with a s
41. the ideal candidate for a direct subjective comparison with ATRAC Ili QUALITY CONSIDERATIONS Audio Coding artifacts differ in their nature and their audibility much from well known audio distorsions wow amp flutter tape saturation crosstalk non linearity group delay distorsion intermodulation etc and they may be very difficult to identify in a networked audio system being a cascade of different signal processing devices Common types of artifacts Unlike analog hi fi equipment or FM broadcasting perceptual encoders when run at too low bit rates or with wrong parameters exhibit sound deficiencies which are in most cases different from noise or dist rsion characteristics we all are used to The KE se QUE An d 2S ek mme Ad m9 2 D ge 3 reason for this is the process generating differences in sound The error introduced by a high frequency resolution perceptual coder is usually best modeled by a time varying in the rhythm of the basic block or frame length error at certain frequencies but not constrained to the harmonics of the music signal So the signal may be sounding distorted but not like harmonic distorsions noisy but with the noise introduced only in a certain frequency range rough If an encoder runs out of bits i e it does not find amp way to encode a block of music data with the required fidelity e g allowed noise per critical band within the bounds of available bit rate some
42. un quantificateur non uniforme on peut obtenir des r sultats meilleurs Une solution tr s simple pour une quantification non uniforme a t choisie pour le syst me de compression propos dans ce travail La s quence obtenue la sortie du d tecteur de seuil de la figure 1 le signal z n est divis e en 32 blocs z n amp 132 de m me longueur En chaque bloc on r alise une quantification uniforme Les plus grandes valeurs des signaux z n zy et z n zu sont d tect es Pour chaque bloc est allou un certain nombre de bits Cette proc dure est bas e sur les valeurs z Pour la valeur z sont allou s 6 bits 2 niveaux de quantification Pour les valeurs Zu sont allou s 21 niveaux de quantification o repr sente la partie enti re Ainsi un nombre de 5 log y 1 bits sont allou s pour chaque chantillon du bloc num ro k La quantification du bloc num ro k est r alis e en Zk n utilisant la transformation up n 5 ag abr n Ainsi une normalisation de niveau est r alis e en chaque bloc La denormalisation correspondante doit tre r alis e pendant la phase de reconstruction avant le calcul de la TPCI Cette op ration est attribu e au bloc D de la figure 1 Le grand avantage de la proc dure de quantification propos e vient de la propri t de decorrelation de la TPC Gr ce cette propri t beaucoup des valeurs zu sont nulles Les valeurs corresp
43. variety of network types and their development is continuous we will check for the beginning the type of existing network after that we will appeal to the specific routines for that types of network This kind of structure would permit an easier adaptation of the application depending on the hardware and software resources of the network My recommendation for the Network routines is to have possibilities to verify the network level of traffic Is necessary to inform the user about this level of traffic and to propose different solution if are some problems Client Server many applications use for communication implementation of client server services These kinds of technologies are useful for job distribution in a heterogenic network to enhance the speed of processing 2 Users Management if an application is accessed by many users in the same time or sequential the routines for this module most to offer a management of users D Integrated technologies This module contains libraries dedicated to working with databases ODBC integration libraries of the objects realized with other applications OLE technologies implementation of different kinds of services client server libraries DDE with optimization functions for different equipments inside PC MCI functions for sound video and CD ROM evaluation libraries of quality service offered by the application on hardware resources by the user function librar
44. 818 vm 18 12262 20 Fig 12 160001 tz 0dB 19500Hz 50dI3 16bit 44 Ill PCM ATRAC encoded signal This possibly reflects the allocation of fixed bits used in the ATRAC bit allocation algorithm which implements a greater number of fixed bits at lower frequencies i e the frequency range where the eir response is most sensitive The results also suggest that the number of fixed bits at lower frequencies is always greater than th combined number of fixed ar d variable bits at high frequencies regardless of the complexity of the signal Note that the bandwidth of the system under test when recording simple waveforms can represent frequencies greater than 19kHz 4 3 2 Test 2 Objective To illustrate the reduction in bandwidth and potential for maximum spectral quantisation noise by encoding a signal which is harmonically rich over the entire audible frequency spectrum Contents of Waveforin 100Hz square wave a OdB approx 4dB full scale digital amplitude A square wave consists of a fundamental frequenc component P and an nfinite series of odd harmonics eg fV Lu This abundance of spectral components extends throughout the recoverabl bandwidth of a 44 1E Hz sample rate recording at eee considerable levels of amplitude with respect to the fundamental As such the application of a square wave can result in a useful indication of the maximum level of increased noise that is likely to occur as a consequence o
45. A Homogeneous Tandems One particular coding algorithm will be applied Al at an identical bit rate e g copies from to digital audio recorders like MD to MD A2 at different bit rates e g DAB network with ISO MPEG Layer2 contribution at 180kbit s ch distribution at 180kbit s ch emission at 128kbit s ch Class B Heterogeneous Tandems Different coding algorithms will be applied BI at an identical bit rate e g ISDN transmission with ISO MPEG Layer2 and Layer3 both at 64kbit s ch B2 at different bit rates e g ISDN transmission with ISO MPEG Layer3 at 64kbit s ch harddisk recorder with Layer2 MD Multiple Generations of Audio Coding in pe i o1 ee SS ue Fig 1 Basic principle of multiple generations audio coding Chief Engineer of Radio Timi oara a regional studio of Romanian Broadcasting Society e mail kpfaff radiotimisoara ro 51 IE OPERATIONAL PRACTICES With increasing deployment of low bitrate audio coding use of audio compression can happen at various stages contribution to production studio distribution between production facilities emission transmission of content to consumers etc Since processing transmission of audio is stil mostly done in uncompressed or even analog representation these leads to repeated cycle of decoding processing and re encoding of the audio content Similarly a change of audio coding formats and bitrates usuaiv requires a deco
46. Beniamin Dr goi Using 4 2 L Compressor in Partial Product Reduction Tree for Fast Bin ry Multipliers oo Kistari aut det ater disci iue det Nr Abe tool Avram Ionel Adrian Phycal Causes of Integrated Circuits Behaviour Degradation under Tonising Irradiation ele eet Ee e Atert le Mata ebe em EE INSTRUMENTATIE Mihaela Lascu Analog Input LabVIEW Applications for Measuring Temperature Ee Sail N ne e pi dene ui doe a Dei a Cr uL Septimiu Mischie Alimpie Ignea Geesen of an AC es using a Lab PC Data ie Board Rae A Ee gi E GE 522 Daniel Belega Septimiu Mischie Considerations Concerning the Practical Behaviour of a Digitising Oscilloscope in Different Signal Recording Modes 27 1 TELECOMUNICATII Muguras D Mocofan Corneliu I Toma New Interactive Multimedia in Model GE GE SEN in a Database c Re DR ei Ioan Snep Dan L Lacrama Corneliu I Toma Optimised Homotropic Structuring Element for Handwriting Characterey Skeleton es sete etes 3 Brun Idae Gas Acre Re ee OL Andrei Cubitchi Une m thode nouvelle pour la compression de la parole 41 Corina Botoca Romanian Bank Notes Recognition Using Cellular Neural Networks 47 Konrad V Pfaff Evaluating Objective Audio Quality for Minidisc and Influence of Multiple Generations Audio Coding for Perceptual Quality ee ST Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria
47. Cow is always 1 in both diagrams according to the function e of a compressor An analysis in terms of speed is made with the help of an analog simulator upon the layout of the circuits using MOSIS models for 0 25 um TSMC process For evaluating the real propagation speed of the carry in the most unfavourable case for both circuits the input signals x3 and x4 are applied like x2 in figure 6 and figure 7 and Cj is applied after a delay which was found for both circuits by simulation too As we can see from the diagrams presented in these figures the signal Sum is obtained faster in the case of the 4 2 L compressor than in the other case The behaviour of the two circuits is obviously different and the most unfavourable case is different for each compressor For the 4 2 L compressor the maxi mum propagation delay was found 0 53 ns while for the other circuit this delay was found 0 6 ns A comparison in terms of speed between our own variant of 4 2 compressor and several other recent variants that were published in the field literature in 0 5 um and 0 25 um CMOS technology is illustrated in the table I Table Compres 2 4 6 7 4 2L type 91 95 96 99 2001 Tech um 0 5 0 5 0 25 0 25 0 25 Delay ns 1 4 1 4 0 6 0 6 0 53 As we can notice the 4 2 L compressor is superior in speed to the other variants even if they are implemented in the same submicronic technology
48. ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Using Masked Logarithm Method for Fast Multiplication and Division Logarithmic Unit Design Lucian A Jurca Rezumat in lucrare este prezentata o metoda originala de structurare a arhitecturii unei unitati de calcul logaritmice denumit metoda logaritmului mascat ce permite efectuarea foarte rapid a opera iilor de nmul ire i mp r ire n simpl precizie Viteza de calcul superioar altor realiz ri prezentate n literatura de specialitate se ob ine n condi iile n care aria ocupat pe cip i erorile de calcul sunt mai mici dec t n cazul referintelor studiate Cuvinte cheie unitate de calcul logaritmic format LNS format virgul flotant simpl precizie INTRODUCTION The circuits for computing logarithms and anti logarithms allow the performing of the multiplication and division operations of two operands A and B by means of addition and subtraction operations A x B antilog logA logB 1 A B antilog logA logB 2 In fact through performing a logarithmic and anti logarithmic operation we accomplish the format conversion from floating point format FP into logarithmic format LNS Logarithm Number System and vice versa Thus a binary number A in FP system in single precision format is written A 1 1 0 M 2777 3 where S represents the sign bit
49. N density The sign of interface state charges is negative for TECMOS with n channel and positive for TECMOS with p channel Q and Qj charges sign have an essential value charges growing in oxide volume contribute for nTECMOS to threshold voltage reduce in a same time interface state charge growing increase threshold voltage Variation of nTECMOS threshold voltage are given by AV AQu AQit y Co 1 were AQ and AQ charge modification in volume and on interface state Co specified gate substrate capacity In TECMOS with p cannel both charges are positive This thing leads to threshold voltage increasing Threshold voltage variation of a pTECMOS is given by AVin AQu AQit J Co 2 Charges accumulation leads to punctual defects apparition in semiconductor structures determinate by chemical bonds broken lonising conditions leads to accumulation of punctual defects under structural defects In long term working under ionising condition charges are all times in bulk state This type of state leads to appearence of IC disfunctionality One way to prevent the appearence of this disfunctionality is to train integrated circuits Training consist in alternation of irradiation period 18 with nonirradiation period This treatment is based on fact that in nonirradiant period IC tends to return at initial state through charges annealing namely through charge carriers recombination This method is called itself anne
50. OR exclusive gates 2 faster than the classical ones having te onexa lt 2tpnanp Will not bring the expected advantages because the propagation through 3 OR exclusive gates in cascaded configuration no longer represents a critical path in the compressor There are lots of cases in which the capacitive load of a logical output is considered to be excessive when it appears on a critical propagation path Thus there are cases in the structure of the compressor in which an output commands three inputs i e minimum six MOS gates To the intrinsic capacity of these there is added the parasite capacity of the three signal paths III THE 4 2 L COMPRESSOR The elimination of the above mentioned disadvantages was made by the design of our own variant of 4 2 compressor presented in figure 3 which is 1096 faster than the compressor presented in 6 as proved by simulation Considering the fact that in a 54 x 54 bit multiplication array there are 5 levels of 4 2 compressors this leads on the whole to a saving of 0 35 ns in 0 25 um CMOS technology In the design of our own variant we started from the idea of finding a general definition for the 4 2 compressor whose inputs and outputs are already defined Thus this circuit has to provide the following functions a the output Sum equation 1 Sum X1 6 x2 x3 vd Cin 1 b the two outputs Cary and Cou have to be of an immediately superior order to S to have a double weigh
51. accurate SINAD estimates when the number of samples is obtained from a number of input signal cycles higher than 20 4 This method is presented in Appendix A For different input signal types the signals acquired in ENV mode are qualitatively appreciated II THE EXPERIMENTAL SETUP The experimental setup used for analysing the practical behaviour of the HM407 in RFR AVM and ENV modes is shown in Fig 1 IEEE 488 Signal Computer generator PC HM 8130 Fig 1 Experimental setup The IEEE Standard 1057 suggests the use of high purity sinewave signals for testing the digitizers 6 The sinewave generator employed for this work was the programmable function generator HM8130 7 The sinewaves used for testing the HM407 in RFR and AVM modes were characterised by frequencies situated in the range 1 20 kHz and peak to peak amplitudes smaller than 6 V The dynamic performance of these sinewaves obtained by testing permits an accurate dynamic evaluation of a waveform digitizer which have an effective number of bits ENOB smaller or equal with 8 bits In the analysis of the practical behaviour of HM407 in ENV mode were used different input signal types The PC via an IEEE 488 interface established the parameters of the signals used At the DSO was selected the AC input coupling Data records acquired by the DSO were transferred to the PC via the RS232C interface The PC contains the program which implements the acquis
52. accurate it avoids a great number of mistakes in the final character recognition stage The process consists of a recursive thinning that usually employs a homotropic set of structuring elements Experiments show that the largely known and used L structuring element does not behave well with the hand printed amp handwritten character documents images Thus the authors proposed another solution coming out from combining standard thinning and L set The results prove to be better and the improvements also echoed in the final recognition stage 40 As researches about skeleton reach the final conclusive state it became clear the need for implementing also a more flexible binarization technique in order to reduce the strokes and curves disruptions Team thought about reevaluating the results after a better binarization method will be find and adapted to the documents images specificities It is highly probable that the two preprocessing steps could be improved not only as separate entities but also as interacting stages of the image preparation for the later automatic interpretation93 REFERENCES 1 Sonka M Hlavac V Boyle R Image Processing Analysis and Machine Vision Chapman amp Hall London 1993 2 D L Lacr m Snep Manipulation of JPEG Images in Java Proceedings of the AEG Symposium Prague 26 27 October 2001 pp 308 311 3 Shapiro V Gluchev G Sgurev V Handwritten Document Image Segment
