Flexible High-Speed Platform for Precision Personnel Location
Contents
1. 50 50 51 52 52 CONTENTS 5 7 3 3 Pulse TraimS 63 7 3 3 Verification of Counter Trigger Method 63 8 Conclusions 65 8 L E ture Work z 4 18 dee Nod se nue GR ERR 65 List of Figures 1 1 1 2 1 3 1 4 2 1 2 2 2 3 2 4 2 5 3 1 3 2 3 3 Block diagram showing communication between system compo Geometric overview of system component placement The OFDM concept o o es Original demonstration photograph and screen capture with mul tiple soluti ns s loe a e Frequency Response of Radio Shack 25kHz Tweeters dB Frequency Response of Radio Shack 27kHz Tweeters dB Frequency Response of Unidentified Radio Shack Tweeters dB Frequency Response of T R 25 18Y Ultrasonic Transducers dB Frequency Response of Vifa 40kHz Tweeters dB Block diagram of audio drivers in system Microphone Preamplifier Schematic Power Amplifier Schematic llle 19 19 20 21 26 LIST OF FIGURES 7 3 4 3 5 4 1 4 2 4 3 44 4 5 6 1 6 2 6 3 6 4 The aluminum cone causes sound to propagate along a horizontal Photographs of an acrylic resonator Left 3 4 view of resonator Right microphone is fit inside a cavity recessed into the tubing and then fit over toroidal base mount Bottom overhead view shows interlocking acrylic rings 22. Reprinted with permission of National Instruments Corporation All rights re served TRIG1 STARTSCAN CONVERT HE Ur nr rd TONER a le de h di signal at whatever speed CONVERT is internally configured Only on the next active edge of STARTSCAN will CONVERT pulses be generated again For a depiction of this sequence of events see Figure 4 5 The most important feature of these signals is that they may be e generated internally and viewed externally on an oscilloscope or e generated externally and routed to the internal control lines The former feature allows for analysis and troubleshooting as well as exportation of timing signals to control other devies the latter feature means that timing signals can come from external control sources allowing for multiple device synchroniza tion For an example of such synchronization see Section 7 3 4 2 3 Analog Output The DAQCard 6062 E provides two channels of analog output at 12 bit resolution Waveforms may be updated at rates as high as 500 ksamples sec but like the analog inputs this rate is cut in half when both channels are being used simultaneously CHAPTER 4 DATA ACQUISITION HARDWARE 37 4 2 4 Analog Output Timing Signals Of the three timing signals available for analog output only one was used during the project The UPDATE signal is very much like the CONVERT signal except it cont
3. set of C functions that can be called directly from a DLL MATLAB 6 5 R13 allows the access of functions defined in Windows standard DLLs through the generic DLL package which provides access to three new MAT LAB functions loadlibrary unloadlibrary and calllib The first two functions load an unload the functions in a DLL while calllib accepts as arguments the li brary name a function name and any variables to be passed to a function For example calling a function add which returns the summation of its two vari ables from FOO DLL would involve the following code in an M file loadlibrary foo dll foo h 3NI DAQ User Manual for PC Compatibles National Instruments October 2000 CHAPTER 5 DATA ACQUISITION SOFTWARE 41 result calllib foo add 2 3 unloadlibrary foo Using MATLAB s generic DLL package allowed us to call low level NI DAQ functions directly 5 2 4 Advantages and Disadvantages The advantage of calling the driver functions directly is the flexibility afforded by register level control of the DAQCard The NI DAQ drivers allow for e software routing of any and all DAQCard signals e external synchronization and control of CONVERT UPDATE and other signals described in Section 4 2 e greater control of software flow due to higher granularity of process control The greatest disadvantage of using the NI DAQ drivers is their complexity relative to the MATLAB Data Acqui
4. the first two were used in this project 5 1 2 Advantages and Disadvantages The major advantage to using the Data Acquisition Toolbox is that it is already integrated with MATLAB Also the first iteration of the demonstration as de scribed in Section 6 1 used the Data Acquisition Toolbox for audio I O using the Data Acquisition Toolbox Quick Reference Guide 1994 2004 The MathWorks Inc CHAPTER 5 DATA ACQUISITION SOFTWARE 40 PC sound card therefore the toolbox made the code transition between Milestone 1 and Milestone 2 almost seamless The key disadvantage to the use of the Data Acquisition Toolbox is its lack of flexibility The toolbox does not allow for external timing signal control nor does it allow for software controlled routing of internal signals which became necessary features during our RF functionality tests as explained in Section 7 5 2 NI DAQ Driver Set v6 9 3 NI DAQ v6 9 3 is a set of functions that control all of the National Instruments plug in DAQ devices for analog I O digital I O timing I O SCXI signal condi tioning and RTSI multiboard synchronization Every command issued to the DAQCard must be passed to the board through these drivers Although the Data Acquisition Toolbox as available in MATLAB does provide its own front end any commands issued through the toolbox are passed to NI DAQ functions which con trol the device directly National Instruments provides the NI DAQ functions as a
5. F Carrier Magnitude Carrier te Noise Mag 10 Loops Logfile Name Path 105 O 5 10 15 20 25 30 Start Locator cMffi testdata html distance in 6 5 Results We performed a series of test runs of the Milestone 3 demonstration collecting statistical data using the HTML statistics logger The tests were run with the fol lowing parameters e 8192 sample length OFDM signal e 101 carriers e carrier range of 2 16 kHz to 7 54 kHz The data in Table 6 1 are derived from 15 test runs and 1000 cycles of the OFDM signal At a considerably poor average SNR of about 12 dB the demonstrator was able to obtain an average positional error standard deviation of about 08 inches or 2 cm which is an order of magnitude better than the desired 2 cm of accuracy A validation of our method can be found in the TOA accuracy which is on the order of microseconds To achieve that level of TOA accuracy using traditional impulse CHAPTER 6 AUDIO DEMONSTRATION 55 Table 6 1 Statistical data from 15 runs of the Milestone 3 demonstration at 1000 cycles per run SNR dB Measured TDOA Predicted TDOA Measured Variance sec Variance sec Dev Location in 1 48e 12 4 26e 12 0624 2 06e 12 4 27e 12 0574 0605 1 29e 12 5 25e 12 1690 CHAPTER 6 AUDIO DEMONSTRATION 56 ultrawideband ranging a microsecond width pulse would need to be generated with signal power concentrated at about 1 MHz plus an
6. PFIS UPDATE PFI2 CONVERT Divide by 8192 N cik clk trigger 8192 samples to MATLAB every 8192 N clock 8192 sample Buffer DACO AINI dad el 92 s a preloaded by MATLAB DAC iM ADC 8192 sample Buffer ey a list of TDOA values those TDOAs should be very close to 0 Performing this verification the TODAs between the noncontinuously sampled single buffer blocks averaged about 1 ns which was an order of magnitude bet ter than we had hoped for We had successfully sampled the OFDM signal syn chronously at a sampling rate of 500 kHz Chapter 8 Conclusions Over the course of a year this project resulted in the successful migration of the locator demonstration from a PC sound card to the NI DAQCard 6062 E In the course of this migration we were not only able to extend the functionality of the demonstration itself but we were also able to gather performance data and run tests to determine the feasibility of further extension of the system on the DAQCard platform 8 1 Future Work Work which can be completed in the near future includes a working RF demon stration Since the DAQCard has been shown to be capable of noncontinuous syn chronous single buffer analog I O at 500 kHz it should be possible to implement a locator demo operating at 500 kHz using 250 kHz of bandwidth via the Leapfrog LF 30S tranceiver set Possibly before an RF demonstration is completed and certainly be
