IHHHi HWWBtoW ■ na liBi mm ■Siwrtll ah n ■H Mill 4 u igJn}B0j ffliiil < Digitized by the Internet Archive in 2013 http://archive.org/details/transmissionofan838wolf /My UIUCDCS-R-77-838 /ricuU^ a&f.zk. UILU-ENG 77 1701 TRANSMISSION OF ANALOG SIGNALS USING BURST TECHNIQUES by MARTIN WOLFF January 1977 DEPARTMENT OF COMPUTER SCIENCE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN • URBANA, ILLINOIS University ot Illinois t Urbana-Cham urn UIUCDCS-R-77 -838 TRANSMISSION OF ANALOG SIGNALS USING BURST TECHNIQUES BY MARTIN WOLFF January 1977 Department of Computer Science University of Illinois Urbana, Illinois 61801 This work was supported in part by Contract No. N00014-75-C-0982 and the Department of Computer Science at the University of Illinois, Urbana- Champaign, and was submitted in partial fulfillment of the requirements for the degree of Master of Science in Electrical Engineering, at the University of Illinois. Ill ACKNOWLEDGMENT The author wishes to thank his thesis advisor Professor W. J. Poppelbaum for the opportunity to be a member of the Information Engineer- ing Laboratory and for his support and guidance throughout the term of the project. He also wishes to thank all the members of the Information Engineering Laboratory for their continual friendship and assistance and their willingness to allow him to appropriate the 350 MHz oscilloscope. Thanks are due to Mrs. Kathy Gee for typing the thesis, Mr. Stan Zundo for doing the drawings, and Mr. Frank Serio and his staff for their help in putting the project together. Finally he would like to offer the classical thanks to his wife Lois for her support and encouragement. IV TABLE OF CONTENTS Page 1. INTRODUCTION 1 2. BURST ENCODING 2 2.0 Burst Basics 2 2.1 Encoding Using Burst Techniques 3 3. BURSTCOMM I 12 3.0 Burstcomm I Description 12 3.1 Burstcomm I Results 16 4. BURSTCOMM II 18 4.0 Burstcomm II Description 18 4.1 Audio and Sensor Circuits 20 4.2 Video Circuits 23 4.3 Optical Link Circuits 24 4.4 Burstcomm II Results 26 5. BIBURST 28 6. CONCLUSIONS 30 APPENDIX CIRCUIT AND TIMING DIAGRAMS 31 LIST OF REFERENCES 42 1. INTRODUCTION The fundamental idea of burst processing is to combine the advantages of stochastic processing (utter simplicity) with the advantages of weighted binary (higher precision with reasonable bandwidth) to obtain a system lying in the middle ground. The result is a system which allows continous processing of non-synchronized signals yielding low precision immediate results and higher precision final results through averaging. The simplicity of circuitry and minimization of synchronization requirements makes burst processing potentially useful in the area of communications. This paper then proposes to investigate various ways in which burst techniques can be used in communications and to give the design considerations and results of three demonstration systems. 2. BURST ENCODING 2.0 Burst Basics A basic building block in burst processing is the Block Sum Registe (see Figure 1). The voltage sum represents an average of the contents of the shift register. The BSR therefore can act as an integrator on the input bit stream. If each group on N bits represents a complete sample of an analog signal, the BSR does indeed function as a perfect integrator. Standard compacted bursts have this property and can be decoded with a simple BSR. Modifying the BSR allows various other encoding schemes to be used. With bidirectional BSRs, a form of delta modulation can be used. Companding becomes easily realizable in this case with weighted non-equal current sources. CLOCK CLEAR INPUT- N BIT SHIFT REGISTER N EQUAL CURRENT SOURCES i ■►OUTPUT VOLTAGE SUM Figure 1. BJock Sum Register 2. 1 Encoding Using Burst Techniques The burst encoder in the original proposal is the ramp encoder (Figure 2a) which consists of a BSR connected as a stairstep generator and a comparator. The comparator compares the stairstep to an analog input generating compacted bursts, i.e. bursts where all ones or all zeros are contiguous within a sample. The basic ramp encoder has the disadvantage of requiring a relatively high sample bit rate. Because the stairstep samples the input at a time dependent on its voltage level, the input must change less than one level of quantization during one stairstep 3 to prevent additional noise from being introduced. Specifically, a lVpp sine wave of frequency f with maximum slope around t = must vary less than one level of quantization 1/N (for N bit bursts and a 1 volt stairstep). Therefore for bit rate f (t = + l/2f ), N27rf -N2uf 1/N > .5 (sin-p-^- si n 2f c ) (D s s or where F is the normalized sampling frequency f s /f~ F s > Ntt/ARCSIN(1/N) (2) For large values of N x F $ equation 2 becomes F $ > N 2 tt (3) Note that all transmissions discussed in this thesis (with the exception of 2e - BIBURST) are non return-to-zero. This implementa- tion is different from transmissions in previous papers on Burst Process- ing which are all return-to-zero. Given external clocking, the non return- to-zero system would be used in real systems. With the addition of sample/hold circuitry (Figure 2b) the bit rate can be reduced to the level of the Nyquist sampling criterion, i.e. samples at twice the signal frequency, F s > 2 N (4) CLOCK CLEAR INPUT © COMPACTED BURSTS Figure 2a, Ramp Encoder CLOCK CLEAR INPUT © COMPACTED BURSTS Figure 2b. Improved Ramp Encoder CLOCK INPUT © UNCOMPACTED BURSTS Figure 2c Delta Block Encoder CLOCK SL - '0 MODE «.