53. al de parole en une succession de segments chaq un ayant une dur e inf rieure 25 ms on obtient une s quence de signaux stationnaires Sur chaque segment le mod le de la parole peut tre de la forme Q x t gt 4 Cos gt 3 q l 42 Cette d composition ressemble beaucoup la d composition du signal x t dans un paquet de cosinus 2 La d composition du m me signal en utilisant une base d ondelettes est de la forme KL all EE k 1 1 o vir sont les ondelettes g n r es par la m re des ondelettes y t Le facteur de compression obtenu utilisant une ondelette m re sp cifi e est plus grand si le nombre de coefficients non nuls dy i gt xs vt 5 de cette d composition N est plus petit Mais dy x 07 D I Roue 0 6 00 Donc ces coefficients repr sentent l intercorrelation des fonctions x 7 et v r calcul e en origine C est la plus grande valeur de cette fonction Cette fonction mesure la ressemblance entre ces deux fonctions Donc la magnitude des coefficients d est plus grande si les signaux x r et v r sont plus ressemblants En utilisant la relation 3 on peut dire que les ondelettes les plus ressemblantes au signal x r sont les l ments des paquets de cosinus Mais si lensemble Wal est une base orthonormale alors l nergie du signal x t peut tre exprim e comme E gt al 7 K L k 1 1 Parce que l nergie du
54. aling In such a way it may prolong life time of integrated circuits in irradiant conditions Conclusions IC stability at irradiation is determinate in the first time on transistor parameters degradation as much in bipolar structures and MOS structures lonising irradiation influence analysis above TECMOS and BJT active elements show that ionising irradiation through charge accumulation determinate in MOS structures significant degradation of follow parameters threshold voltage Vin mutual conductance slope gm carriers life time lt for bipolar structures ionising irradiation determinate increasing of base current lg and decreasing of amplified coefficient h c determinate by increasing of speed recombination and structural defects called clusters 1 References 1 Avram Adrian Simularea si prognozarea function rii elementelor si circuitelor integrate sub actiunea iradierii ionizante Tez de doctor Chisin u 2000 2 R M Barnett et al Review of particle properties Phzs Lett D54 1996 3 Basumos B C Jle crBHe m3uiyuenui roiyrnposo iuukp M Du3matru3 1963 263c 4 Setz F On the disordering of Solids bz the Action of Fast Particles Discussion of the Faradaz Society London 1949 vol 5 p 271 282 5 Schleiser K M Shaw J M Benyon J M AO as a Radiation Tolerant CMOS Dielectric RCA Rev 1976 vol 37 No 3 p 350 388 6 Grove A S Snow E N A model for
55. alse PP Clear past data if Analysis is turned off Allow user to reset Histogram Parameters Fig 10 False State of the Case Structure In Measuring Temperature in Two Points I didn t use Analog Input software devices only I demonstrated with a Temperature System in Two Points a temperature aquisition and analysis application This VI reads two simulated temperatures alarms if one and or the other is outside a given range performs a statistical mean standard deviation and histogram of the temperature history concerning the two points 5 CONCLUSIONS This paper presents three modalities of acquisition monitoring and analysing temperature The first example shows a situation in which a SCXI modul an aquisition board Lab PC and Analog Input VI instruments are used The second example is only with a Lab PC board and Analog Input Virtual Instruments In many practical situations it is usefull to create a simulated example that runs without failure and only after this step it is necessary to use an aquisition board Of course when it is necessary to do acquisitions from transducers and sensors it is obligatory to use as well a SCXI modul For all the examples I created HTML files File C Fis HTML scxi 1100 html File C Fis HTML Trad TermocContin html File C Fis HTML Temp2pct html that are very usefull in analysing and comparing virtual instruments 22 7 REFERENCES 1 L Toma Sisteme de ac
56. analog converters DAC with output range 5V and E is acquired using a 12 bit analog to digital converter ADC with input range 5V The three converters are on the Lab PC Data Fig 5 Block diagram of algorithm Acquisition Board This board is see as an input output device for the uP microprocessor of the IBM PC wu t w Asin t w5 Acos at 6 computer and this controlls the Lab PC or in complex representation The Lab PC has 28 input output registers which have consecutive addresses The address of first register or base address was chosen 260H The U wU jmU 7 From 5 and 7 follows that microprocessor reads or writes these registers and thus Zx R wi jw2 8 the three converters can be controlled or he bridge bal oe 9 B The bridge balancin S 5 X Rw The bridge balancing is equivalent with modifying The reference voltage U is achieved as a digital U magnitude B and phase so E 0 For this purpose sinus is used the LMS Least Mean Square algorithm Fig 5 a2 qe alg 8 asai E 10 n 24 where n is the number of points per period The weights w and w at time k 1 can be calculate using the LMS algorithm as follows 4 i ijs 7 c E n 11 27 w2 k 1 Wo k uEA cos k n where yw is the convergence factor which controls the stability and adaptation rate C Software structure The program has been written in C language First the values of the U was calculated in
57. ation and Analysis Pattern Recognition Letters North Holland vol 14 No 1 January 1993 pp 71 78 4 Deseilligny M P Stamon G Suen C Y Veinerization A New Shape Description for Flexible Skeletonization IEEE Trans on PAMI vol 20 No 5 May 1998 Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Une methode nouvelle pour la compression de la parole Andrei Cubitchi Resum On pr sente une m thode nouvelle pour la compression de la parole qui est meilleure du point de vue du facteur de compression et de la transparence que la m thode de compression utilis e en GSM Cette nouvelle m thode de compression utilise une transform e en paquets de cosinus adaptative Elle r alise la s quence d op rations suivante le calcul de la transform en paquets de cosinus du signal comprimer la s lection adaptative des plus importants coefficients obtenus et la quantification adaptative des coefficients obtenus ainsi Pour la reconstruction du signal de d part en utilisant sa version comprim e ont t consid r s des op rations inverses Bien sur la reconstruction n est pas parfaite mais les distorsions introduites par la compression sont n gligeables donc la m thode consid r e est transparente Le but des op rations adaptatives est de maximiser le facteur de compression
58. carried out In fact there is often no reason for leaving the coded domain and going via PCM e g for copying purposes Even when different algorithm are involved transcoding the conversion in the coded domain can probably improve the outcome of tandem coding If however further processing of the signal is required such as level change equalization or reverberation a return to the PCM domain is forced Unfortunately provisions for interfacing in the coded domain are not yet widely available today If decoding re encoding of the compressed audio content is necessary it must be clear that a degradation in signa quality will happen thus in oreder for the final coded audio to meet a desired targel quality the coding quality at intermediate 58 coding steps must be significantly better than the target quality In this way quality losses due to tandem coding can be compensated by increased coding quality and required bitrate in intermediate coding steps 7 Data reduction systems necessarily reduce the amount of data so it stands to reason that multigeneration processing will lead to quality loss The MPEG algorithms seem to be slightly less prone to this but it remains a general problem with all systems Things are worst when different data reduction system are cascaded in which case the quality deteriorates very quickly indeed The symptoms usually become apparent after two or three generations in the form of vague ste
59. carry on the horizontal in the compressor block from figure 2 is limited to a single bit single position to the left The vertical propagation time of the whole block is actually the vertical propagation time through one compressor This obviously means that the reduction to half of the number of partial products is made in a time equal to the propagation time on the critical path through the 4 2 compressor For this reason the fast propagation of signals through this circuit is essential and practically determines the propagation speed through the whole Wallace tree The optimised 4 2 compressor described in 6 and presented in figure 1 has a series of limitations which affect the propagation time of the carry 12 The 3 input NAND gates in advanced submicronic technologies which need smaller supply voltages can no longer be implemented or they are implemented through two 2 input NAND gates by the addition of one more logical level In this way the critical propagation path through the circuit becomes the one which includes the fast output Cout of the current compressor and the input Ci and the output Ca of the left hand compressor So this path becomes slower compared to the propagation time for x1 and x2 figure 1 through 3 successive OR exclusive gates Even in technologies which allow the integration of 3 input NAND gates its propagation time is bigger than through a 2 input gate The use of unconventional
60. catiilor multimedia ce permite 0 bun reutilizare a codului scris si in alte aplicatii si stocarea obiectelor utilizate in baze de date Cuvinte cheie multimedia tehnici de programare aplicatii interactive I INTRODUCTION This paper present a new interactive multimedia application model witch combine the advantages of object oriented programming with aspect programming The model proposes a structure for multimedia applications with a good reutilization of code in other applications and objects storage in databases II THE ARHITECTURE OF MULTIMEDIA APPLICATION MODEL In the multimedia applications model architecture are four modules having a very strong interaction Every module is composing by other small modules and these small modules can be used very easy in other applications I use four categories of modules for the application kernel The application kernel generally is a specific part for a multimedia application is not possible to reutilize the code and for users is a close part don t accept modifications This kernel is the manager of entire application He use the functions procedure routines witch are stored in modules In many cases the module is the equivalent of a functions library One of the kernel jobs is to verify and to validate all the new functions implemented by the users The architectures modules categories are Objects processing functions User interactions Input Output Op
61. ch shorter The global time spend to realize an entire multimedia application 1s generally shorter If is used the real time content generation concept for pages the time of physical implementations degrease very well For example to realize a multimedia presentation for a company witch produce hundred goods is necessary 36 to make hundred pages one page for every product If all the information s about the product is stored in a database the implementation task for all pages is to make the model for a single page and after that to connect this page to database is not necessary to realize hundred different pages In this dynamic way the reduction of the implementation time is considerable In plus the size of the executing file is smaller because the object are loaded only when are used This paper present a new interactive multimedia application model witch combines the advantages of objects oriented programming with aspect programming The model proposes a structure for multimedia applications with a good reutilization of code in other applications and objects storage in databases Using this model is possible to develop very performing multimedia applications BIBLIOGRAFIE 1 Kiczales G Hilsdale E Hugunin J Kersten M Palm J Griswold G W Getting Started with AspectJ Raport of Defence Advanced Research Projects Agency F30602 C 0246 2001 Mocofan M Pescaru D Parallelisation of a textu
62. control logic to allow the implementation of both addition and subtraction In this way the adders which provide the explicit values of the logarithms of the mantissas for the two operands were practically eliminated These values become hidden being contained implicitly by the 4 partial pseudo products and they contribute to the process of obtaining the final result That is why this operation method was called the masked logarithm method The implementation of this method leads to the generating of a big Wallace tree which has 30 pseudo partial products and which has its lower end in the ALU We used as reduction blocks for the pseudo partial products 4 2 compressor blocks and CSA blocks of 3 2 full adders in such a way as to obtain the most efficient structure in which there are no unoperated intermediary pseudo partial products on any of the levels of the tree The resulting structure is presented in figure 4 Y2A AEyA O Control Y2B AEy8 0 Control bit Inverter Cmpria 2 En bes Cmpri4 2 Fig 4 Wallace tree for pseudo partial product reduction The CSA block groups up the terms y and AE with the most significant partial product of y2xAE in order to avoid the excessive growth of the size of this block and so the size of the blocks connected after it as well as that of the wiring For the antilogarithm computation circuit the procedure applied is the same the data words
63. corresponding memory locations Baring in mind that the error in floating point single precision format i e the value of the least significant bit provided by any output of the ROMA or ROMC is 1 2 x 107 it means that to the calculated values of E we can add corrections of one LSB after which they are directly memorised ROMA respectively they are transformed in two s complements and then memorised ROMC Theoretically the total conversion error can be reduced to half if to the memorised value E we add an amount equal with the half of the maximum conversion error i e 1 5 x 107 This amount can not be represented on an integer number of bits which could have been 1 or 2 The compromise is thus the adding of 1 one LSB in those locations for which the conversion error is bigger than one LSB i e bigger than 1 2 x 107 In this way the conversion error both for logarithm and for antilogarithm computation decreased to 1 8 x 107 As far as the computation of the product AEyxy is concerned we can notice from figure 3 in which the multiplication area is presented that if we perform the calculation of the truncation error in the most disadvantageous case when all bits of a smaller or equal order with 28 the bits of the right side of the vertical line are equal to 1 we obtain the value 2x107 3x10 4x107 amp 5x107 610 710 8x107 9x107 5 96x10 This value represents half of the representation err
64. ction with a greater value than the maximum value o the interest interval interval Reconstruction Fig 3 The color filtering flow chart In the second step there are converted in black all the objects that have a greater value than the minimum value of the interest interval of colors and all the objects that have a greater value than the maximum value of the interest interval of colors In the third step it is processed a XOR operation between the images resulted in the second step Finally there are reconstructed all the objects that are in the desired interval interest and eliminates all the others In the first step is performed pre filtering of the image In Fig 4 there are presented the effects of color filtering on the 10000 lei banknote Size classification subroutine selects from the output image of color filtering all the objects that have a dimension between an interval of pixels and deletes all the other objects The steps of the size classification subroutine are illustrated in Fig 5 From its output image the subroutines objects containing many holes extraction and objects with one hole extraction will eliminate all the objects that have more than one hole and other uninteresting characteristics and will retain only objects with one hole The two images obtained at the output will be applied as image a and b to the reconstruction subroutine as it can be seen in Fig 6 The output image will contain the group