7. RTSIBus Timing 1 Anab Digital UO 1 B s l OM netos K Digtal 1 0 8 Timing Contfo Interface E ESSE a oY AO Control A DAC FIFO Calibration DACs CHAPTER 4 DATA ACQUISITION HARDWARE 34 Figure 4 2 Analog input block diagram Reprinted with permission of National Instruments Corporation All rights reserved AlSense Calibration ACHO AIGND DACs ACH1 Dither gt Ja a 2e ACH7 Selection ACH8 2 Acne 15 ADC FIFO ae a ACH11_ r ACH12_ Gain ACH13_ Polarity J Channel Type Calibration Sources Cal Mux e Channel Number for all the channels so true simultaneous sampling is not possible the channels are multiplexed into the ADC See Figure 4 2 for a block diagram of the analog input subsystem 4 2 2 Analog Input Timing Signals Of the 8 timing signals available on the DAQCard 6062 E for control and analysis of analog input only 3 were used in this project Those signals include TRIG1 CONVERT and STARTSCAN The TRIGI signal as used in our project initiates the data acquisition sequence This can mean different things in different situations If the DAQCard is configured for a single buffer acquitision of 1000 samples then TRIGI initiates the on
8. generation at rates much higher than 25 kHz The best consumer sound cards available can sample signals at 46 kHz but as we planned for a robust system that could be tested in a wide range of frequencies we had to look at hardware designed specifically for PC based data acquisition 4 1 Considerations for Hardware Selection The major deciding factor in choosing our data acquisition hardware was the fact that our demonstration system needed to be portable that is we required mobility to demonstrate the rapid deployment capabilities of the final system In order to accomplish this we envisioned an ideal demonstration involving 5 receivers and 1 transmitter with each reference node connected to its own laptop Since the sensors would be interfaced to the data acquisition hardware and then to the laptop 31 CHAPTER 4 DATA ACQUISITION HARDWARE 32 a PCMCIA bus solution seemed the most appropriate choice for interfacing with a laptop After conducting research we concluded that the only company currently offering data acquisition hardware for a PCMCIA bus at the analog output and input transfer rates we required was National Instruments and the specific piece of hardware we purchased was the DAQCard 6062 E 4 2 DAQCard Operation The National Instruments DAQCard 6062 E is a multifunction analog digital and timing I O data acquisition device The DAQCard features a 12 bit A D con verter ADC two 12 bit D A converte
9. large numbers of runs without having to manually record all the data ourselves To solve this problem a piece of code was appended to the end of the main audio demonstration code which put all of the relevant statistical data in a single matrix and used the MATLAB command dlmwrite to append the matrix to an existing TXT file which contains the statistical data from any previous runs since the last time statistics were cleared from the file The TXT file is then read by dlmread which puts all of the data from all the recorded runs into a single matrix Finally an unofficial tool called the MATLAB HTML Toolbox takes the contents of that matrix and outputs it to a file in the form of an HTML table HTML header data is manually added using MATLAB s file I O functionality As a result of this every time the duplex4 code is run a row is appended to a table in an HTML document with the newest statistics The result is easy to read and interpret see Figure 6 5 http www vision ime usp br casado matlab htmltoolbox CHAPTER 6 AUDIO DEMONSTRATION 53 Figure 6 5 Output of the HTML statistics generator A Elle Edit View Navigation Bookmarks Window Help aa a AD Identity as Opera Address he darius Media Docs MyDocs ECE MQP Data testdata Jan 8 04 html amp o Google search Search 100 Num Cycles Recv
10. T T T T T T 20r Peak Peak Voltage at Receiver for 1V peak peak at Transmitter dB o 30 1 1 1 1 1 1 1 1 21 22 23 24 28 29 30 25 26 27 Frequency Transmitted kHz Figure 2 4 Frequency Response of T R 25 18 Y Ultrasonic Transducers dB Although this bandwidth performance was unacceptable at least the frequency response seen in Figure 2 4 showed that the parts performed to their specification 2 4 Vifa 40 kHz Tweeters The most expensive pair of transducers tested was a pair of symmetric Vifa 40 kHz tweeters As seen in Figure 2 5 measuring from our defined lower bound of 20 kHz the 3dB bandwidth was 40 kHz which was exactly in the desired frequency range 2 5 Final Transducer Analysis If we were to use any ultrasonic transducers the Vifa set with its excellent fre quency characteristics would certainly have been the one to use as its perfor CHAPTER 2 ULTRASONIC TRANSDUCERS 22 Frequency Response of 40 kHz Tweeters dB 21 T T T T Y Y e o a bal a pk pk Voltage Response to 1V pk pk Input Sine Wave dB o Figure 2 5 Frequency Response of Vifa 40kHz Tweeters dB mance justified its price tag at 200 a pair Ultimately we chose not to construct an ultrasonic based test system The choice was made on the basis of the analog that could be established between an audio frequency wavelength in air and an RF wavelength through vaccuum The accuracy of the location algor
11. Tweeters dB Frequency Response of Radio Shack 27kHz rated Tweeter T T T 5L 4 pk pk Voltage Response to 2V pk pk Input Sine Wave dB 19 20 21 22 23 24 25 26 27 f kHz Figure 2 2 Frequency Response of Radio Shack 27kHz Tweeters dB CHAPTER 2 ULTRASONIC TRANSDUCERS 20 Frequency Response of the Other Radio Shack Tweeters T T T pk pk Voltage Response to 2V pk pk Input Sine Wave dB 34 1 1 19 20 21 1 1 22 23 24 25 f kHz Figure 2 3 Frequency Response of Unidentified Radio Shack Tweeters dB again subjected to a 2 V sinusoid Although the 3 dB bandwidth is again about 500 kHz the 10 dB bandwidth ranges from 23 kHz to 25 5 kHz Finally Figure 2 3 shows the response to a 2 V sinusoid of a pair of Radio Shack tweeters of unknown specification the tweeters lacked any accompanying docu mentation The response of these tweeters peaked around 20 5 kHz with a 3 dB bandwidth of 1 KHz The 10 dB bandwidth ranges from 19 kHz to 21 5 kHz 2 3 Ultrasonic Transducer Pair A more expensive model of transducer tested was a pair of T R 25 18Y ultrasonic transducers This pair of tranducers was asymmetric with T 25 18Y designating the transmitter and R 25 18Y designating the receiver The tests yielded about 700 Hz of bandwidth between the 3dB points with peak response at 25 6 kHz CHAPTER 2 ULTRASONIC TRANSDUCERS 21 Frequency Response of Coupled T R 25 18Y Ultrasonic Transducers 30
12. Used Signal Power Signal Noise SNR Bua Dev of SS Estimator Var actual Var pred Std Dev Bard Solver 100 0 3 1e 004 1 04e 005 10 5 1 92e 006 3 67e 012 4 23e 012 0 0727 100 0 6 13e 004 1 02e 006 24 4 5 02e 006 2 52e 011 2 09e 011 0 0924 100 0 6 01e 004 2 19e 005 11 2 1 23e 006 1 52e 012 4 58e 012 0 0546 100 0 685e 004 2 55e 006 31 4 2 69e 006 725 012 4 66e 011 0 0891 100 0 6 44e 004 9 38e 005 23 3 1 23e 006 1 52e 012 1 83e 011 0 0864 100 0 4 28e 004 4 19e 005 19 8 2 12e 006 4 5e 012 1 23e 011 0 0569 100 0 4 49e 004 5 3e 005 214 2 56e 006 6 57e 012 1 48e 011 0 0762 100 0 7 25e 004 4 82e 005 16 5 2 09e 006 4 37e 012 8 35e 012 0 0824 100 0 4 85e 004 3 34e 005 16 8 1 22e 006 1 49e 012 8 64e 012 0 0498 100 0 5 44e 004 127e 005 7 39 1 09e 006 1 2e 012 2 94e 012 0 067 100 0 3 188 004 2434005 17 7 2 5e 006 6 24e 012 9 58e 012 0 152 100 0 4 42e 004 1 51e 005 10 7 1 61e 006 2 58e 012 4 29e 012 0 0889 100 0 4 168 004 ES 241 314e006 984e 012 202e 011 0368 100 0 5 14e 004 1 278 005 7 87 1 12e 006 1 25e 012 3 11e 012 0 0647 100 0 3 06e 004 1 02e 005 10 5 1 25e 006 1 55e 012 4 18e 012 0 162 100 0 6 09e 004 2 52e 005 12 3 1 39e 006 1 93e 012 5 2e 012 0 0622 100 0 2 92e 004 1 36e 005 13 3 2 79e 006 7 78e 012 5 82e 012 0 0629 100 0 5 5e 004 2 76e 005 14 1 62e 006 2 64e 012 6 29e 012 0 1 100 0 3 64e 004 205e 005 15 3 15e 006 9 93e 012 7 07e 012 0 0788 100 O 3 88e 00