65. ctronica Aplicat Bd V P rvan Timisoara 1900 e mail adiavram hotmail com 17 oxide in electrodes At the same time take place less mobile carriers capture holes in oxide by trapped centres that are formed in broken bonds process between Si and O atoms Trapped holes process continue until take place the compensation of external electric field intensity with induced irradiation field of charges in Qo volume and ending of electron hole pair divide process These models permit a qualitative explication of saturation effect of Qa D that was experimental observable in dependence of dose were D absorption dose Also these models permit to understand the growing effect of Qa at saturation in condition of positive potential Vs growing on metalic electrode of irradiated MOS structure 1 Holes that not be trapped slowly begin to move to negative electrode Surface state forming process take place in two phase in phase one holes are formed that will be capture by traps and in second phase take place positive ions movement to Si SiO interface After second phase interface state are formed The problem of forming interface state have a principial contribution because interface forming process take place after ending irradiation process and after forming Qu charge 5 Electrons and holes are separated in oxide by electric field and positive charge Qa appear in oxide volume with Ns density and interface state charge Qi with
66. ding re encoding cycle for format conversion The introduction of MPEG coding into broadcasting brings benefits and problems We are able to do things that were previously impossible but there are things that we used to do that are made more difficult The case where equipment using MPEG compression has to be integrated with existing conventional equipment presents one kind of problem to be overcome The other problem is to have equipment that can perform processes on MPEG compressed signals that would normally be associated with processing in the uncompressed domain It is the conversion from MPEG format into uncompressed signals that introduces impairements into our signals Keeping them in MPEG coded form preserves the quality This is similar to the situation with analog and digital signals Once converted to digitel it is conversions back to analog that introduce impairement Keeping the signals in digital form preserves the quality provided that we can perform in the digital domain all the operations to which we are accustomed in the analog domain 3 When coding previously coded audio using modera psycho acoustic based coders the audio quality deteriorates The reasons include the packet or frame structure imposed on the audio the time to frequency transformation applied to the signal and the number of bits allocated to each part of the spectrum which is detertnined by a psychoacoustic model PCM bitstream PCM bitstream PCM
67. e frequency and time domains the progressive signal degradation due to ATRAC encoding over multiple generations Contents of Waveform I6kHz sine wave at OdB masking tone 19 5kHz sine wave masked tone at 70dB relative to the masker 5 Synopsis of results The effect of multivle generation recording using ATRAC encoding is shown to result in a progressive degradation of the masking tone level from the first generation 10dB digital transfer 16dB analogue transfer The total level degradation by the tenth generation digital transfer is about 24 for the Sharp MD and 18dB for SONY MD In the case of analogue transfer after three generation the level dropped down by 3CdB An interesting observation is that of a level raising of the higher frequency masked spectral component frem the original 70dB to about 30dB 35dB shown in fig 25 After ten generations this level degrades only with 5dB 40dB digital transfer The increase in spectral quantisation noise by the sixth generation is about i2 dB fig 27 compared to the PCM original fig 24 In addition to the change in spectral content analysis of later generation waveforms in the time domain displays significant glitching artefacts i e signal peaks of excessive amplitude with respect to the average overall waveform amplitude that are unrelated in time placement or frequency to the original signal content These are shown in the time domain graphs of figures 26
68. ection method based on the memorisation in the same points of the values of the function derivative too after which the interpolation is performed 1 2 Thus we note y OM and we partition the mantissa y in two parts y4 containing the most significant 11 bits and y containing the least significant 12 bits The values of the function log 1 y y in these 2 2048 points are memorised in internal ROM ROMA as correction values E provided through the application of the address y1 Thus we obtain the following approximation log 1 y y Ey AE x y 6 The second look up table ROMA of the memory is much smaller because AE lt lt Ey Considering for AE a 12 bit representation the complete conversion between the two formats is made through a reading in the look up tables two 23 bit additions and a 12x12 bit multiplication Electronica Aplicat Bd V P rvan Timisoara 1900 e mail jlucian ee utt ro The calculation of the antilogarithm is made in the same way Considering C the result of finding the antilogarithm we can write C DE 127 0 M DE 127 2 7 where E represents the integer part in LNS format and M represents the fractional part In this approach we do not extract the bias value from the exponents of the two operands For this we extend the ALU with one bit to the left while the bias value is extracted or added to the resulted exponent depending on the performed operation multiplication or d
69. ed situation of measuring the temperature in two points Of course this structure can be extended for measuring the temperature multipoint 21 mH igal Themoneervi i HU pe E S enpate Hein tn HE rea eal mom p P did xia bit IES m d ia Leg T i i bp br bei Disabled Low Lin Konst Ec E Sa M tee Mean Tena prow enm mim H us SEP Pat virent fDisebled TEE e Ead p HTML Fie Path arom Char Bereta happend b Fame f Cie fa image Foma us Image Depth m i ipse Fein t p x E i Fig 7 Connector Pane for Measuring Temperature in Two Points Current Temperature Temperature History Pp Pe Statistics Se 1 I U 1 1 H H Minimum bin Maximum bin 70 00 2130 00 i i i i i i Fig 8 Front Panel for Measuring Temperature in Two Points D Acquistion Loop Analysis Stuctue b IN Tne PI a Fig 9 Block Diagram for Measuring Temperature in Two Points eid F
70. equence 0 determines the offset of your SCXI 1100 module and plug in data acquisition board It passes this amplifier offset to the Scaling Constant Tuner VI in sequence frame 2 The string CALGND is used in the offset channel string to tell LabVIEW to ground the amplifier inputs and then measure the offset Several binary readings are acquired and averaged to produce a single amplifier offset value Fig 3 1 Block Diagram Sequence 1 for SCXI 1100 Eo Oe PD previous error cluster j Pass on error cluster The voltage read from the temperature sensor on the SCX terminal blocks in units of volts Sequence 1 reads the CJC Cold Junction Compensation voltage Several readings are taken from the sensor on your SCXI terminal block indicated by the string MTEMP These readings are averaged to yield one value and passed to the next sequence CIC Sensor The temperature sensor voltage will range from 0 5 to 0 5 if IC sensor 2 to ZV if Thermistor cold junction voltage Fig 3 2 Block Diagram Sequence 2 for SCXI 1100 2 0 2 binary amplifier offset from frame O L HER colin iu Kanech DT buffer size LAS input limits 77777 in deg C 50 to0degC N number of chansc Less ON previous error cluster LE 7 H device 1 SCH cono acq scan rate engt clusterc TC type J scan rate number of sa
71. er save and exit menu items and a display image place as shown in Figure 2 1 Consequently W class also has the methods for performing opening saving and exiting E Imports com sun image codec jpeg ava awt java awt image 5 java io t javax swing ERR Sb Wy ImageFilter S Sb calc int v proctimage im An Jw EX w p JFrame B Runnable wi String titlu action Event e Object arg main String aras paint Graphics d e rund M cont Fig 2 2 The program class structure as shown in Borland JBuilder 4 environement s Structure Window The Sb class is focused on the thinning and contains the set of if else if statements imple menting the Hit or miss procedures needed for 39 finding the pixels to be changed from letter pixels to background ones The sequence of successive thinning is realized through the Thread mechanism inside a while statement It stops when Sb report no more changes has been done over the last thinning cycle The contour is also declared incremented and displayed within the W class shows the number of cycles in real time In the end its figure represents the skeleton stage thinning cycles needed for the currently processed image Final image saving has been done using the method in 2 It gave the possibility to preserve in JPEG format the skeletonized images for the further processing steps IV EXPERIMENTAL RESULTS
72. erations Integrated technologies The multimedia applications model architecture is presented in Fig 1 A Objects processing functions For Objects processing functions are two main categories Temporal processing the temporal aspect of objects is used in processing Objects processing for this category are two types of objects processing Single object processing Multiple objects processing In a multimedia application we run in the same time different sounds video sequences and animations The applications run correctly when are free hardware resources s necessary to manage very well the application and the hardware resources this can be realized by many temporal references The Time Line routines must to by very simple and easy to use for a user with medium knowledge in the field of programming The main jobs of the Time Line routines are Management for time constraint objects and for the hardware resources used for playing that Verification if the hardware resources are free to use If is not free for use is necessary to stop playing the current object Temporal references list construction for effects and transitions between objects Management of temporal references jumping in collaboration with sub modules GUI and Events Here is useful to use the aspect programming techniques 1 The very good results are achieved if are mixed the object oriented programming techniqu
73. ers these parameters describe the object its dimensions the way of compression the application that generated the object useful for example in the number of colors dimensions resolution images etc 4 Global parameters witch describe the content of objects thru many features These parameters are useful in content database searching 5 User parameters specified by each user and useful only in some applications Multimedia Object Multimedia Object Structure Component values Multimedia object interpretation Complex Simple Objects Object Sequential Sequential Animation Video Sound Sunet Image Sen Image ane Video Animation Multimedia Text Sound Image Video Animation Fig 2 Multimedia Object Descriptive Model After describing the object follows the description of the primary elements witch are compose the object The general parameters of description of the primary elements are Primary_Element Parameters Primary_Element Identification Primary_Element General Information s General Attributes Primary_Element_Global_Features Parameters Primary Element Description Specify Information s Attributes for Content Object Description of E Primary_Element _User_Features Parameters User ID Primary Element Details User Personal Information User Attributes for Content Description of Object
74. es for multimedia objects description and aspect programming techniques for temporal management of the events Object Processing Functions for reducing the space that a multimedia application requires in the context of increasing the power of calculation I propose the utilization of on line processing concept That Facultatea de Electronica si Telecomunicatii Departamentul Comunicatii Bd V Parvan Timisoara 1900 e mail toma hermes ee utt ro Facultatea de Electronica si Telecomunicatii Departamentul Comunicatii Bd V Parvan Timisoara 1900 e mail mugur mail opendrum utt ro 31 suppose the processing of the objects is made during the running of application So it can be used only one physical object in many apostasies in several scenes it can wear a variety of forms it can be scaled it can be rotate or deformed etc during the running of an application If the application uses a very large number of objects witch suffers the same processing methods handling this objects can be simplified by using the idea of processing directly the information and not store it We can meet two situations the new generated object can be stored the handler can personalize the multimedia objects starting from a model object or the object can be destroyed after utilization if we need it again the object can be regenerated If the processes are very complex there is the risk of loosing a lot of time wi
75. ession de la figure 1 est le facteur de compression f Celui ci repr sente le rapport entre les nombres de bits des signaux x n et vn Le d tecteur adaptatif de seuil et le quantificateur adaptatif de la figure 1 sont con us pour maximiser f en gardant snr sup rieur 20 dB III L UTILISATION DE LA TPC POUR LA COMPRESSION DE LA PAROLE l y a un grand nombre de transform es qui peuvent tre utilis es pour la compression des donn es Le choix d une transform e particuli re doit tre fait en accord avec la sp cificit du signal comprimer Ce travail est concentr sur les transform es en ondelettes orthogonales utilis es pour la compression de la parole Le signal de parole a un mod le math matique qui fait l utilisation des paquets de cosinus tr s utile pour les applications de compression A Le mod le math matique de la parole Chaque phrase est une s quence de tons qui ont diff rentes intensit s fr quences et dur es Chaque ton est un signal sinuso dal avec amplitude fr quence et dur e sp cifiques C est le mod le sinuso dal de la parole Une description math matique pour ce mod le est ole x t 2 Aq cos 7 2 q 6 o les composantes sont nomm es partiales Chaque terme de cette somme est un signal double modulation Donc il ne s agit pas de signaux stationnaires Mais fr quemment la parole est consid r e comme une succession de signaux stationnaires Divisant le sign
76. f ATRAC encoding Synopsis of results Two immediate differences occur between the reference fig 13 signal and the ATRAC encoded signal Primarily a high level of quantisation noise 50dB relative to the reference signal is spread across the entire frequency spectrum as shown in fig 14 Additionally the bandwidth of the signal is truncated by almost 4kHz at the high end of the spectrum Fig 14 100 z PC n square wave at Odi ATRAC encoded signal This suggest that total number of bits available during the ATRAC bit allocation process is fully required to represent the square wave and the number of bits allocated to the higher extremes of the frequency spectrum is sacrificed to accomodate as much of the preceding bandwidth as possible 4 3 3 Test 3 Objective To attempt to demonstrate the allocation of bits performed during ATRAC encoding with respect to sub band frequency components and their amplitudes Contents of Waveforms Waveform I 400Hz Masking tone 0dB 150Hz 250Hz 355Hz 450Hz 560Hz 720Hz 1kHz IkHz 1 6kEHz 1 9kHz 2 1I kHz 2 5kHz 2 9kHz 3 4kHz all at 60dB Waveform 2 400Hz Masking tone 0dB 150Hz 250Hz 355Hz 450Hz 560Hz 720Hz 1kHz 1 1kHz 1 6kHz 1 9kHz 2 1 kHz 2 5kHz 2 9kHz 3 4kHz all at 40dB Waveform 3 400Hz Masking tone 0dB 150Hz 250Hz 355Hz 450Hz 560Hz 720Hz IkHz 1 1kHz 1 6kHz 1 9kHz 2 1kHz 2 5kHz 2 9kHz 3 4kHz 3 7kHz 4 4kHz S 8kHz 6 4kHz 7 7kHz 9 5