13. and DLLs which provide the user with a convenient front end to the driver software for a given hardware device While the user interacts with the toolbox the toolbox is passing data to and receiving data from the device drivers The device drivers communicate directly with the hardware in the background Data Acquisition Toolbox documentation 1994 2004 The Math Works Inc 38 CHAPTER 5 DATA ACQUISITION SOFTWARE 39 5 1 1 The Data Acquisition Session Any program which requires data from an external device be it a PC sound card or a multifunction DAQ has a program flow as follows 1 Initialization device objects are created 2 Configuration object properties are written such that number of I O chan nels voltage ranges sampling frequencies etc are set to desired values 3 Execution device objects are armed at which point functionality may be triggered internally or externally 4 Termination device objects are deleted A device object behaves like any object in MATLAB An object constructor can be called to create a device object which then has certain properties depending on which constructor was used to create the object Three kinds of objects are available for use in the Data Acquisition Toolbox analoginput analogoutput and digitalio Each object type is associated only with one particular subsystem on the hardware device with which it interfaces Of the three object types listed only
14. connections 6 3 2 Software At this point in the project the code for the locator algorithm had been extended to accept a variable number of input channels previously it was hard coded to accept 2 inputs new system wide variable named N out was added to the software This variable simply specified the number of input channels to use Not only was N out passed to the locator algorithm but it was also used when physically adding channels to the hardware With this implemented the locator algorithm would by definition expect the same number of input channels as are allotted in hardware Also the channelSkew value divided by 2 This reflects the fact that this milestone included 2 times as many receivers as were included in the last milestone and thus the channels needed to be multiplexed twice as fast CHAPTER 6 AUDIO DEMONSTRATION 52 6 4 Other Revisions Several features were added to the demonstration software in the course of the project that were not directly related to improving position estimation 6 4 1 HTML Statistics Logger Once the audio demonstration was complete we began collecting statistical data during each run of the demonstrator At first the program simply output the relevant statistical data to the MATLAB command window at the end of each run Eventually we determined that an automatically logged record of all statistics would be desirable so that we could easily calculate performance averages over
15. of the signals used in the audio demon stration versus the upper and lower bounds of a comparable RF system The 5 kHz wide signal used in the audio demonstrations corresponds to a 5 GHz ultra wideband signal Hence by using audio frequencies in the 2 16 kHz to 7 54 kHz range we attain a 1 1 scale model of an RF system in our target band The only necessary change in the location algorithm is the replacement of the speed of light by the speed of sound in air when solving for position from the TDOAs CHAPTER 2 ULTRASONIC TRANSDUCERS Table 2 1 Audio RF Correspondences Audio Freq in Air Wavelength Radio Freq in Vaccuum 2 16 kHz 1588 m 1 88 GHz 7 54 kHz 0455 m 6 57 GHz 24 Chapter 3 Audio Driver As the data aquisition card contains no onboard driver circuitry it was necessary to design and to implement preamplifiers for the microphones and a power amplifier for the speaker 3 1 Microphone Preamplifiers The variable gain preamplifiers as shown in Figure 3 2 is a simple non inverting op amp configuration using the LM324 a commonly available single supply op amp Of note is the voltage divider which creates a reference voltage of Kee or 2 5 V in this case By referencing the op amp to this voltage the output of the pream plifier maintains a DC offset of Kec thus centering the output and preventing clipping from occurring 25 CHAPTER 3 AUDIO DRIVER 26 Figure 3 1 Block diagram of audio drivers in
16. system Laptop Transmitter DAQCard 6062 E B gt Power Amplifier gt 1 i Preamplifiers Reference Voltage Divider Microphone Preamp Circuit 9y 9V o E R1 z 1k 1 45V e m 100u R5 S100 R3 1k amp 0 Mo R6 1k 4 5V Figure 3 2 Microphone Preamplifier Schematic CHAPTER 3 AUDIO DRIVER 27 Power amplifier for speakers T13V R1 100k c1 14H DAC OUT ANN l 214 DUF xpo z 10uF VE ps R GND u 1k E l T bypass gt l 0 0 nic qo Sele A 1 R3 gt ZN 4 C2 i our d ca 429 LM380 de AUF D le N VY b SPEAKER Figure 3 3 Power Amplifier Schematic 3 2 Audio Power Amplifier The audio power amplifier is based on the LM380 a fixed gain 2W audio power amp All details about specific connections can be gathered from the schematic in Figure 3 3 but there are some physical constraints that are of note We discovered after testing initial circuit designs that the output of the LM380 tended to oscillate rapidly between the supply rails After some troubleshooting it was determined that this instability was due to feedback caused by power supply coupling To correct for this problem we had to keep all connecting wires as short as possible In addition we required 100uF capacitors on the power bus at pin 14 V by placing the capacitors on the power supply near the supply pins for the IC we e
17. 00 kHz bandwidth signal we needed transduc ers with appropriate characteristics Although an RF transmitter for our applica tion was being developed at the same time by another group associated with the DOJ sponsored efforts we needed something immediately so that we could test the robustness of our signal when its frequency is stepped up transmitted at RF stepped down and then analyzed by MATLAB We looked at several commercial RF transmitters until we found one that we felt would meet our specifications The Terk Leapfrog LF 30S wireless audio video transmitter and receiver system is acommercially available device whose purpose is to transmit A V signals wire lessly around a home at 2 4 GHz The LF 30S broadcasts using FM modulation at frequencies from 2 4 GHz to 2 48 GHz The system can be set to use one of 4 selectable channels in its broadcast range During tests we discovered that the 2 4 GHz carrier frequency of the wireless IEEE 802 11b network in our building interfered with the operations of the LF 30S but by changing the transmission channel we were able to discover a channel where interference was negligible Before using the LF 30S we needed to find its transmission bandwidth To mea sure the bandwidth we connected the tracking oscillator output from a spectrum analyzer to the input of the video link transmitter then connected the video link receiver output to the input of the spectrum analyzer We obtained the frequ
18. 2 V peak to peak amplitude depending on the transducers being measured The peak to peak voltage of the received signal on channel 2 was recorded and the function generator was advanced manually to the next desired frequency In the case of asymmetric sets 1 e where the manufacturer designated a transmit ter and a receiver the same method was used except care was taken to use the given devices as they had been designated by the manufacturer Recorded data was entered into MATLAB and the following frequency response plots were generated 2 2 Radio Shack Tweeters There were three Radio Shack brand tweeters tested for their ultrasonic response Although not much performance was expected from the tweeters in terms of band width they were cheap and easy to acquire and as such at least provided a com parison point to see if more expensive specialized equipment was truly worth the cost Figure 2 1 shows the response of a 25 kHz rated tweeter to a 2 V sinusoid of varying frequency The peak response occurs at 19 5 kHz with a 3 dB bandwidth of about 500 Hz The 27 kHz rated tweeter in Figure 2 2 shows a more pronounced peak at 24 kHz CHAPTER 2 ULTRASONIC TRANSDUCERS 19 Frequency Response of Radio Shack 25kHz Tweeters 5 T T T T T pk pk Voltage Response to 2V pk pk Input Sine Wave dB S o T 1 35 1 1 1 1 1 1 1 19 20 21 22 23 24 25 26 27 f kHz Figure 2 1 Frequency Response of Radio Shack 25kHz
19. 4 1 83e 005 13 5 1 19e 006 1 42e 012 532e012 0 0774 20 0 2 8e 004 1 59e 005 15 1 1 9e 006 3 62e 012 7 12e 012 0 0989 E Blank p BAN of our t El Transfers QHTMatLab Additionally the path and filename of the HTML log file to be written was eventu ally incorporated into the locator GUI as seen near the bottom of Figure 6 6 6 4 2 Graphical User Interface A graphical user interface GUI was created as an intuitive and user friendly front end to the audio locator program As can be seen in Figure 6 6 the GUI not only provides the geometric representation of the fixed receiver locations and the predicted transmitter location but it also provides the user the ability to start stop and configure new test runs The user can pick from graphical and text display options choose whether to display statistical data and troubleshoot the program among other options CHAPTER 6 AUDIO DEMONSTRATION 54 Figure 6 6 Screen capture of the audio locator GUI Audio Locator Demo System a 10 xf Location Solver Text Display Graphical Display I Rx Tx Distance in Tx Location Transmitter location using 4 receivers T TE T T TxLocation T TOA Circles I Variances 0 Z Ax 1 Display TOA Sorting Method 1 Signale Present Thresholding Y Bx and Tx Setup C Troubleshooting 4 z ax o Fe re 7 Calibration Y Carrier Phase distance in
20. 9V battery CHAPTER 6 AUDIO DEMONSTRATION 48 6 2 20 Software Much of the code for interfacing with the hardware systems remained the same between Milestone 1 and Milestone 2 The portability of the Milestone 1 code was due to the fact that every device object created by the Data Acquisition Toolbox has the same set of base properties The object oriented nature of the toolbox sim ply necessitated that the objects were constructed to communicate with a different driver set in this case the NI DAQ 6 9 3 drivers What follows are those changes that were made ai analoginput nidaq 1 ao analogoutput nidaq 1 These two commands create analog input and analog output objects associated with the I O subsystems on the DAQCard 6062 E although the commands would be compatible with any NI device using the NI DAQ drivers addchannel ai 0 1 addchannel ao 1 One analog output channel and two analog input channels are created for the 2 receivers and 1 transmitter necessary for the demonstration Also of note is that channels on the NI DAQ device are identified starting from 0 whereas the PC sound card channels as seen in Section 6 1 1 are indentified starting from 1 set ai Channel InputRange 0 10 set ai InputType SingleEnded set ai ChannelSkew 2 265e 005 2 CHAPTER 6 AUDIO DEMONSTRATION 49 The first line of code sets the voltage range expected at th
21. Acknowledgements I would like to thank the Department of Justice and the National Institute of Justice for funding support without which this project would have been impossible to undertake I would also like to thank Professor David Cyganski for providing me with guidance and insight over the year I spent on this project Abstract This paper discusses the creation of a demonstration of precision TDOA based geolocation on an extensible general purpose test platform incorporating multiple receivers operating at a wide range of frequencies Data was collected to show that nanosecond order TDOA accuracy can be achieved using frequencies in the audio range Contents 1 Introduction 10 11 System Architecture 2 2 o o 11 1 1 1 The Transmitter Signal 13 1 12 System Overview suos Rom Ow a 13 12 ProectGoal eA 14 13 Original Demonstration lee 14 T MOP Go ec iii a bet drap a WAS uos 15 2 Ultrasonic Transducers 17 2 1 Test Configuration een 17 2 2 Radio Shack Tweeters leen 18 2 3 Ultrasonic Transducer Pair o 20 2 4 Vifa40 kHz Tweeters ee 21 2 5 Final Transducer Analysis o 21 CONTENTS 3 Audio Driver 3 1 Microphone Preamplifiers 3 2 Audio Power Amplifier 3 2 1 Aluminum Reflector 3 3 Acrylic Resonators 4 Data Acquisition Hardware 4 1 Consideration
22. after the object is deleted As a result of this finding a vector of zeroes was written to the analog output device just before disarmament and deletion which ensured no distorition due to transients would occur in consecutive runs of the program CHAPTER 6 AUDIO DEMONSTRATION 50 Figure 6 3 Photograph of the Milestone 3 configuration 6 3 Milestone 3 This iteration of the indoor geolocator was configured for four audio receivers and one audio transmitter The four receivers remained at fixed locations the coordi nates of which are specified in code Each microphone was embedded in an acrylic resonator as described in Section 3 3 See Figure 6 3 for a photograph of the hardware configuration 6 3 1 Hardware Not very many changes to the hardware were necessary for this milestone due to the inherent extensibility of our system most of the changes were made in soft ware The exception to this was the creation of two additional microphone pream CHAPTER 6 AUDIO DEMONSTRATION 51 Figure 6 4 Milestone 3 signal connections Note the addition of two analog input channels and the removal of an analog output Mic Preamp 9 ACHO DACOOUT 3J3 Speaker Driver Mic Preamp J9 ACHI Mic Preamp 3 gt gt ACH2 Mic Preamp 3J9 ACH3 AIGND AOGND plifier circuits and resonators necessary for the addition of new channels See Figure 6 4 for
23. ber of reference nodes placed outside a target area receiving and collaborating with one another the data from transmitters on the personnel inside a target area The reference nodes send received data to a command and control unit which uses that data to extrap olate the position of each transmitter with a high degree of accuracy See Figures 1 1 and 1 2 for depictions of the reference node and transmitter configuration The reference nodes are all synchronized with each other and they collaborate time of arrival TOA data relative to a master clock to obtain time difference of arrival TDOA data Using receiver relative TDOA data means that the transmit ters need not be synchronous with the rest of the system which leads to greatly simplified circuitry on the transmitter side Thus the complex equipment can re main outside of the target area of the emergency The transmitter position is estimated using the TDOA data and the positions of the reference nodes relative to an arbitrarily defined origin Thus reference nodes can be placed anywhere outside a target area The command and control unit can be defined as an origin and the relative positions of the reference nodes can be determined using a positioning algorithm specifically used to determine node po sitions This process defines an ad hoc coordinate system to be used in all further calculations Cyganski D et al A Multi Carrier Technique for Precision Geolocation for Ind
24. e inputs of either chan nel to between 0 V and 10 V The analog inputs are then configured to referenced single ended RSE as described in Section 4 2 1 Finally the channelSkew pa rameter is set to an empirically determined value The value of channelSkew deter mines the time between sampling of consecutive channels during a channel scan In terms of hardware control channelSkew simply dictates the amount of time be tween pulses of the CONVERT signal during a STARTSCAN cycle as outlined in Section 4 2 2 ZZZ zeros 3 1 putdata ao ZZZ These last lines occur just before the device objects are disarmed and deleted and were added as the result of a problem encountered during initial device tests We conducted our first tests of the DAQCard not using the OFDM signal as our output signal but rather we used a 1 kHz sinusoid We quickly discovered that upon loading MATLAB the first time we would attempt to run a program whose output waveform was a sinusoid everything would work as predicted However the next consecutive run of the program caused significant distortion and overdrive of the outputs This distortion was identified as a transient on the output and after further testing we discovered that disarming and deleting an object does not necessarily reset the state of its outputs So if the analog output device is disarmed at the crest of gen erating a sinusoid of a 2 V amplitude the output remains at 2 V indefinitely
25. e shot acquisition In the case of a double buffered acquisition then TRIG1 initiates the continuous waveform acquisition which operates indefinitely until a stop com CHAPTER 4 DATA ACQUISITION HARDWARE 35 Figure 4 3 TRIGI timing diagram Reprinted with permission of National Instru ments Corporation All rights reserved Rising Edge Polarity Falling Edge Polarity ty 10 ns minimum Figure 4 4 CONVERT timing diagram Reprinted with permission of National Instruments Corporation All rights reserved ta gt Rising Edge Polarity Falling Edge Polarity i ty 10ns minimum mand is issued TRIGI is edge triggered and may be configured as rising edge or falling edge in software See Figure 4 3 for timing data The CONVERT signal controls the rate at which data aqcuisition occurs After TRIGI is initiated the rising or falling edge depending on user configuration of CONVERT causes the ADC to convert the analog input to a digital value and send it to the onboard FIFO to be later read by the PC see Figure 4 2 See Figure 4 4 for timing data The STARTSCAN signal is relevant when there is more than one analog input to be read and when CONVERT is internally generated For N channels to be read each rising or falling edge of STARTSCAN initiates N pulses of the CONVERT CHAPTER 4 DATA ACQUISITION HARDWARE 36 Figure 4 5 Timing sequence of TRIGI STARTSCAN and CONVERT