77. follows 7 10ms or f 1 T 100Hz That is impedances will be measure at a frequency of 100 Hz and for the acquisition of E the ratio F F is 100 Also for inductive impedances measurement because L had values of order of 0 1 2H in order to increase Z frequency of U and U was 200 Hz by appropriate programming of 8253 circuit 25 Another important problems was values of impedances from the two arms of bridge and amplitude of reference voltage U First because maximum load current of each DAC s is 2mA follows that minimum value of sum between unknown impedance and reference impedance must be about 5kQ Second the value of reference impedance R must be choose so the bridge be near of the balance i e value of E will be maximum of 5V amplitude in order the ADC not will be saturate Also due to 9 R and R or X and R must be of approximate the same order of magnitude If w or wz have values very different of 1 for exemple highest of 2 5 that is U from 6 can result highest of 5V amplitude From these observations amplitude of U must be setting to maximum middle of output range of DAC Third in order that the bridge has maximum sensibility Z and R and then U and U must be have the same order of magnitude From these many constraints authors must select one middle solution Thus 1 value of reference resistor R is choosen function of approximate value of Z 2 amplitude of U is choosen about to 2 5V
78. ftware Environment ultimedia Application Public Library Multimedia Environment Public Library Multimedia Application an E o a E o gt vo a Dn E E Q 2 2 lt a mi Ki o E 5 z Objects Database B Multimedia Applications Developers Fig 4 Multimedia Environment in the model context For the objects storage can be utilize a standard database using a generalization of the ODBC concept This model realizes a dynamic content for multimedia applications The pages included in applications are generated dynamic from a database in real time this pages exist if the application run and are generated in concordance with the user options Using the client server technology many users acces the application in the same time using remote commands I propose to include in every multimedia environment tools for creation of new functions library In this way is possible to develop the multimedia software environment V CONCLUSIONS In this days is very important to have an open environment for multimedia applications development because are many application developers with a good experience in this field and this experience can be use for the functions library development Fig 4 From design point of view the time necessary to realize a good concept and the modeling for application is much longer From implementation point of view the time for creation is in this case mu
79. he test outcome Once recorded onto MD the test signals were then transferred back to the PC using the optical digital output S PDIF of a SONY MDS JB920 and saved as new 16bit 44 1kHz sound files WAV format The new sound files of the coded signals were subject to Fast Fourier Transform FFT analysis using SoundForge 5 0 The original source signals were also analysed for comparative measures using the same software package PC Waveform Generator and FFF Analyser V MD Recorder Sharp MD722 gt 7 Soundcard C H SB Live 4 Paya SONY i SPDIFIN d EEr Fig 4 General set up for transfer of objective test waveforms The multiple generation test was made also in two more ways a analogue transfer to and from the Sharp MD722 to see the effect of D A and A D conversion the analogue circuitry on the test outcome as shown in fig 5 PC Waveforn Generator and FFT Analyser MD Recorder Sharp MD722 SB Live daa Line IN Fig 5 General set up for analogue transfer of objective test waveforms b digital transfer for playback and recording using oaly the SONY MDS JB 920 as shown in fig 6 PC Waveform Generator and FFT Analyser recens D ae apt te Ps a SONY meee P SPDIF IN Fig 6 General set up for digital transfer of objective test waveforms 4 3 Basic Waveform Tests 4 3 1 Test 1 Objective To illustrate the increased quantisation noise of perceptually masked
80. hizitie si prelucrare numeric a semnalelor Editura de Vest Timisoara 1996 2 A Ignea M surarea electric a m rimilor neelectrice Editura de Vest Timisoara 1996 3 National Instruments LabVIEW User Manual 1999 4 National Instruments LabVIEW Measurements Manual 1999 5 National Instruments LabVIEW Development Guidelines 1999 Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Implementation of an AC Bridge using a Lab PC Data Acquisition Board Septimiu Mischie Alimpie Ignea Abstract The paper presents implementation of an AC bridge based on a LAB PC data acquisition board The balance of the bridge is made using the LMS algorithm A real time implementation of the bridge is presented and tests results obtained are incorporated in this paper Index terms balance of AC bridge virtual instrument adptive LMS algorithm I INTRODUCTION Usual an alternative current AC bridge for impedances measurement contains four impedances When the bridge is balanced the unknown impedance can be computed function of the known impedances This method has the disadvantage of manually balanced and this process can have a long time Also the price can be expensive because the tree impedance must be adjustable and accuracy Fig l AC bridge with two sinewaves ge
81. ies witch allowed searching object and resources witch are used in the multimedia application etc Dynamic Data Exchange DDE allowed the construction of client server applications DDE allows data actualization in other application and makes easier the utilization of remote commands Dynamic Link Library DLL use the extern files that contains routines and methods witch are used by Windows applications Object Linking and Embedding OLE is a techniques witch allows creation of objects and this objects can be exported and reused in other multimedia applications Open Database Connectivity ODBC Microsoft proposed the concept for generalizing the way of communications with databases I suggest a generalization in using this concept so the multimedia applications can use this way of connecting for accessing their objects from a databases The ODBC technology allows a commune interface for accessing heterogenic SQL databases SQL language is a standard of accessing databases This interface gives a higher level of interoperability one application can access a variety database system management witch use SQL language This allows users to realize and distribute a client server application without restrictions from a system databases management 3 MCI an operating system has to allow the use of multimedia resources for various applications witch runs in the same time Is necessary that their data stream to be synchron
82. iguration the number of channels available on the DAQ device limits the total number of analog input channels In some cases however you can cable more than one DAQ device to separate modules in an SCXI chassis When you want LabVIEW to acquire data from SCXI analog input channels you use the analog input VIs in the same way that you acquire data from onboard channels You also read and write to your SCXI relays and digital channels using the digital VIs in the same way that you read and write to onboard digital channels You can write voltages to your SCXI analog output channels using the analog output VIs III MEASURING TEMPERATURE WITH THERMOCOUPLES If you want to measure the temperature of the environment you can use the temperature sensors in the terminal blocks If you want to measure the temperature of an object away from the SCXI chassis you must use a transducer such as a thermocouple A thermocouple is a junction of two dissimilar metals that gives varying voltages based on the temperature However when using thermocouples you need to compensate for the thermocouple voltages produced at the screw terminal because the junction with the screw terminals itself forms another thermocouple You can use the resulting voltage from the temperature sensor on the terminal block for cold junction compensation CJC The CJC voltage is used when linearizing voltage readings from thermocouples into temperature values To read the
83. implementation the CNN provide an efficient solution Recognizing bank notes requirements to avoid counterfeiting bank notes orientation of the image or background to permit copying any other documents has to satisfy two in any III THE RECOGNITION PROGRAM The recognition program consists of many algorithms according to the flow chart from Fig 2 e color filtering that generates the set of suspected objects from the image size classification objects containing more extraction e objects containing more concavities extraction objects with one hole extraction reconstruction of the extracted objects Each module from Fig 2 is in fact a subroutine performed by a sequence of elementary locally operations named templates Some of these subroutines will be presented in the following Input image than one hole Size classification Objects with Objects containing many holes one hole extraction Sum of subresults Reconstruction Fig2 The flow chart of the recognition program for the 10 000 lei banknote The color filtering subroutine flow chart is illustrated in Fig 3 Each pixel of the image has a value related to its color in 1 for white and 1 for black interval The color filtering selects all the objects that have colors values between an interval of 48 Prefiltering Objects extraction with a greater value than the minimum value o the interest Objects extra
84. in 9 5 kHz and Vo 0 V In Fig 5 a the number of recordings used for averaging was M 128 and in Fig 5 b was M 512 By comparison of the Modulo Time Plots of the residual errors presented in Fig 5 with the Modulo Time Plot of the residual error shown in Fig 4 results that the averaging reduces the noise As consequence in AVM mode SINAD and ENOB increases and so the performance of the HM407 increases From Fig 5 results that the dynamic performance increases when M increases Therefore we recommend the use the maximum available value for Modulo Time Plot Amplitude ZFS Frequency kHz Phase deg GA Offset mV bution Function SINAD dB ENOB THD dBc Amplitude ZFS Frequency kHz Phase deg E i 0 0 02 0 04 0 08 0 eta Offset mV Nb of cycl ution Function SINAD dB ENOB THD dBc Fig 5 The dynamic performance obtained with a M 128 b M 512 IV THE BEHAVIOUR OF HM407 IN ENV MODE Fig 6 shows the signals acquired in ENV mode for different input signal types as a function of time for a vertical gain of 500 mV div and with a HM407 sampling rate f 1 MHz The input signals were characterised by an amplitude A 4 V fi 5 6 kHz and Vo 0 The number of samples acquired was M 2048 Both plots presented in Fig 6 were obtained by means of MATLAB program In both plots shown in Fig 6 is evident that the input signal amplitude and frequency vary These plots
85. ing 0 5 0 5 0 5 0 0 0 A 105 4 05 Best 4 04 I 2 5 05 O35 0 5 0 0 0 Fig 7 Object reconstruction based on signature a the original image b the signature c reconstructed object The processing time increases proportional with the size of the reconstructed object V EXPERIMENTAL RESULTS The recognition program was implemented in MATCNN a toolbox of Matlab for CNN It was tested on different Romanian banknotes on the 5000 10 000 50 000 and 100000 lei banknote with different backgrounds In some situations it was enough to apply only some of the subroutines from Fig 2 to obtain the recognition characteristics If the presence of an banknote was confirmed in the input image the recognition characteristics see example from Fig 1 were generated at the output So a copy machine equipped with a a CNN will stop functioning whenever it detects a fraud tentative VI CONCLUSIONS The significant features of Romanian banknotes were selected it is presented a method that use the local logic of CNN and developed a program in Matlab capable to prevent them to be copied by color copy machine The algorithm benefit from the high speed of CNN and a possible hard implementation that can offer a real time operation VII REFERENCES 1 L O Chua L Yang Cellular Neural Networks Theory IEEE Transactions on Circuits and Systems vol 36 nr 10 octombrie pag 1257 1272 1988 2 J Miller T Roska T Szir nyi K Crounse
86. ing with all structuring elements in this paper is the same thus the comparisons among them is made on an objective basis The L structuring element commonly gives better skeletons approximation when it is employed with handwriting but it is slower The M structuring element is always quicker but the curves are coarser and disconnection errors occur 4 bog Pss WU ta a b Fig 1 3 a Image processed with M structuring element b Image processed with L structuring element The slow quick discussion generally tend to become obsolete with the processing speed and the vast resources of actual computers Therefore the L element technique is preferred in almost all cases The problem resides in the fact that the L structuring element does not behave quite well with the handwritten documents processing and the resulting skeletons are still coarse and sometimes thorny II THE H STRUCTURING ELEMENT Therefore the authors tried to improve the procedure performances employing an optimized structuring element resulting from combining the standard skeleton structuring element and the L structuring element We started from the straightforward simplification of the standard structuring element 1 1 Each pair of matrixes can by mathematically described by a common model as shown in 2 1 1 1 1 1 0 0 11i fa iE bL ob d Ss 2 1 000 100 Consequently the eight matrixes in the initial set
87. ion times on the pipeline stages should be as close as possible Through this new approach the terms ya Eya respectively yg Eyg which are added to the products AEyA x Van respectively AEyg x yog see equation 6 will be considered as initial partial products of the above mentioned products These terms will thus be introduced in the Wallace tree for the reduction of the number of partial products beside the 12 initial partial products of each product In this way we eliminated the final multiplication adders and the two resulting terms in the lower end of the trees in the sum and carry sections will implicitly contain the product value too distributed between the two sections Using this redundant representation the proper values of the two products AE xyz will no longer be found in the structure explicitly thus becoming masked values The problem which is still to be solved is that of the situation in which in equation 6 the term AE x y2 is negative and its two s complement conversion i e of all the 12 partial products would be necessary This happens starting from the address 907 to 2047 of ROMA and ROMA situation which corresponds to the negative slope on the diagram of the function log 1 y y presented in figure 1 We managed to avoid this severe shortcoming through an artifice which allows the total elimination of the cases in which the product AE x yo must be subtracted As shown in figure 2 we can write
88. ions THD ENOB was estimated by the method proposed in 1 eq A 4 from Appendix A THD was estimated by the Interpolated FFT algorithm 5 Also are presented the parameters of the best sine fit that corresponds to the digitizer output signal When M is a power of 2 for correct estimation of the dynamic performance it is necessary to overcome the situations in which the number of input sinewave cycles is equal or very close to a number which is an integral power of 2 In these situations are not tested all the digitizer levels The performance shown in Fig 4 were obtained in the same conditions like the ones used in the case of Fig 3 but with f 200 kHz and fin 6 25 kHz Modulo Time Plot Amplitude ZFS Frequency kHz Phase deg 70 0 0 15 0 2 Offset mV Nb of cycles SINAD dB ENOB THD dBc Fig 4 The dynamic performance obtained in the situation in which the number of input sinewave cycles is 64 From the Modulo Time Plots is clearly evident that in this situation remains digitizer levels which are not tested If are not tested all the digitizer levels we are not sure that the performance obtained rest the same Therefore these performance corresponds only to the digitizer levels tested and not to the digitizer Fig 5 shows the dynamic performance obtained in AVM mode for a vertical gain of 50 mV div and with a HM407 sampling rate f 400 kHz The sinewave test signal parameters were f
89. iourilor de radiodifuziune de la contri butie inclusiv inregistr rile primare din teren prin studioul de produc ie p n la distribu ie i emisie a ap rut problema cod rilor multiple succesive sau in tandem cum mai sunt denumite ntr un lant digital de transmitere a materialelor audio pot exista combina ii diferite de scheme perceptuale de codare succesive S a observat c n func ie de algoritmii si debitele binare fo losite calitatea final a sunetului poate fi serios afectat n partea experimental a lucr rii a fost evaluat obiectiv calitatea audio a cod rii pentru algoritmul ATRAC minidiscul fiind n prezent principalul instrument pentru nregistr rile pe teren L INTRODUCTION The central objective in audio coding or audio compression is to represent a digital audio signal with a minimum of bits per sample while keeping transparent not distinguishable from the criginal under given listening conditions The rapid acceptance and deployment of low bitrate audio coding systems in the broadcast transmission network has raised new issues with respect to maintaining the audio quality of the original program signal Tandem coding is one of the most important issues related to audio performance in the low bitrate digital network Quality degradation due to transcoding between different formats can be avoided by adhering to a common system and by following the ITU R International Telecommunications Unio
90. itical propagation path is also shorter In principle a 4 2 compressor is obtained through the interconnection of two 3 2 full adders whose structure is presented in figure 1 doubly and which presents a propagation time equivalent to 4 OR exclusive gates in cascaded configuration A major improvement of the structure of the 4 2 compressor was brought by D Villeger in 3 and 6 by the exploiting of the fast inputs and outputs of the 3 2 full adders in such a way that the maximum propagation time becomes the equivalent of only 3 OR exclusive gates in cascaded configuration This optimised structure is presented in figure 1 For this compressor the 4 input bits are x1 x2 x3 and x4 while the 2 output bits are Sy and Cary The carry in is Ci and the carry out is Cou The Cary and Cox bits have a double weight in comparison with Sum The Sum bit is obtained through the implementation of the OR exclusive function between the 5 inputs of the circuit Supposing that the propagation time of the carry through an OR exclusive gate is twice bigger than through an NAND or NOR gate we can notice in figure 1 that indifferent of the propagation path of the input signals including the Cn generated by the right hand compressor the maximum propagation time is not longer than the equivalent of 6 simple logical gates or 3 OR exclusive gates We can also notice that the output Cout does not depend on the input Cn and so the propagation of the
91. ition process and the program package presented in 4 In this program package is implemented the method proposed in 1 For this method the program package provides a number of graphical pages The program that implements the acquisition process is written in C language III DYNAMIC TESTING OF HM407 IN RFR AND AVM MODES To ensure that the HM407 in RFR and AVM modes is correctly tested the sinewave generator was adjusted so that the peak to peak amplitude of the sinewave detected by the HM407 coincides or is slightly smaller than its amplitude range Different settings of the HM407 time base and vertical gain were tested Several acquisitions were made at different test frequencies In all the situations analysed in both modes the number of samples used was M 2048 and was obtained from a number of input sinewave cycles higher than 20 Also in all the situations analysed in both modes was used the 4 term maximum decay window 8 Fig 2 reports the SINAD estimates obtained in RFR mode at a given frequency Each plot refers to a different HM407 vertical gain 100 mvidi so 50 mwidis J 20mVidiv x 10mWwidiv o 5mVidiv m 45 S T i d un N LE ia NER a 4d0E x 4 q a Es z tA 3235F an M aM LP 5 10 15 20 Frequency kHz Fig 2 SINAD estimates as a function of frequency for different HM407 vertical gains From Fig 2 results that the SINAD estimates remai
92. ivision 4 5 Obviously the implementation of the square root operation supposes the extraction of the bias value just in this case from one of the two data inputs Y is partitioned in the same way and we use a ROM ROMC for memorising the conversion error E in 2048 points as well as the difference AE ROMC The final result of conversion is 2 1 y Ey AE x y2 8 The correction values Ey for both log 1 y y and 1 y 2 are represented in figure 1 0 10 log 1 y 1 y 2 0 08 0 06 0 04 0 0 0 2 0 4 0 6 0 8 1 0 Fig 1 The conversion errors between the log 1 y and y and the 1 y and 2 Implementing the equations 1 2 6 and 8 led to a 6 stage pipeline structure 1 2 which allowed a 100 MHz clock frequency in 0 8 um CMOS technology However the structure which was obtained has the following disadvantages from the point of view of the computation speed The propagation times on the pipeline stages were not the same The speed advantage resulted from the vertical carry propagation in the multiplication area organised in a Wallace tree was diminished by the time necessary for the proper addition of the last two intermediate partial products when the carry propagation is horizontal even if this is done with a fast adder in the third stage of the structure there are two successive proper additions which together lead to a unique final result consuming a lot
93. ize equal to that of the least significant partial product having the value AEyn will be added If the control bit is 0 then the generation of partial products will be done with y2 unreversed and the bits py of the first pseudo partial product from figure 3 will all be O The Wallace tree will thus have as inputs 15 initial pseudo partial products and it will provide two data words sum and carry If in this stage we did the proper addition of these we would obtain the value of the logarithm of the mantissa of each operand applied to the input of the two logarithm computation circuits working in parallel Further on the two fractional numbers obtained would be concatenated to the exponents of the two operands in order to obtain the logarithms of the operands Finally the two logarithms would be applied to the ALU in order to be added or subtracted However in this case too we would have two consecutive proper additions which slows down the process of obtaining the result For avoiding this situation we can also consider the two pairs of data words sum and carry as pseudo partial products and thus they are again introduced in a new reduction block This will provide in the end two data words the final sum and carry which will be possible to be added then with the help of a fast adder This final reduction block of the last pseudo partial products will be included in the ALU as it should act under a
94. ized and for that you must use MCI functions 33 QoS Quality services this module describes the needs of a multimedia application from the point of view of components functionality 4 I think that the model of a multimedia application has to contain evaluation elements of the hardware resource performances QoS module has to realize an initial testing of the equipments that is going to run Is necessary to inform the user about how the application will run in these conditions eventually it may propose solutions The biggest problems in multimedia applications are the synchronization constraints of different types of media witch runs simultaneous I propose that the next parameters witch describe the quality of a service of the way the program runs in a hardware environment to be verified and respected by the programmer 5 The rate of application run speed The level of hardware utilization The level of jitter The temporal cumulate errors The rate of environment errors Browse functions because multimedia applications are becoming more and more complexes using more than 100 000 objects is necessary to introduce a new module witch contains browsing functions It is very difficult to go thru the entire application to find information that s why we prefer automatically solutions for searching For this kind of applications is necessary to have an objects search functions library This ro
95. kHz 15 SkHz all at 40dB 4 9kHz 5 8kHz 8 4kHz 12 6kHz 13 5kHz 17 7kHz all at 60dB In addition to differences between adjacent partials of approximately one critical band the partials are grouped into sub bands Waveform contains partials 55 located in the lower sub band of frequencies 0 5 5kHz waveform 2 has the same partials at increased amplitudes and waveform 3 introduces frequency components in the middle and higher sub bands 5 5kHz 5 Synopsis of results The spectral components based on the lower sub band frequencies are recreated with no apparent increase in quantisation noise when at an amplitude of 60dB below the masking tone fig 16 An increase in the amplitude of the masked components of 20dB results in a measurable amount of spectral quantisation noise spread throughout the entire sub band fig 18 This illustrates that the bit allocation algorithm has distributed bits according to the magnitude of the spectral components in the same sub band The introduction of frequency components in the higher sub band of frequencies shows a significant increase in spectral quantisation noise up to 20dB fig 20 Additionally the highest frequency component has not been reproduced at all further illustrating the importance of lower frequency partials during bit allocation process S TERET F n LOCH rji 3 LION RUIT Fig 18 Waveform 2 b 16 bit 44 IKHz ATRAC encoded signal
96. leura E est suffisant de choisir la valeur de seuil 16 La constante a peut tre mise en relation avec le rapport signal bruit des signaux u et s qui ont la m me nergie sn Donc on peut crire E Snro 10 logig 7 2 10 logig a 17 Ainsi une borne inf rieure du rapport signal bruit qui d pende de a B 10 log a 18 a t tablie En prenant le signe gal dans la derni re relation on peut obtenir une borne inf rieure pour la constante a SE a 10 19 En utilisant cette valeur et la relation 16 une borne inf rieure de la valeur du seuil t peut tre obtenue S 20 En conclusion en choisissant pour le seuil t une valeur superieure on obtient une valeur du rapport signal bruit la sortie superieure f Malheureusement la valeur exacte du snr ne sera pas connue C est le motif pour lequel un algorithme adaptatif pour le choix de la valeur du seuil est recommand Cet algorithme peut utiliser la valeur pour initialisation En commencent avec cette valeur t est augment A chaque it ration la valeur du snr est calcul e Si cette valeur est superieure le processus d augmentation est continu L algorithme prends 44 fin quand pour la premi re fois la valeur de snr devient inf rieure V LE QUANTIFICATEUR Les consid rations faites dans le paragraphe precedent sont bas es sur l hypoth se d une quantification uniforme Quand on utilise
97. lysis of a Hybrid Number System Processor IEEE Transactions on Circuits and Systems Vol 46 No 6 ICSPE5 June 1993 pp 382 392 4 Jurca L A Some Considerations Regarding the Design of a Hybrid Logarithmic Floating Point Mathematical Processor Buletinul UPT Tom 45 59 Fasc 1 2000 pp 169 172 5 Jurca L A Logarithmic Unit for Fast Performing of Multiplication and Division Operations in Single Precision Format Buletinul UPT Tom 45 59 Fasc 2 2000 pp 29 32 6 Fujii H amp all A Floating Point Cell Library and a 100 MFLOPS Image Signal Processor IEEE Journal of Solide State Circuits Vol 27 No 7 July 1992 pp 1080 1088 7 Mori J amp all A 10 ns 54x54 b Parallel Structured Full Array Multiplier with 0 5 um CMOS Tehnology IEEE Journal of Solid State Circuits Vol 26 No 4 April 1991 pp 600 605 10 Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Using 4 2 L Compressor in Partial Product Reduction Tree for Fast Binary Multipliers Lucian A Jurca Beniamin Dr goi Rezumat n lucrare este prezentat un nou tip de compresor 4 2 denumit compresor 4 2 L ce permite o cre tere cu aproximativ 10 a vitezei de calcul a unui multiplicator binar la care produsele par iale sunt organizate n arbore Wallace Sinteza compresorului 4 2 L a avut la
98. m the last 4 2 compressor block from figure 4 will definitely need to be included in the ALU It will perform the reduction of the last four pseudo partial products after the control logic of the adder subtracter circuit has intervened on the subtractor The subtractor is represented by two pseudo partial products whose addition is no longer performed exactly with the purpose of increasing the propagation speed of the carry through the logarithmic unit This means that both terms must be converted in two s complement code The avoidance of the reconversion from two s complement in sign magnitude code of the result in the case in which it is negative is done using the same method as in 6 only modified for 4 operands We note Aj Ao respectively B4 B the 4 pseudo partial products and A A4 represents the subtractend while B B4 B represents the subtractor in the case in which a subtraction is performed Considering the convention of representing B negated the data word obtained through the inversion of all the bits of B as being IB we can write the two terms which are simultaneously computed in the adder subtracter circuit A B A B4 B2 A 7 B 7 B2 2 11 KB A A B B2 1 12 Equation 12 can be checked as follows in equation 13 below A B B A B A 1 13 we replace the term B A with the value obtained from equation 12 we indeed obtain A B A Ba B 1
99. mples to average for each data Ho point 100 enor out latest temperature dal gt error cluster so PE error out 3 cold junction voltage from frame 1 Init Er Cluster a m TC type HI Ee v H Pat Print Vi To HTML b HTML File Path Append Format mage Format Image Depth Image Directory Fig 3 3 Block Diagram Sequence 3 for SCXI 1100 To reduce the noise on the slowly varying signals produced by thermocouples you can average the data and then linearize it For greater accuracy you can measure the amplifier offset which helps scale the data and lets you eliminate the offset error from your measurement After measuring the amplifier offset measure the temperature sensor for CJC Both the amplifier offset and cold junction measurements should be taken before any thermocouple measurements are taken Use the Acquire and Average VI to measure temperature sensors The main differences between the amplifier offset measurement and temperature sensor measurement are the channel string and the input limits If you set the temperature sensor in mtemp mode the most common mode you access the temperature by using mtemp If you set the temperature sensor in dtemp mode you read the corresponding DAQ device onboard channel Make sure you use the