26. ency CHAPTER 7 RF DEMONSTRATION 59 Figure 7 1 Frequency response of the LeapFrog A V transceiver system response shown in Figure 7 1 As can be seen in the figure the LF 30S has a 6 MHz tranmission bandwidth far exceeding our 250 kHz bandwidth requirement Unfortunately once the LF 30S was integrated into the system we discovered that the DAQCard was incapable of sampling the waveform at the rates we desired CHAPTER 7 RF DEMONSTRATION 60 7 2 Data Acquisition Although the DAQCard 6062 E is specified as being capable of providing 500 ksamples sec of continuous I O actually obtaining that level of throughput for our RF signal proved to be far more difficult than initially imagined 7 2 1 Throughput of NI DAQ 6 9 3 Drivers The first roadblock in the process of developing a 500 ksample sec acquisition sys tem was the implementation of the NI DAQ 6 9 3 driver function set see Section 5 2 for a detailed description of the software driver functionality All of our in terfacing with the DAQCard is done through this particular set of driver functions regardless of whether or not we use the MATLAB front end or we access the C functions in the DLLs that are called from MATLAB using the loadlibrary func tion In order to initiate continuous I O with an external clock source we needed to use the DAQ_DB_ Transfer function Unfortunately it was empirically determined by changing the frequency of the external sampling cloc
27. f the line in on a PC sound card but passed through an amplification stage before they got there see Section 3 1 Once initialized the transmitter could be moved around the test area and was tracked by the software The transmitter did not require an external power amplification stage as it was being driven by the speaker out of a sound card As seen in Figure 6 1 we required foam padding around the test area to absorb sound that would otherwise reflect off of the nearby computers or refrigerator We also had to elevate the transmitter about 5 inches off the ground so that the horizon tal plane of the speaker was the same as the horizonal plane created by the tops of prototype wooden resonators on the receivers This was to ensure omnidirectional ity of the receivers see Section 3 3 The transmitter used the aluminum reflector described in Section 3 2 1 6 1 1 Software This milestone demonstration exclusively used the MATLAB Data Acquisition Toolbox to control the interface with the PC sound card For a general overview of this process see Section 5 1 1 What follows is a walkthrough of a simplified version of the code used to create Milestone 1 ai analoginput winsound ao analogoutput winsound CHAPTER 6 AUDIO DEMONSTRATION 44 Figure 6 1 Milestone 1 two Rx tracking one Tx These two commands create analog input and analog output objects associated with the I O subsystems on a PC s
28. fore moving 65 CHAPTER 8 CONCLUSIONS 66 up to higher carrier frequencies and bandwidths it will be desirable to have each transmitter and reference node operating on its own laptop computer This requires synchronization of all reference nodes to each other in order to assure accurate TDOA extraction which can be achieved by synchronizing the GPCTR control signals for our counter trigger architecture Finally a three dimensional demonstration will be needed Although some pre liminary work has been done with the audio demonstrator to test feasibility the directionality of the speaker microphone coupling system made this rather difficult to conduct With the completion of the RF system it should be possible to explore full 3D operation with greater ease Once the demonstration is operating at high frequencies with a high bandwidth complete reference node synchronization and in three dimensions we will have a model that provides a powerful analog to the envisioned Precision Personnel Locator Appendix 67
29. gnal 1 1 2 System Overview The Precision Personnel Locator will CHAPTER 1 INTRODUCTION 14 e estimate the location of a user in three dimensions relative to a chosen refer ence point with approximately 2 cm of accuracy e provide a graphical display of user positions and paths taken at a command station e provide an audio communication link with users so that a commander can issue self rescue instructions and other vital information e operate in at least the range of a city block Possible extensions to the system include the addition of physiologic telemetry so commanders can keep track of vital signs of users as well as the addition of Geographic Information Systems GIS data overlays 1 2 Project Goal The goal of the Precision Personnel Locator project is to create technology to aid and rescue those who put the mission of our safety before their personal safety 1 3 Original Demonstration An initial demonstration was constructed in the winter of 2001 by Prof David Cy ganski and Joshua Resnick at Worcester Polytechnic Institute The demonstration consisted of two stationary speakers transmitting the OFDM signal to a single re ceiver The OFDM signal consisted of 100 carriers spaced between 2 kHz and 7 5 kHz Both of the speakers and the microphone were connected to the sound card of a tower PC and both were controlled using MATLAB s Data Acquisition Toolbox CHAPTER 1 INTRODUCTION 15 Figure 1 4 O
30. ithm is based in part on the wavelength of the transmitted location signal and wavelengths depend on both the frequency of a signal and the medium through which a signal is sent While audio signals re quire a medium of propagation such as air or water RF signals propagate through vaccuum Although RF signals oscillate at much higher frequencies than audio signals the physical wavelength of an audio signal through air corresponds to the physical wavelength of a higher frequency RF signal through vacuum This relationship is best illustrated by the equation CHAPTER 2 ULTRASONIC TRANSDUCERS 23 speed frequency wavelength Sound travels at 343 m s in air while an RF signal travels through vaccuum at the speed of light 2 998 10 m s We can substitute the speed of sound along with an arbitrary audio frequency to get a wavelength We can substitute the wavelength back into the equation with the speed of light to get the corresponding radio fre quency For example the lowest frequency used in our audio demonstration was 2 16 kHz The following equations yield the corresponding radio frequency for that wavelength through vaccuum speed 343 m s length 1588 waveleng aaa 2160 s 1 m speed 2 998 108 m s71 inne 1 8810 1 88 GH frequency adog 1588 m s GHz Using these equations we can calculate the data shown in Table 2 1 which con tains the lower and upper frequency bounds
31. k that the function would fail at rates above roughly 200kHz of thoughput If the sampling rate is set through external hardware to be any higher than that the computer stalls on the DAQ_DB_ Transfer call A conjecture which was confirmed by National Instruments engineers is that this was due to the use of a while loop in the DAQ_DB_Transfer call which slowed down the entire continuous I O process We could not alter the drivers themselves to remove the while loop as they came as precompiled libraries CHAPTER 7 RF DEMONSTRATION 61 7 2 2 NiDAQmx Testing National Instruments engineers informed us that a new version of NI DAQ called NiDAQmx offered roughly a 20 fold performance increase over our current driver set Since all we required was a 3 fold increase even a conservative performance increase in the new driver set would have been satisfactory Unfortunately any performance gains in the new driver set would never be seen After two weeks of testing the new driver set we discovered that the loadlibrary function embedded in MATLAB 6 5 R13SP1 could not call the C functions provided by the new NiDAQmx This is because the new drivers use 64 bit integers int64 uint64 int64Ptr and uint64Ptr with which MATLAB is not yet fully compatible 7 3 Counter Trigger Synchronization Method The fact that the NiDAQmx drivers were not compatible with MATLAB made it seem as though we had hit a roadblock that we could not navigate pa