100. n Radio communications sector recommended bitrates throughout the network transmission chain The motivation for this paper was derived from reports criticizing the sound quality of audio codecs within tandem configurations While systems incorporating low bit rate audio coding will exhibit proper performance when tested alone the sound quality may significantly decrease along with multiple generations coding The fact that a perceptual codec cannot be assessed by traditional audio measurements e g SNR THD N pinpcints another problem The performance of codecs within tandem configurations must be evaluated by subjective listening tests Degradation of the sound quality caused by cascaded audio channels is well known from analog recording Each step of copying will introduce additional tape noise linear and non linear distorsions Up to a certain level where no bit errors occur digital audio is free of such analog domain artefacts By introducing perceptual coding systems however due to quantization processes a different situation arises Fig 1 shows the basic signal path through a cascade of n audio coding devices Every codec employs a coding algorithm referred to as A B at a specific bit rate bl b2 Because multiple generations audio coding can employ one or more types of codecs operating at identical or different bit rates there are defined the following classification for tandem conditions 2 Class
101. n de la parole Dans 9 est prouv e la convergence asymptotique de cette transformation vers la transform e de Karhunen Loeve la meilleure transform e pour la compression d un signal Un nouvel algorithme de compression a t pr sent Cet algorithme est caracteris par l utilisation d un systeme de detection de seuil adaptatif Celui ci realise le choix du seuil en gardant l errure quadratique moyenne de reconstruction sous une valeur impos e L algorithme de selection du seuil peut etre initialis en utilisant la valeur de la relation 20 La methode de compression decrite se base aussi sur l utilisation d un systeme de quantification adaptatif A la m thode de compression pr sent e il faut ajouter une proc dure de codage des valeurs et positions des chantillons obtenues la sortie du quantificateur Cette proc dure n est pas encore crite mais la m thode de compression propos e semble tre tr s utile parce que sans cette proc dure de codage la nouvelle m thode est d j superieure la m thode de compression utilis e dans le standard GSM IX REFERENCES 1 N Moreau Techniques de compression des signaux Masson 1995 2 M V Wickerhauser Adapted Wavelet Analysis from Theory to Software A K Peters Wesley 1994 3 Y Meyer Ondelettes et algorithmes concurrents Herman Paris 1993 4 H S Malvar Lapped Transforms for Efficient Transform Subband Coding IEEE Trans on ASSP vol 38 pp
102. nd Cout 0 or Cary 0 and Cou 1 In other words all these possible combinations which produce either Cary 1 or Cor 1 have to be found in two disjuncted areas This possibility insures the redundant character of the outputs of the circuit Exploiting of this property we will be able to enlarge the family of 4 2 compressors f All possible combinations of input bits for which their sum is 4 or 5 have to lead to Car Cout 1 Taking into consideration all these functions considered as compulsory but also the disadvantages of the 4 2 compressor of reference it resulted the efficient structure of the 4 2 compressor from figure 3 called the 42 L compressor Due to the big number of inputs for the generation of the Sym bit according to equation 1 we adopted an arborescent structure of OR exclusive gates which groups up the inputs x1 x2 x3 x4 and afterwards Cin Then based on the grouping of the first 4 inputs we implemented Cout through the simple logical function Cout X1 X2 x3 x4 3 We thus obtain the generation of the fast output Cou by the crossing of only two logical levels obtained only with intrinsic gates with only two inputs applying the De Morgan equations None of the input lines is loaded with more than two logical inputs anymore We can notice that equation 3 fulfils the demands d and f In order to obtain the output Cary we implemented equation 4 Cary x1 E x2 x3 O x4 Ci x10
103. nerators and two impedances Fig 1 shows another version of AC bridge namely with virtual arms 1 2 It contains in two arms two sinewave generators U and U and two impedances Z and Z in another two arms If the bridge is balanced the error voltage E 0 0 and the next expression can be write U X e SS ty 1 U Sg S Mischie A Ignea are with the Electronics and Telecommunications Faculty Dept of Measurement and Optical Electronics Bd V P rvan 1900 Timi oara e mail mischie meo ee utt ro a_ignea meo ee utt ro 23 where all variables are complexe quantities In process of balancing of the bridge if U is fixed reference voltage U must be modified so E 0 and the equation 1 is satisfied If the two sinewave generators are implemented using two digital to analog converters and E will be measure with an analog to digital converter the balancing process of AC bridge can be made using a computer which controls the three converters Thus the balancing process can be made automatically in a very short time and with a minimum of accuracy impedances For simplicity Z is chosen to be a pure resistor Z R reference resistor The momentane values of two sinewaves can be write as follows u t Asin at u t Bsin ot 6 The parameters B and of the U generator must be modified in order to balance of the bridge Q From 1 and 2 Z x can be write as follows z 489909 _ x B
104. nfluence of IC structure organisation Difficulty of this problem is determinate by using different technologies high integrated degree functional complexity and by high standard of necessary levels to assure IC stability under ionising irradiation action 1 To assure electronic devices fiability that works at various type of irradiation like R ntgen y quanta electrons protons neutrins and other natural and artificial type source is an actual problem because these types of irradiation leads to defects appearence in semiconductor devices that may by the cause of IC fiability reduction 1 2 Variety of effects that appear in IC at ionising irradiation are determinate by many factors that exist in irradiant environment Space is a real irradiant environment were exist galactic irradiation protons flow a particles etc with energy between 10 10 MeV solar irradiation o particles and protons flow with energy below galactic energy but with high flow density Earth irradiation due Earth magnetic field protons flow with energy low to 700MeV and electrons flow with energy low to 1MeV 1 3 Artificial irradiation source may be made using reactors and other ionising irradiation equipment 1 4 Facultatea de Electronic si Telecomunicatii Departamentul Discution Principal causes of IC based on TECMOS and bipolar transistors functional characteristics degradation under ionising irradiation action are determina
105. noise masking characteristics could result in a signal in which the quantization noise become audible Monophonic compatibility has been identified as a requirement for high quality audio coding algorithms The mixing of left and right channels of an encoded stereo signal should cause no audible impairement of the resulting signal other than the loss of the stereo effect The development of audio processing algorithms and equipment frequently requires subjective testing to evaluate the quality of the processed audio The lack of correlation of perceived quality with classic objective audio measurement such as SNR THD and IMD is now well known and the increasing importance of psychoacoustic based audio processing is rendering classic measures obsolete for many applications Analog interface design is important since the performance of A D and D A conversions can be a larger factor in analog tandem performance than actual coding artifacts depending on the bitrate used Sufficient precision must be provided to ensure that analog noise summation is not a limiting factor Adaptive Transform Acoustic Coding ATRAC used in minidisc MD is one of a number of high quality digital audio compression schemes that are based on psycho acoustic principles Other schemes of this nature include the Motion Picture Expert Group MPEG 1 Layer 3 coding scheme commonly known as MP3 The current popularity of MP3 and the most used free player WinAmp make it
106. ns relatively constant in the considered frequency range For an ideal 8 bit digitizer SINAD 49 92 dB Thus from Fig 2 results that the electrical noise within the digitizer is higher than the quantization noise especially for small vertical gains Fig 3 shows the dynamic performance obtained in RFR mode for a vertical gain of 50 mV div and with a HM407 sampling rate f 400 kHz The sinewave test signal was characterised by a frequency fin 9 5 kHz and an offset He 0 V Fig 3 was carried out using the program package presented in 4 28 Modulo Time Plot Amplitude ZFS Frequency kHz Phase deg 0 04 0 0 EN Offset mV ution Function SINAD dB ENOB THD dBc Fig 3 The dynamic performance obtained in the situation fs 400 kHz and f 9 5 kHz In top graph of Fig 3 are presented the Modulo Time Plots 9 of the output signal and of the residual error 4 The amplitude of the output signal cycle plotted has no meaning This cycle is displayed to show the phase of each residual with respect to the output signal By means of these plots is characterised the acquisition process 9 In bottom graph of Fig 3 is represented the power spectral distribution function PSDF of the residual error This plot put very clear in evidence the distortions which affects the output signal 10 In this graphical page are presented the dynamic parameters of the digitizer SINAD ENOB and total harmonic distort
107. nsive training period IV EXPERIMENTAL DETAILS Assessment of high quality reduced bit rate perceptual audio coding schemes is not always accurate using traditional methods of electro acoustic testing e g applying sweeping sine waves to measure signal to noise ratio The nature of perceptual coding using ATRAC as an example implies that if the input signal to the encoder is simple in spectral content more bits will be available for it during the allocation process Hence the amount of audio information that is converted to quantisation noise or removed is reduced Additionally coding systems that use an adaptive MDCT transform block size exhibit greater frequency resolution with slow transient response signals as a result of a longer transform block Therefore there is a need to create test signals that can be used to assess the performance of the coding scheme over a variety of programme types These might include basic complex waveforms with harmonic content derived of psycho acoustics waveforms that are chosen to illustrate deficiencies specific to the type of coding scheme being evaluated 4 1 Experimental Aims The objectives of the experiments executed in order objectively to evaluate the ATRAC coding system as implemented on Sharp MD722 portable recorder were as follows to demonstrate the increased quantisation noise of spectral components as a result of ATRAC encoding using 16bit 44 1kHz source signals to
108. of 4 2 compressor does not reflect exactly the features of this circuit because the sum of the 4 input bits of equal weight can not be represented in all possible cases with the help of the two output bits The circuit also provides a carry out on a supplementary output line and thus needs an input for the introduction of the carry in In this way there is a 4 2 vertical propagation path and a 1 1 horizontal propagation path However the horizontal propagation of the carry is practically limited to 1 bit because the carry out is generated in such a way that it does not depend on the carry in Thus on the scale of the whole multiplier the use of this circuit permits the shift of the critical propagation 2 Facultatea de Electronic i Telecomunicatii Departamentul Electronic Aplicat Bd V P rvan Timisoara 1900 e mail jlucian ee utt ro e mail bdragoi yahoo com 11 path from the centre of the multiplication area where the degree of partial product overlap is the maximum towards the left end where this degree of overlap becomes less and less Furthermore the combination of CSA blocks of 3 2 full adders and 4 2 compressor blocks allows an efficient design of the Wallace tree by the elimination of the cases in which one or two initial or intermediate partial products remained ungrouped cross a level of the tree without being operated 5 The number of levels of the Wallace tree thus becomes smaller and the cr
109. of pseudo partial products The concatenation of the exponents will be done at the terms A and B4 obtaining the final pseudo partial products A and B4 while in the 9 bit positions of integer part corresponding to A and B of fractional weight it will be completed with zeros Thus we will obtain the other two final pseudo partial products A and B The block diagram of the new adder subtracter designed with the purpose of the complete implementation of the masked logarithm method is presented in figure 5 The bit line Sop has the logical value 0 in the case of addition and 1 in the case of subtraction The three blocks Invert will be transparent when an addition is performed and they will invert the bits of the data from the inputs when a subtraction is performed The block Selector will select the result from the Adder1 unchanged in the case of addition and in the case of subtraction it will select the result from Adder2 or from Adder1 inverted depending on the MSB of Adder2 5 The Adder1 and Adder2 are 36 bit adders The four least significant bits from the outputs of the two adders will be lost and thus at the output of the circuit we will regain the single precision format plus one bit in the MSB position which avoids the overflow due to the accumulation of bias values in the case of multiplication The justification of the 36 bit length for the adders and the compressor block will be done in the next paragraph
110. oisi me paragraphe est la TPC Le paragraphe 4 est d di la pr sentation d une m thode de quantification tr s simple On peut trouver quelques r sultats de simulation dans le paragraphe suivant Quelques conclusions sont pr sent es dans le dernier paragraphe II LE SYSTEME DE COMPRESSION Le syst me de compression propos est pr sent la figure 1 x n yin z n n Figure 1 Le syst me de compression propos Le premier bloc CPT calcule la transform e en paquets de cosinus du signal d entr e x n ve Le d tecteur de seuil TD r alise l op ration zn m gt Si h gt t 1 0 si non o est la valeur d un seuil Q repr sente un quantificateur adaptatif Le codeur C r alise une compression sans pertes suppl mentaire La sortie du syst me de compression est d crite par le signal v n Le d codeur D est le syst me inverse du codeur C A la sortie du syst me qui calcule la transform e en paquets de cosinus inverse ICPT on obtient le signal reconstruit n La qualit de la reconstruction peut tre appr ci e l aide du rapport signal bruit de ce signal snr L erreur de reconstruction repr sente la distorsion introduite par la m thode de compression et peut tre regard e comme une perturbation additive du signal d entr e Sa puissance repr sente la puissance du bruit dans l expression du rapport signal bruit L autre param tre important du syst me de compr
111. ondantes y et b sont nulles aussi Donc le nombre total de bits allou s aux chantillons du signal w n N est tr s petit par rapport au nombre de bits du signal x n Cette proc dure de quantification a un petit d savantage aussi Pour la transmission ou pour le stockage de chaque bloc 22 du signal uln w n il faut ajouter aux coordonn es de chaque chantillon quelques valeurs suppl mentaires les valeurs z y En utilisant ces valeurs et la relation 21 les nombres yy peuvent tre calcul s aussi pendant la phase de reconstruction A l aide des param tres z et y les op rations inverses aux op rations d crites par la relation 22 peuvent tre r alis es Mais le nombre de bits demand s pour la repr sentation des valeurs z est tr s petit par rapport N Donc le facteur de compression n est pas affect par la n cessit d ajouter les valeurs suppl mentaires pour chaque bloc wull La valeur du facteur de compression r alis par le syst me de la figure 1 peut tre calcul en utilisant la relation 16 N e 23 fe N N B 23 o N repr sente le nombre de bits demand s pour le codage des positions des chantillons du signal un et B represente le nombre de bits demand s pour le codage des param tres Les nombres AN et B peuvent tre calcul s en utilisant les relations suivantes 32 Ne Y N K bj k 1 24 o N k repr sente le nombre d chantillons non nuls dans
112. or in single precision format The removal of all bits from this area i e the elimination of the hard structures from the whole Wallace tree and ALU which correspond to this area does not introduce a supplementary error if the values E which are memorised in ROMA are rounded to 23 bits toward the nearest higher value and the values E whose two s complement is memorised in ROMC are rounded through truncation V CONCLUSIONS The use of the masked logarithm method in the implementation of the logarithmic unit allows a very efficient organisation in a 6 stage pipeline structure as shown in figure 6 The total time for the obtaining of the result we have taken into consideration the maximum propagation times of all the blocks from the critical path is tiot troma hoa tee taru tRowc talog tree trinsum 4ns 4ns 4ns 4ns 4ns 4ns 24 ns Input operand A Sa Ea ya Input operand B Ss Es ys M I Y18 T 4ns EA YA ROMA ROMA ROMB ROMB ys E Eya AEya AEys Eys 4ns Wallace Tree Branch A Wallace Tree Branch B SI e J A S 22 OP B B Sa 5 ALU Se 4ns B o I ET 4ns Wallace Tree antilog Output C Fig 6 The architecture of the logarithmic unit We can see that in this case unlike in the case of the most performant implementation described in the field literature by F Lai and E Wu 1 2 3 the propagation times through the blocks from the c
113. ore digitizing it The manipulation of signals to prepare them for digitizing is called signal conditioning Common types of signal conditioning include the following amplification linearization transducer excitation isolation and filtering II SCXI OPERATING MODES The SCXI operating mode determines the way that DAQ devices access signals There are two basic operating modes for SCXI modules multiplexed and parallel When an analog input module operates in multiplexed mode all of its input channels are multiplexed to one module output When you cable a DAQ device to a multiplexed analog input module the DAQ device has access to multiplexed output of that module as well as all other modules in the chassis Facultatea de Electronic si Telecomunicatii Departamentul MEO Bd V P rvan Timi oara 1900 e mail mihaela meo ee utt ro 19 through the SCXIbus The analog input VIs route the multiplexed analog signals on the SCXIbus for you transparently So if you operate all modules in the chassis in multiplexed mode you need to cable only one of the modules directly to the DAQ device When an analog input module operates in parallel mode the module sends each of its channels directly to a separate analog input channel of the DAQ device cabled to the module You cannot multiplex parallel outputs of a module on the SCXIbus You must cable a DAQ device directly to a module in parallel mode to access its input channels In this conf
114. osition suivante Proposition 1 Une borne superieure de la distorsion du signal reconstruit obtenu apr s la compression adaptative bas e sur la TPC d crite dans ce travail est N t 0u N repr sente le nombre d chantillons du signal d entr e Preuve L erreur quadratique moyenne d approximation du signal yn par le signal z n est K Y k l 11 o n repr sente la position des chantillons du signal y n avec la valeur absolue inf rieure au seuil t Il y a K tels chantillons Soit ofn le signal obtenu apr s la mise en ordre croissante des chantillons du signal ul L erreur quadratique moyenne devient K So k lt K t k l 12 On consid re comme premi re approximation que le bloc Q r alise une quantification uniforme de pas t En utilisant une quantification non uniforme les r sultats seront mieux L erreur quadratique moyenne est N ez D eln uln 13 n l Pour chaque chantillon de la s quence o k k 1 K on associe un z ro dans la s quence fn Pour les autres chantillons du signal z n la diff rence z n v n est inf rieure la valeur t C est le motif pour lequel on peut crire N SE N K t k K 1 lt E2 14 Parce que la distorsion d finie en 9 peut tre crite dans la forme Dee 15 tenant compte des relations 12 et 14 on peut conclure que la proposition est prouv e Donc pour garder la distorsion sous la va
115. r bloc de la figure 1 calcule la TPCI Le r sultat est le signal n C est le resultat de la procedure de reconstruction En utilisant ce signal on peut calculer la distorsion D Toutes les op rations d j d crites sont r p t es pour diff rentes valeurs du seuil t pour la maximisation du avec la contrainte que snr soit sup rieur 5 VII RESULTATS DE SIMULATION Le syst me d crit plus haut a t simul en Matlab Un morceau de parole a t segment en blocs de 1024 chantillons Pour chaque segment la proc dure pr sent e a t appliqu e La TPC a t calcul e en utilisant une ou deux it rations Il y a deux types de segments les segments tonals et segments de bruit o la forme d onde ressemble celle d un bruit La valeur moyenne pour le facteur de compression de ces 20 segments est 11 54 C est le motif pour lequel on peut dire que la m thode de compression propos e ici est superieure la m thode utilis e en GSM Le plus petit rapport signal bruit enregistr sur ces 20 segments est de 18 49 dB Le facteur de compression peut tre augment sans perdre la qualit si chaque TPC d un segment tonal est calcul e en utilisant une seule it ration Une autre possibilit d augmenter le facteur de compression est d utiliser des segments plus longs par exemple contenant 2048 chantillons VIII CONCLUSIONS Dans ce travail on a t prouv e l efficience de la TPC pour la compressio
116. rectly the system of events witch might appear and also the actions witches are realized after this It is necessary that we define one hierarchy of priorities in execution incase of mixing over the events in the same time in the case in witch one event might provoke more than one action C Input Output Operations Input Output Operations contains also one component witch permits the network work the development of services client server and a system of user validation Input Output Operations represent the way of interaction between the application and the out world especially with the hardware resources of the computer system on witch the application runs We can found these functions in all the programming languages I propose the writing of small routines witch can verify the peripheral state before their utilization the release of some occupied resources because of some damage and the determination of peripheral performances In the relations with the functions from QoS library checking services for the quality of the provided service the check for a good run of a multimedia application on a certain system it s imperious the introduction of a routine for error detection that appears in the functionality Together with QoS functions there will be generated messages witch will inform the user on the eventual problems that might appear and proposes the remove of problems Network because now is a big
117. red colour image segmentation alghorithm Transactions on Automatic Control and Computer Science Fourth International Conference on Technical Informatics CONTI 2000 12 13 October Timi oara Romania pp 139 144 2000 Khoshafian S Baker B Multimedia and Imaging Databases Ed Morgan Kaufmann Publisher San Francisco California 1996 Ghinea G Thomas J P QoS Impact on User Perception and Understanding of Multimedia Clips Proceeding of The 6 th ACM International Multimedia Conference Bristol England 12 16 September pp 49 54 1998 5 Mocofan M Pescaru D Aplica ii interactive multimedia Programarea OpenScript Ed Politehnica Timisoara 2001 2 3 4 Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 Optimised Homotropic Structuring Element for Handwriting Characters Skeleton loan Snep Dan L Lacr m Corneliu Toma Abstract This paper is focused on the study of skeleton transformations involved in the hand printed amp handwritten documents images preprocessing A new homotropic transformation structural element is proposed in order to optimize the sequential thinning procedure reduced number of thinning cycles and less coarse skeleton lines Keywords Hand written Character Recognition Java Programming
118. reo imaging noise modulation a harshness to the the sound aliasing and increased background quantising noise In general it is wise to avoid recording a complex mix on a data reduced if any subsequent processing or copying is envisaged It would also be sensible to avoid recording stereo source material on data reduced systems wherever possible but single mono Sources will not be noticeably affected by data reduction VI REFERENCES 1 RJ Beaton Network Iriplications of Low Bitrate Audio Codecs Bericht 18 Tonmeistertagung Karlsruhe 1994 pp 553 12 M Kevhl J Herre C Schmidmer NMR Mesurements on Multiple Generations Audio Coding presented at the 96 AES Convention 1994 February 26 March 01 Amsterdam preprint 3803 774 6 13 AJ Mason AK McParland N J Yeadon The Atlantic Audio Demostration Equipment J Audio Eng Soc Vol 48 No 1 2 2000 Januarv February pp 106 118 4 K Brandenburg MP3 and AAC Explained presented at AES 17th International Conference on High Quality Audio Coding 5 D Spector Evaluating Adaptive Transform Acoustic Coding for Mini Disc presented at the 108th AES Convention 2000 February 19 22 Paris Preprint nr 5082 B 4 16 H Robjohns Cutting it Fine in Sound on Sound august 1998 7 M Erne Perceptual Audio Coders What To Listen For presented at the 101 AES Convention 2001 September 21 24 New York USA 8 M Erne Digital Audio
119. ritical path are considerably smaller and are much more balanced The clock frequency of the system thus reaches 250 MHz which represents a speed increase of over 60 in comparison with the above mentioned reference Furthermore we remarkable features can notice some other the number of logical gates from the entire structure is significantly smaller due to the elimination of 8 fast adders of minimum 18 bits with the price of the supplementation of three fast adders with 4 bits the propagation time through the critical stage of the pipeline decreased up to the minimum physical level given by the access time of the memories the latch size between the pipeline stages decreased with roughly 15 due to the arborescent structure of the logarithmic unit a supplementary saving of roughly 25 was achieved at the implementation of the three multiplication operations of the type AEyxy2 a possibility which has been demonstrated through an error calculation in the case of the above mentioned product BIBLIOGRAPHY 1 Lai F A 10 ns Hybrid Number System Data Execution Unit for Digital Signal Processing Systems IEEE Journal of Solid State Circuits Vol 26 No 4 Apr 1991 pp 590 599 2 Lai F Wu C F E A Hybrid Number System Processor with Geometric and Complex Arithmetic Capabilities IEEE Transactions on Computers Vol 40 No 8 Aug 1991 pp 952 961 3 Lai F The Efficient Implementation and Ana
120. s has an immeasurable amount of pre echo as seen in fig 22 The MP3 encoded signals at 128kbit s have approximate 150ms of pre echo signal fig 23 The superior performance of ATRAC in this test reflects the effectivness of the adaptive transform block size Fig 21 Castanets 16 bit 44 IKHz PCM signal EE A M EH SU GS jo anm Fag 22 Castanets ATRAU encoded s gnal equi 292kbits s i FS ER Re ee ae a Fig 23 Castanets Audio Catalyst MP3 Lincoder ut 28kbits s 4 3 Multiple Generation ATRAC encoding One of the major issues concerning the MD recording system is its performance over multiple generation of copying to and from another MD recording systera This is analogous to tape to tape cassette recordin To execute the multiple generation tests a single signal was recorded io MD from the computer and the MD recording transferred back to the computer This was the first generat on recording which then become the source of the next MD recording with the pattern repeating up to ten generations The resultant signals for all ten generstions were analysed in both frequency and time domains The creation of these sound files was made possi Je by the fact that the PC digital audio interface which was used for repeated transfer of the test wavefortas ignores the copyright flag included in the S PLAF consumer channel data structure Test 4 Objective To demonstrate in both th
121. s in this way the users succeed to personalize their applications Also is good that for experimented developers of multimedia applications to exist instructions on the way to implement new functions witch later can be distributed in new ways of multimedia applications processing parameters The functions in this library Area use generally only one object during the processing part Users User Interactions Time_Line Objects processing functions Temporal processing Processing_Functions Kernel Transitions Input Output Operations Client Server Integrated technologies Objects processing Objects Fig 1 Multimedia application model architecture 32 B User interaction This module is characterized through a graphical interfaces and events generated from interactions between application and user GUI the graphical interface with the user is made from different graphical elements objects witch offers interactivity between user and application buttons links images little animations etc These objects have attached routine witch start when one defined event appears in a library of functions methods and events vents Events The events system modify the way of application run depends on the actions made by the user in the application This makes an event and for them we will have a specified action It is very important that we define cor
122. se analogue multidimensional processing array 1 Its dynamics is described by the following equations XO O DA Yat 2 Ba Ua Hi KHN KIEN 1 1 y t 2 2 where xjis the state of the C cell y is the output of the Cj cell ux is the independent input in the Cu cell Aju is the weight of the feedback connection from the C4 cell to the Cj cell Bix is the weight of the control connection from the Cy cell to the Ci cell Ij is the bias current of the Cj cell N is the neighborhood of C cell The CNN most important characteristics are their geometrical and electrical regular structure the locality of interconnections between the processing elements and their programmability These characteristics permit to a CNN to solve complex image processing tasks through the implementation of analogic algorithms that can be combined according to a flowchart Some of the applications are interpolation and three dimensional image reconstruction 3 image compression and decompression 6 microscopic and topographic image processing 2 4 character recognition Canadian bank notes recognition 5 xy 0 1 ho 1 Facultatea de Electronica si Telecomunicatii Departamentul In this paper it is presented a method to stop the operation of a color copy machine whenever is detected a fraud tentative using different analogic algorithms and local logic of CNN cell Similar algorithms are presented in
123. sum and carry being obtained after only 3 levels of compressors They are then added with a fast adder in order to obtain the mantissa of the final result from the output of the logarithmic unit In this case too we will take measures to avoid the subtraction of the term AEyxy2 but we also take into consideration the fact that in equation 8 the term Ey must be subtracted too The product Avon is subtracted in the cases which correspond to the positive slope on the diagram of O the function 1 y 2 presented in figure 1 while it is added in the other cases As equation 9 can no longer be used the sum of the two negative terms from equation 8 will be written as follows Ey n AEY nyxy2 EyintA Y nXY2 Ey ny AEyq AEyqyx yz 1 EynytAEyn AEy n xy2 AEyqy Ey n 1 tAEY nyxY2t AEY n 10 Thus in the ROMC locations from the address 0 to the address 1082 the two s complement of the quantities EYm 1 which should have been found at the next address as well as the value 1 for the control bit are memorised from address 1083 to address 2047 when AEyxys is positive the two s complement of the values Ev as well as the value O for the control bit will be memorised directly III ALU DESIGN Due to the fact that we must implement in the ALU besides addition the subtraction too which implies the conversion into a two s complement of the subtractor the number which is subtracted from the other ter