32. llle 30 Block diagram of DAQCard 6062 E operation Reprinted with permission of National Instruments Corporation All rights reserved 33 Analog input block diagram Reprinted with permission of Na tional Instruments Corporation All rights reserved 34 TRIGI timing diagram Reprinted with permission of National Instruments Corporation All rights reserved 35 CONVERT timing diagram Reprinted with permission of Na tional Instruments Corporation All rights reserved 35 Timing sequence of TRIG1 STARTSCAN and CONVERT Reprinted with permission of National Instruments Corporation All rights Teservedy 3 ae aser E Ba Ak OE a ade Co 36 Milestone 1 two Rx tracking one Tx 44 Connections to the SCB 68 connector block Note that the config uration is flexible for a2 Tx 1 Rx or 1 Tx 2 Rx demonstration 47 Photograph of the Milestone 3 configuration 50 Milestone 3 signal connections Note the addition of two analog input channels and the removal of an analog output 51 LIST OF FIGURES 8 6 5 Output of the HTML statistics generator 53 6 6 Screen capture of the audio locator GUI 54 7 1 Frequency response of the LeapFrog A V transceiver system 59 7 2 Block diagram of counter trigger control configuration 64 List of Tables 2 1 Audio RF Correspondences a 6 1 Statistical data fro
33. m 15 runs of the Milestone 3 demonstration at 1000 cycles per run so mo rr ee a Chapter 1 Introduction On December 3 1999 six firefighters died in a warehouse fire in Worcester MA Due to the lack of any visibility whatsoever inside the warehouse two firefighters and the search teams that entered the building to rescue them perished literally 100 feet from safety This posed a question to engineers how can technology be developed to prevent this kind of tragedy from happening in the future The Worcester cold storage fire underscored the importance of precision position ing information to the men and women who risk their lives by being the first people on the scene of a disaster Not only do first responders require position informa tion but any technology that they use needs to be quickly deployable so that time is not spent on equipment calibration but rather spent on the urgent task at hand A precision positioning system that quickly and automatically configures to new environments has the potential to help not only firefighters but law enforcement officers military responders and corrections officers The Department of Justice and the National Institute of Justice provided our group with funding for the cre ation of a demonstration based on techniques documented in papers by Cyganski 10 CHAPTER 1 INTRODUCTION 11 et al 1 1 System Architecture At its core the Precision Personnel Locator consists of a num
34. nates at the frequencies that need to be amplified the frequency is determined by the length of the shaft created by the hole which creates a physical band boost filter in effect amplifying sound via resonance at around 5 kHz More importantly than applying a physical band boost filter the resonators caused the received signal to be omnidirectionally phase coherent The microphone regis tered the sound passing over the resonator opening from any direction For stability each resonator was equipped with a base The base is a rough cut 2 square of 1 4 thick Lexon with a mounting toroid affixed to it with epoxy The toroid is a 1 8 cut of the second largest acrylic tube This allows the largest tube to fit tightly over it creating a base that can be fixed to the table If a microphone needs servicing the microphone unit can be removed from the base without com promising the position in which it had previously been set For photographs of the final result see Figure 3 5 CHAPTER 3 AUDIO DRIVER 30 Figure 3 5 Photographs of an acrylic resonator Left 3 4 view of resonator Right microphone is fit inside a cavity recessed into the tubing and then fit over toroidal base mount Bottom overhead view shows interlocking acrylic rings Chapter 4 Data Acquisition Hardware Although the original demonstration used a standard PC sound card to handle sig nal I O sound cards are not designed to handle data acquitision and waveform
35. nsured an AC ground at the chip 3 2 1 Aluminum Reflector In order to ensure that sound generated by the speaker which faces up away from the table propagates towards the microphones in the manner of a two dimensional CHAPTER 3 AUDIO DRIVER 28 Figure 3 4 The aluminum cone causes sound to propagate along a horizontal plane speaker omnidirectional point source a 45 degree metal cone was crafted from aluminum sheeting The cone is suspended over but not touching the center of the speaker See Figure 3 4 for an illustration 3 3 Acrylic Resonators The microphones were found to be extremely directionally sensitive so a set of audio resonators was built to alleviate this problem The acrylic resonators were made from acrylic tubing acquired from the ECE Shop They consist of three tubes of decreasing diameter one fitting tightly inside the other The smallest tube has an internal diameter of just under 3 8 The micro phone s diameter is just over 3 8 so after a little bit of filing done on the internal aperture the microphone fit snugly inside The internal two tubes are cut to 3 whereas the outer tube is 3 1 2 This gives 1 2 of space inside for the base mount and the wires A notch was cut into the side of the resonator which provides room for external wires to reach the internally mounted microphone CHAPTER 3 AUDIO DRIVER 29 Sound that passes over the top side of the acrylic tube reso
36. oor Multipath Environments ION GPS DNSS 2003 Conference September 9 12 2003 Portland USA Cyganski D et al Performance of a Precision Indoor Positioning System Using a Multi Carrier Approach ION National Technical Meeting 2004 San Diego California January 26 28 2004 CHAPTER 1 INTRODUCTION 12 Figure 1 1 Block diagram showing communication between system components GPS Signal os Personnel e MN Unit Miia pe Reference oe Phys Unit known a aN Moni P MU Monitor location J Jur A P EA Reference Unit known Command and Control location Unit y Reference Unit known location GPS reference dd Positioning signal System control User Commander link Figure 1 2 Geometric overview of system component placement e 802 11 Data Side Channels Control Unit CHAPTER 1 INTRODUCTION 13 Figure 1 3 The OFDM concept Constellation and bit rate and power chosen per channel Channels spaced Af Pulses transmitted at rate 1 Af 1 1 1 The Transmitter Signal The transmitters generate a periodic orthogonal frequency division multiplexed OFDM signal OFDM is a digial modulation technique which typically involves sending modulated carriers in parallel with frequency spacing Af such that Af is the minimum frequency spacing required for the signal to satisfy orthogonal ity and hence detectability conditions See Figure 1 3 for an illustration of an OFDM si
37. ound card addchannel ai 1 addchannel ao 1 2 One analog input channel and two analog output channels are created for the 2 transmitters and 1 receiver necessary for the demonstration set ai ao TriggerType Manual This sets the trigger type to manual meaning that once the device is armed a software trigger command has issued before any data acquisition or waveform generation may begin CHAPTER 6 AUDIO DEMONSTRATION 45 set ai SampleRate 44100 set ai BitsPerSample 16 set ai SamplesPerTrigger inf set ao SampleRate 44100 set ao BitsPerSample 16 set ao RepeatOutput runloopt4 The first three lines tell the sound card drivers to sample incoming analog audio data at 44 1 kHz with 16 bits of resolution continuously until a step command is issued The last three lines tell the sound card drivers to generate any waveforms at 44 1 kHz with 16 bits of resolution and to repeat the waveform being generated runloop 4 times where runloop is an integer set earlier in the demonstration code putdata ao data_out start lai ao trigger lai ao The putdata function queues the contents of the data_out vector into the data ac quisition engine so that it can be directly accessed by the driver software in this case for the purpose of analog output The start function arms the analog input and output subsystems of the hardware and the trigger f
38. period of the OFDM signal and is 8192 samples in length In addition to generating the waveform it also generates the UPDATE signal When configured as an output UPDATE reflects the actual update pulse that is connected to the DACs The OFDM signal is generated on the DACO line and is connected to the analog input channel 1 line ACH1 on computer B s DAQCard The UPDATE signal is asserted to the PFIS UPDATE line which is connected to the PFI CONVERT line on computer B s DAQCard Although the ADC on computer B is continually reading data at its input and is clocked at the same rate as computer A is generating data it does not begin analog to digital conversion until it receives a start trigger The start trigger comes from the general purpose counter GPCTRO which is configured to generate a pulse train DAQCard 6062 E User Manual National Instruments June 2002 CHAPTER 7 RF DEMONSTRATION 63 7 3 2 Pulse Trains The general purpose counters on the DAQCard 6062 E have a set of control reg isters which can be written to cause different behaviors and any of these counters can be configured to generate a pulse train A pulse train needs to know three things to function a valid source identifier an integer hiCount and an integer loCount The counter is triggered to count on the rising edge of the signal specified by source The counter s output is high 5 V until it reaches the value specified by hiCount at which poin
39. riginal demonstration photograph and screen capture with multiple solutions Eb Edt Vew Inset Tools Window Heb JDSGUSBRAAS POD Transmitter location HS two receivers The position of the microphone was correctly tracked within a specific test area and the algorithm correctly identified located obstacles creating reflections near the microphone Also due to the number of reference nodes the algorithm solved for position using TOA rather than TDOA data Although the positioning solutions were correct they were certainly not accurate as the predicted location of the microphone experienced roughly 2 cm of jitter Nor was the solution unique Due to the presence of only two reference nodes the algorithm yielded either 0 1 or 2 solutions depending on the location being solved for see Figure 1 4 1 4 MQP Goal The original demonstration consisted of a two dimensional TOA based solution limited to 2 transmitters and 1 receiver at the audio transmission range These limitations were based on the platform on which the demonstration was imple mented a consumer PC sound card The purpose of this MQP was to migrate CHAPTER 1 INTRODUCTION 16 the demonstration to a more extensible general purpose test platform so that the demonstration could incorporate multiple receivers operating at a wide range of frequencies could be executed across multiple computers could be synchronized to an externally genera
40. rols the rate at which the DAC is updating its output Since the transmitter in our system is asynchronous the UPDATE signal was never externally controlled and was only used as a timing reference for the input conversion rate of other DAQCard units as explained in Section 7 3 4 2 5 General Purpose Counter The DAQCard comes with the DAQ STC System Timing Control application specific integrated circuit ASIC with seven 24 bit counters and three 16 bit coun ters While almost all of these counters are reserved for use by the analog input and analog output subsystems the DAQCard leaves two 24 bit registers for general use These general purpose counters are referred to as GPCTRO and GPCTRI and can be accessed through a family of gpctr functions via the NI DAQ driver set see Section 5 2 for a description of the software drivers For a detailed description of general purpose counter usage in this project see Section 7 3 Chapter 5 Data Acquisition Software The DAQCard 6062 E can be controlled several ways National Instruments pro vides a set of dynamically linked libraries DLLs with C functions that can be called from any program that includes the libraries In addition MATLAB has a Data Acquisition Toolbox which supports the NI DAQ 6 9 3 driver set and NI E series devices Initially we chose to use the Data Acquisition Toolbox 5 1 MATLAB Data Acquisition Toolbox The Data Acquisition Toolbox is a set of M files
41. rs DACs eight lines of TTL compatible digital I O DIO and two 24 bit counter timers for timing I O TIO For a block diagram of the DAQCard see Figure 4 1 4 2 1 Analog Input The DAQCard features 16 analog input lines at 12 bit resolution which can be configured as 16 referenced single ended RSE or 8 differential inputs Since our microphone preamplifiers are single supply we use the analog inputs in RSE mode The input voltage range is programmable but can be chosen from several sets of ranges between 10 V and 10 V in single ended mode The ADC which samples the analog input lines can operate at conversion rates of up to 500 N ksamples sec where N is the number of channels being simultane ously sampled This is because the DAQCard contains only one ADC onboard CHAPTER 4 DATA ACQUISITION HARDWARE 33 Figure 4 1 Block diagram of DAQCard 6062 E operation Reprinted with per mission of National Instruments Corporation All rights reserved ZN Voltage Calibration ws REF DACS 8 gt Analog Mux Mode Selection Switches 8 x M 7 39 S 2 9 c 3 o lt S Trigger gt 5 Circuitry z hb T T Tage Analog Input 2 PFI Trigger Timing Contro Request Amb lecppom mps 1 Cantal Co Counter Bus Timing vo DAQ STC intortace Analog Output
42. s for Hardware Selection 4 2 DAQCard Operation 4 2 1 4 2 2 4 2 3 4 2 4 4 2 5 Analog Input Analog Input Timing Signals Analog Output Analog Output Timing Signals General Purpose Counter 5 Data Acquisition Software 5 1 MATLAB Data Acquisition Toolbox 5 1 1 5 1 2 The Data Acquisition Session Advantages and Disadvantages 5 2 NI DAQ Driver Set v6 9 3 5 2 1 Advantages and Disadvantages 31 31 32 32 34 36 37 37 38 CONTENTS 6 Audio Demonstration 6 1 6 2 6 3 6 4 6 5 Milestone L 24423 eee Gr 4 Ges dde HO eS 6 1 0 1 Hardware 0 GAl SoftWare aa ar das AMR hack RM ee a Milestone2 2 2 lees 62 1 Hardware 0 0 00 00000000 62 2 Software 4 268 o Rd Be ee Ee e Milestone 3 7 X uuu eme qued VERD S 63 1 Hardware 2 000000 0008 4 6 3 2 Sottwarte 3 edes BUS e I ord aA Other REVISIONS J ies Bia BLO hoe PP aye Dane dhe yes 64 1 HTML Statistics Logger 642 Graphical User Interface o Results i as a E E Taube A od s 7 RF Demonstration 7 1 7 2 7 3 Terk Leapfrog LF 30S o o e Data Acquisition e eo 7 2 Throughput of NI DAQ 6 9 3 Drivers 12 2 NIDAQMX Testing ies e eed eus Counter Trigger Synchronization Method 7 3 1 Counter Trigger Test Architecture
43. sition Toolbox The NI DAQ implementation of a single getdata command in the toolbox requires roughly five function calls and the creation of a while loop Chapter 6 Audio Demonstration As explained in Section 1 3 the original audio demonstration used two transmit ters and one receiver connected to a sound card The location of the receiver was tracked around the test area which measured 22 x 22 The audio demonstration was built over the course of a year from Summer 2003 to Spring 2004 as a series of milestone designs Each milestone represents a sig nificant change in the hardware configuration and functionality and as such this chapter covers the milestones in chronological order beginning where the original demonstration see Section 1 3 left off 6 1 Milestone 1 The first new iteration of the demonstration was completed in July of 2003 The purpose of this iteration was to confirm the theory that the location algorithm would still work if the function of the receivers and transmitters were reversed i e fixed receivers tracking the position of a moving transmitter 42 CHAPTER 6 AUDIO DEMONSTRATION 43 6 1 0 1 Hardware This iteration of the indoor geolocator involved two receivers audio microphones and one transmitter audio speaker The two receivers remained at fixed locations each on an opposite diagonal corner of a 22 inch square The two receivers were connected to the left and right channels o