124. supposes the following steps step 1 Is determined the best sine fit corresponding to the digitizer output data signal by means of interpolated fast Fourier transform Interpolated FFT algorithm 5 m Asin 27m I ep td s A 1 m 0 M 1 where fins d are the amplitude frequency phase and offset of the best sine fit estimated by Interpolated FFT algorithm step 2 Is computed the residual error err m between the digitizer output data signal y m and the best sine fit m err m y m Gel m 01 M 1 A2 step 3 SINAD and ENOB are estimated by the following relationships 1 92 SINAD es dB 20 log 19 A 3 rms err ENOBes n lo du AA m Sa ideal rms err Ka where n is the resolution of the analog to digital converter ADC of the digitizer ideal rms err gt q is the ideal code bin width of the ADC 30 2 3 4 5 6 7 8 10 REFERENCES D Belega Accurate dynamic characterization of digitizing signal analyzers Proceedings of the Sympozium on Electronics and Telecommunications ETC 2000 vol II pp 77 80 Timisoara 2000 Hameg Instruments Manual for Oscilloscope HM407 1996 Tektronics User Manual for TDS210 Digital Real Time Oscilloscope 1998 D Belega Contributions at the testing of analog to digital converters Ph D thesis Universitatea Politehnica Timisoara 2001 C Offelli D Petri Interpola
125. t so that the maximum sum of the 5 input bits which equals 5 could be represented In this way the equation 2 will be satisfied has to be generated by the X1 X2 X3 xX4 4Cin Canyt Court Sum 2 Indeed when all the input bits are 1 we obtain 5 10 10 1 INCLUDE C MsimS2 compres stm fast input fast input Fig 1 Optimized 4 2 compressor presented in 6 x1 k X2k X3k x4 x d RM A EE em SE ZK D ao Ho ES gt rst Fig 3 The 4 2 L compressor 13 Because the outputs Cin and Cary have the same order they provide a redundant representation at the output of the 4 2 compressor c One of the bit lines of output carry has to be generated by simple logical functions which to include a minimal number of logical levels and not to depend on the carry in Cin It will be called fast output and will insure the carry in for the compressor with the immediately superior order In this way the horizontal propagation of the carry will be limited to only one bit and will not influence the speed of result generation at the output of a cascaded 4 2 compressor block d All possible combinations of input bits for which their sum is O or 1 have to lead to Cary Cout 0 e All possible combinations of input bits for which their sum is 2 or 3 have to lead to Ce 1 a
126. t 1 bit full adders The CSA blocks accepted 3 partial products at the input and they provided 2 intermediate partial products in a time equal to the one necessary for the propagation through one full adder The great computation speed is obtained on the one hand due to the solely vertical propagation of the carry in the multiplication area and on the other hand due to the simultaneous calculation of more initial or intermediate partial products But using only CSA blocks leads to the situation in which 1 or 2 partial products remain unoperated on one of the levels of the Wallace tree This fact results in a certain inefficiency due to the further delay which appears with the increase of the number of levels of the Wallace tree A more efficient structure is obtained if instead of the CSA blocks we use 4 2 compressor blocks or a combination of 4 2 compressors and CSA blocks The 4 2 compressor was introduced in the field literature by Weinberger in 1981 However the first implementation on the chip which used this circuit appeared only in 1991 in the form of the 54 x 54 bit multiplier designed by J Mori 2 possible to be achieved as a result of the increase of integration density Since then other improved variants of 4 2 compressors have appeared variants which will be compared at the end of this paragraph with our own variant of this circuit named the 4 2 L compressor II THE 4 2 COMPRESSOR CIRCUIT The generalised name
127. te by charges accumulation in oxide volume and substrate and at oxide semiconductor interface TECMOS case These charges can be classified such 1 Interface trapped charges Qi localised at Si SiO interface This type of charges induced energetic state in Si forbidden band 2 Fixed oxide charges Qs localised near Si SiO interface This type of charges are imobile in electric field 3 Oxide trapped charges Q localised near Si SiO interface or near oxide metal interface Oxide trapped charges are created by exposition at ionising irradiation 4 Mobile ionic charges Qm determinate by presence of alcaline metal ionised atoms such sodium and potassium This type of charges are localised near Si SiO interface and gets there by moving in electric field Using gate dependence last three categories can be grouped together in one unique class oxide charges Q different as interface trapped charges that depend by applied tension on gate an will be call interface charges Qi The effect of positive charge accumulation in MOS structures oxide under ionising irradiation action was study by Grove A S Snow E I 6 and Stanley A G 7 In concordance with these models under ionising irradiation action in MOS structures oxide take place electron hole pair generation followed by separation of these pair under external electrical field according with passing of most mobile carriers electrons from Ele
128. temperature sensor on the SCXI 1100 use the standard SCXI string syntax in the channels array with mtemp substituted for the channel number as shown in the following example To read the grounded amplifier on the SCXI 1100 use the standard SCXI string syntax in the channels array with calgnd substituted for the channel number as shown in the following example SGXT OO Te mm Fig 1 Connector Pane for SCXI 1100 device 1 WE S 1 For instructions select File gt gt VI Properties gt gt Documentation channels ob sci md1 0 3 Temperature Strip Chart Celsius SO Robot set mat 103 m A offset channel Deh ob0 sc mdi calgnd cold junction channel ob0 sc mdi mtemp B ob sc1 mdi mtemp ae number of samples to average 10 0 for each data point 100 a100 You must stop 0 0 scan rate the acquisition to 0 50 512scans sec Change any input latest temperature data STOP 26120 parameters TC type JF J chan iio input limits in deg C 50100 deg C CIC Sensor JC Sensor 0 error out high limit en limit del source 350 000 zum ERROR 70401 JAI Group Config Fig 2 Front Panel for SCXI 1 100 E GIEL binary amplifier offset pass on error cluster Always use the same input limits and clock tates that you will be using in your acquisition when you determine the amplifier offset S
129. th calculation and the result cannot be obtained instantaneous The processing functions have a big diversity and depend of the element that is processed Sometimes its implementation can be very heavy I recommend structuring of the elements in a library that contains processing functions which can be accessed later For users can be allowed the personalization of these functions this is realizable by modification of the Transitions during one multimedia application you don t know the moment in witch the user decides to go to another scene of the application To realize a very interesting switch between scenes frequently are used transitions or effects between objects When in those scenes are objects like videos and animations those contain different positions in each moment that s why you cannot recalculate the transition between the two scenes the transition between the objects from the two scenes The processing part it will be done in the moment decided by the user when he wants the transition In this case it will be a process on line witch will use the system resources where the application is running and we will not talk about storing the results of the process For a new utilization of the processes effects and transitions I recommend the organization of a library with distinctive processing functions Transitions Is necessary for these functions to exist a transparency of processing parameter
130. tion Techniques for Real Time Multifrequency Waveforms Analysis IEEE Trans Instrum Meas vol 39 pp 106 111 Feb 1990 S J Tilden T E Linnenbrink P J Green Overview of IEEE STD 1241 Standard for Terminology and Test Methods for Analog to Digital Converters Instrum and Meas Technology Conference Venice vol 3 pp 1498 1503 May 1999 Hameg Instruments Manual for Function Generator HM8130 1996 A H Nutall Some windows with very good sidelobe behaviour EEE Trans Acoust Speech Signal Processing vol ASSP 29 pp 84 91 Feb 1981 F H Irons D M Hummels The Modulo Time Plot A Useful Data Acquisition Diagnostic Tool JEEE Trans Instrum Meas vol 45 pp 734 738 June 1996 J J Blair A Method for Characterizing Waveform Recorder Errors Using the Power Spectral Distribution IEEE Trans Instrum Meas vol 41 pp 604 610 October 1992 Buletinul tiin ific al Universit ii Politehnica din Timi oara Seria ELECTRONIC i TELECOMUNICA II TRANSACTIONS on ELECTRONICS and COMMUNICATIONS Tom 46 60 Fascicola 1 2001 New Interactive Multimedia Application Model Using Objects Storage in a Database Muguras D Mocofan Corneliu I Toma Rezumat Lucrarea prezint un model de aplicatii multimedia ce permite utilizarea celor mai moderne tehnici de programare programarea orientat pe obiecte si cea orientat pe aspecte Modelul propune un mod de structurare al apli
131. ults that the LMS algorithm is very fast and converged in a very short time and measuring process is a real time Tab 1 Measured values R Measured values Measured values Weights Magnitude with AC bridge with insruments of U LSB 3 38kQ R 76811 kQ 6 84 KQ w1 2 0152 800 w5 0 0003 3 38kQ C 0 3064 uF 0 33 uF w1 0 038 800 R 110Q w 1 537 21 6kO R 7 245 kO 7 25 KQ w 0 3354 1200 w2 0 0001 21 6kO C 0 3181 uF 0 33 uF w 0 0061 1000 R 1320 w 0 2310 21 6kQ C 0 1291 uF 0 12 uF w1 0 015 1000 R 3280 W2 0 569 3 38kQ L 0 931 H 10H w1 0 027 700 R 93 Q 105 Q w2 0 346 3 38kQ L 1912H 2 0 H w1 0 073 700 R 222 Q 215 Q w2 0 711 500 amplitude samples 1500 Fig 6 Variation of E voltage IV CONCLUSIONS This paper presented implementation of an virtual AC bridge for impedance measurement using an IBM PC which contains a data acquisition board with two DAC s and one ADC This virtual instrument works in real time using a C language program Experimental tests were carried out on resistors capacitive impedances and inductive impedances The obtained results have very small errors 26 REFERENCES 1 M Dutta A Rakshid S N Battacharyya and J K Choudhury An application of the LMS adaptive algorithm for a digital AC bridge JEEE Trans Instrum Meas vol 36 pp 894 897 1987 2 Selim S Awad Natarajan Narasimhamurthu and Wiliam H Ward
132. unfavourable case in the case of the 4 2 compressor presented in 6 15 e de D yl 130 Ais DIS pa 1 13 SINE POST P Ap e e xl Om d D d 05m SInNEWEUST ci Time ns e D n dx 05 yl 009 y2 009 os pm SMNEWFOST E r Y 8 E i E i Time ns Fig 7 The logic state of the outputs Sum Cary and Cu in the most unfavourable case in the case of the BIBLIOGRAPHY 1 Stallings W Computer Architecture and Organization Prentice Hall Inc 1996 2 Mori J amp all A 10 ns 54x54 b Parallel Structured Full Array Multiplier with 0 5 um CMOS Tehnology IEEE Journal of Solid State Circuits Vol 26 No 4 April 1991 pp 600 605 3 Villeger D Fast Parallel Multipliers Final Report Ecole Superieure d Ingenieurs en Electrotechnique et Electronique Noisy le Grand May 1993 4 Ohkubo N amp all A 4 4 ns CMOS 54x54 b Multiplier Using Pass Transistor Multiplexer IEEE Journal of Solid State Circuits Vol 30 No 3 March 1995 pp 251 257 5 Wang Z Jullien H A Miller W C A New Design Tehnique for Column Compression Multipliers IEEE Transactions on Computers Vol 44 No 8 August 1995 6 Oklobdzija V G Villeger D Liu S S A Method for Speed Optimized Partial Product Reduction and Generation of Fast Parallel Multipliers Using an Algorithmic Approach IEEE Transaction on Computers Vol 45 No 3 March 1996
133. utines are very complexes sometimes are necessary to implement also content based methods A correct structuring of the objects and also the implementation of the searching functions offers many advantages to the users increasing the attractiveness of the application The Browse library has to be open for new developments III MULTIMEDIA OBJECT DESCRIPTIVE MODEL Each multimedia object has a value a logical structure and an interpretation of the content for that value The value of a multimedia object his logical structure and the model of representation and storages must be described in the multimedia object s model The multimedia describing model contains information that is referred to the logical composition of this object the ways of synchronization and temporization for the components and the necessary parameters for displaying The model of interpretation has to describe exactly the real world thru the sub objects that are composing it and thru the relationship between them and between each sub object and the real world 34 One multimedia object may be composed with other multimedia objects or with simple elements un composed multimedia objects One multimedia object may be composing with the elements Texts mages Sounds Videos Graphics Animations Scripts The model of description can be synthesized like this A number of simple or complexes sub objects compose each multimedia object
134. ware for IEEE standards 1057 and 1241 I INTRODUCTION The acquisition process of an analog signal is often made by means of a digitising oscilloscope DSO The actual DSOs have in addition some signal recording modes like for example computation and displaying the average value of several signal recordings displaying the envelope curve of the acquired signal peak detector etc 2 3 For efficiently using of such DSO it is very useful to know based on the experimental results the behaviour of DSO in each signal recording mode Also it is very important to know with high precision the dynamic performance of the DSO obtained in different signal recording modes In this paper is experimentally analysed the behaviour of an actual DSO HM407 in different signal recording modes Refresh RFR Average AVM and Envelope ENV HM407 is an 8 bit DSO with 40 MHz analog bandwidth a 100 MS s maximum sampling rate and a maximum record length of 2048 samples 2 Also HM407 has only an interface the serial interface RS232C for the communication with a PC One of the most important dynamic parameter of a digitizer is the signal to noise plus distortion ratio SINAD Using the method proposed in 1 was estimated the SINAD of HM407 in the RFR and 1 E TT Faculty of Electronics and Telecommunications Dept of Measurements and Optical Electronics e mail belega ag meo ee utt ro 27 AVM modes This method leads to very
135. x2 Ci x3 x4 x1 x2 x3 x4 4 The first three terms of the logical sum insures the fulfilment of the demand e related to Cou while the fourth term insures the fulfilment of the demand f from the above list We can also notice that the demand d is fulfilled by all four terms Using the three inverters from the figure 3 permitted the complete elimination of the 14 disadvantage of supplementary capacitive load of certain logical outputs one output does not command more than two inputs and in the same time it permitted the implementation of the De Morgan equations for the exclusive use of intrinsic gates in CMOS technology IV SIMULATIONS RESULTS CONCLUSIONS The circuits presented in figure 1 and figure 3 was simulated for all possible logic combination of the inputs The logic state diagram for the outputs Sum Cary and Cour for the 4 2 compressor described in 6 is presented in figure 4 and for the 4 2 L compressor in figure 5 For each input combination the analysis time was initially established at 10 ns i e much greater than the propagation time through the compressor As we can see from figure 4 and figure 5 the output Sim is the same according to the function a in our definition for the 4 2 compressor The outputs Cary and Cour present the same logic state in both figures if the sum x1 x2 x3 x4 C is equal with 0 or 1 or 5 according to functions d and f If the above sum is equal with 2 or 3 than Camyt

pdf - Departamentul de Comunicatii

Contents

Download Pdf Manuals

Related Search

Related Contents