44. st Sev eral weeks were spent wrestling with this problem until we took a step back and looked at the way the entire project functioned During a brainstorming session it occurred to us that we are only updating position information on the user end from 2 to possibly 4 times a second hence we do not need to continuously acquire data at a rate higher than a single 8192 sized sample block 2 to 4 times a second If we could figure out a way to aquire discontinuous 8192 sample blocks such that each sampled block occurred at the exact same time offset with respect to the con tinuous and repeating transmitted OFDM signal then we could replace continuous double buffered sampling with highly synchronous single buffered one shot sam pling With single buffered sampling we could use the original NI DAQ 6 9 3 CHAPTER 7 RF DEMONSTRATION 62 drivers and acquire a signal at a sampling frequency of up to 500 kHz It was at this point that we came up with our counter trigger solution which com pletely bypassed the need for a continuous software loop by handling almost all of the processing in hardware onboard the DAQCard A test system was built which proved that the single buffering solution is feasible 7 3 1 Counter Trigger Test Architecture As illustrated in Figure 7 2 the architecture of the counter trigger system consists of two computers The DAQ board attached to computer A continuously outputs its preloaded buffer which consists of one
45. t the counter resets and the output is set to low 0 V The output remains low until the counter reaches the loCount value when the counter resets and the output is inverted once again What this means is that we can set a general purpose counter so that source is the sampling clock of the OFDM waveform and so that hiCount loCount 8192 N If the counter ouput is then tied to the ADC trigger and the trigger is set as a low to high signal which means that single buffer sampling would occur once every 8192 2N samples are output read the output read count is the same because the clocks on both the continuous output computer A the counter trigger input computer B and synchronous N can be any natural number and can be tweaked to achieve the desired number of single buffer reads per second 7 3 3 Verification of Counter Trigger Method To verify that our new sampling method worked data was collected from a system set up in the manner depicted in Figure 7 2 If the time offset of the data sampled on computer B was truly constant relative to the period of the repeating waveform output by computer A then if we take multiple 8192 sample blocks of data and run them as separate simultaneous signals through the state space estimator to obtain CHAPTER 7 RF DEMONSTRATION 64 Figure 7 2 Block diagram of counter trigger control configuration Computer A Computer B Continuous Output Counter Trigger Input Clock
46. ted clock signal and could be extended to three dimensional positioning Chapter 2 Ultrasonic Transducers As we already had a set of speakers and microphones that worked in the 2 kHz to 8 kHz range we needed to find a set of ultrasonic transducers that could be used to create an ultrasonic version of the demonstration A wide range of transducers were examined from commercial tweeters for home stereo systems to expensive ultrasonic specific tweeters The tests were conducted looking for a set of trans ducers with a 3 dB bandwidth of at least 6 kHz where the lower bound of the bandwidth was greater than 20 kHz However a 10 dB bandwidth of 6 kHz would also be acceptable as within this range the signal can be amplified in software with an acceptable increase in noise levels 2 1 Test Configuration All of the transducers were tested using one of two methods In the case of symmetric sets i e no manufacturer differentiation between a trans mitter and a receiver two transducers of the same kind were designated as trans 17 CHAPTER 2 ULTRASONIC TRANSDUCERS 18 mitter and receiver The transmitter input was connected to a function generator and to channel 1 of an oscilloscope and the receiver ouput was hooked up to chan nel 2 of the oscilloscope The transmit receive pair were acoustically coupled The function generator was manually set to a sine wave of a certain frequency as mea sured on channel 1 and set to 1 V or
47. to run with two receivers and one transmitter In Figure 6 2 one will notice that there are two microphone preamplifiers and two power amplifiers The CHAPTER 6 AUDIO DEMONSTRATION 47 Figure 6 2 Connections to the SCB 68 connector block Note that the configura tion is flexible for a 2 Tx 1 Rx or 1 Tx 2 Rx demonstration Mic Preamp JM9 ACHO DACOOUT 3J3 Speaker Driver Mic Preamp 9 ACHI DACIOUT J Speaker Driver AIGND AOGND NI DAQCard was configured to run with 2 of one transducer and 1 of the other transducer so the MATLAB code runs both the 2 Tx 1 Rx demonstration and the 2 Rx 1 Tx demonstration 6 2 1 Hardware The hardware configuration was as follows the computer was connected to the DAQCard 6062 E data acquisition card via the PCMCIA bus There was a shielded cable that connected the PCMCIA card to the SCB 68 I O connector block The connector block had its analog inputs ACHO and ACHI connected to the micro phone preamps The block s analog outputs DACO and DACI were connected to the speaker amplifiers which became necessary to drive the transmitter as the DAQCard does not have any onboard power amplification See Figure 6 2 for sig nal connections Of note is the power supply for the speaker amplifier circuit a 36V 10A HP Harrison 6433B DC Power Supply which was borrowed from the shop The microphone preamplifier was powered by a
48. unction causes the subsystems to begin data acquisition and waveform generation respectively while data getdata ai 2 power process data CHAPTER 6 AUDIO DEMONSTRATION 46 end The main loop of the program occurs immediately after the device is armed and triggered For each iteration of the while loop getdata is called which writes 2P samples from the data acquisition engine to the vector data The data vector is then manipulated to the specifications of the location algorithm stop ai ao delete ai ao At this point sfop disarms the subsystems and delete removes the device objects from the MATLAB workspace 6 2 Milestone 2 The second iteration of the indoor geolocator used one receiver and two transmit ters which is the same configuration as Resnick s original demonstration Section 1 3 The two transmitters remained at fixed locations each on an opposite diag onal corner of a 22 inch square The transmitters were oriented facing toward the initialization point where the microphone was located during the initialization loop at the start of code runtime The microphone was moved about the test area once the initialization loop was completed This demonstration was effectively the same as Resnick s demonstration except all signals are processed through the DAQCard 6062 E instead of the sound card Also of note is that at the time of this milestone the hardware was configured with the option
49. y harmonics That we managed similar TOA accuracy using signals at frequencies less than 10 kHz is a testament to the utility of our solution method Chapter 7 RF Demonstration After completing the third audio demonstration milestone and proving to our satis faction that the algorithm works at audio frequencies we decided that it was time to test our system in the RF range Switching to RF meant not only a change in the frequency of signal transmission but also a change in bandwidth for the signal As mentioned in Section 2 5 the wavelength of our audio signals in air correspond physically to the wavelength of certain RF signals in vaccuum As illustrated in Table 2 1 the approximately 5 kHz of bandwidth we use in the audio range corresponds to approximately 5 GHz of bandwidth in the RF range While jumping to a 5 GHz bandwidth signal may have been desirable it certainly was not feasible as our DAQCard has a maximum analog I O transfer rate of 500 kHz We decided to test the limits of the card by attempting to construct a demon stration using a 250 kHz wide OFDM signal We encountered a number of problems in our implementation of the high bandwidth system Two problems discussed herein include the physical transmission of such 57 CHAPTER 7 RF DEMONSTRATION 58 signals and the signal processing capability of the data aquisition hardware and software 7 1 Terk Leapfrog LF 30S In order to transmit and receive a 5
Download Pdf Manuals
Related Search