« ', SR - 1 i 1 1 s ■»'— BIDIRECTIONAL BSR [>- UNCOMPACTED INPU1 r n X BURSTS Figure 2d. Improved Delta Block Encoder SL CLOCK i SR BIDIRECTIONAL BSR 8 = 7 STEP INPUT© ii •o' 8 -1 + 8 + i - :=Q 3 LEVEL OUTPUT H - SR GND- HOLD L - SL Figure 2e. Biburst Encoder A different encoder using the same basic components as the ramp encoder is the delta block encoder (Figure 2c). The BSR tracks the input signal by clocking in ones or zeros depending on whether it is below or above the input. Again, as in the basic ramp encoder, the burst bit rate must be sufficiently high so that the input changes less than one level of quantization per sample. Since a new sample is effectively taken with each bit, it would seem that F s > tt/ARCSIN(1/N) (5) or for large N x F F s > ttN (6) is sufficient. Unfortunately, the BSR only changes its value if the input bit is the opposite of the final bit. If the bit falling off is the same as the bit coming in the BSR does not change. Since the input signifies a desired change, this error must be eliminated. The tracking of the input must be made independent of the history of the signal. Figure 2d using bidirectional shift registers is one solution. An increase of the BSR is guaranteed regardless of the state of the BSR (assuming of course the BSR is not already full of ones or zeros). Equa- tion 5 is now valid for a single frequency input. For a bandlimited input, obtaining a slope overload relation becomes more difficult. Extensive work has been done analyzing noise in conventional delta-modulation systems which the system in Figure 2d closely resembles. The major difference is that using a BSR as an integrator gives a maximum and a minimum output (and hence input if the BSR is to track the whole input) DC voltage. The theoretical analyses of conventional delta-mod have assumed no absolute maximum DC input voltage, instead they have dealt with an input of a certain rms power. A straightforward method of determining a slope overload criterion for the BSR system will be given after a short review of some useful results in the literature. Noise in delta-mod systems can be grouped into two categories; slope overload noise or noise caused by failure to track the input in the regions of maximum slope, and granular or quantization noise inherent in quantizing an analog signal. Since the input is not excursion voltage limited, any input slope is possible for a bandlimited signal and slope 5 overload results must be expressed in terms of probabilities. O'Neal and Protonotarious have obtained results for slope overload noise with good agreement with experimental results. Using curves they have calculated, one can determine an optimal step size given a sampling frequency. Below this step size slope overload noise predominates and above the step size granular noise increases. For a RC limited gaussian input and F = 64 for example, the optimum step size for a IV input is approximately .047. rr This corresponds to a 21 bit BSR in a BSR system even though the calcula- tions are not exactly correct for the DC maximum limited BSR system. To totally eliminate slope overload noise, a frequently used criterion is to set the maximum slew rate of the encoder to four times 7 8 the rms slope of the input. ' For gaussian signals, the probability that slope overload error can occur under these conditions is less than -5 9 4 x 10 . Expressing this relation for a IV system with step size rr 1/N and F s = f s /f c where f c is the cutoff of the gaussian input, F s > 4tt N//6 (7) 5 7 Granular noise has been examined rigorously * but a crude method exists which gives quite accurate results. The noise spectrum of granular noise is assumed to be white over a bandwidth equal to f s . Given that noise error can vary from -1/N to +1/N at any time with equal probability, total average noise power is equal to N , 1/N x 2 dx - 1/(3N 2 ). (8) 2 M/n The non-filterable portion of this (frequencies within the signal band- width) is clearly the total power x f c /f s - Therefore, for the non- filterable noise power N , N g = 1/(N 2 F S 2 3) (9) Since the maximum signal power is 1/8 for this IV system, for the maximum signal to noise ratio S/N, S/N = | (NF S ) 2 (10) A straightforward slope overload criterion is possible for the BSR integrator system because the input must be maximum amplitude limited. The maximum slope of a bandlimited input is dependent upon the maximum slope possible at the output of the bandlimiting filter with some finite amplitude input. In the case of cascaded RC sections, the maximum possible slope occurs with a step input. Specifically, for step size 1/N and normalized sampling frequency F = f_/f r with 2nf being each RC pole, 1 . e" 27Tf s - 2TTF s e _2TTF s - (2ttF s ) 2 \r e" 2 ^ ... < I (11) or, where m is the number of identical cascaded RCs, 1 - e" 27TF s I (2 F s )'/;: < A (12) i=0 IN A more meaningful F s for higher order filters is the ratio of f s to the -3dB point of the overall filter. For m cascaded RCs with an overall half power point of f c ' , (1 + (f c 7f c ) 2 ) m/2 = 2l/2 (13) or f c ' = f e ( e ln(2)/m - 1) 1/2 (14) Equation 12 can be used substituting F s ' for F s where F s ' = F s x (e n ^ ' /m - 1) ' . Figure 3 shows the curves for m = 1 and m = 2 with F s ' instead of F s . The value of N is the maximum possible where the only source of noise is granular noise. Figure 2e is a modification of Figure 2d involving an additional transmission level yielding synchronization and bandwidth advantages. It shall be referred to as Biburst - uncompacted bursts with bidirectional (plus, ground, minus) levels of transmission. A Biburst system transmits a bit only when the input differs from the BSR value by more than half a step. The receiver therefore does not need any clock at all unlike the previous systems. It derives its clock from the edge of the input pulses and if nothing comes in, it holds its current value. The average bit rate requirement is obviously reduced for Biburst systems. For a IV signal with cutoff frequency f c , the average r r absolute value slope is f c . With step size 1/N, the system need transmit only N bits per cycle. A normal delta mod system would transmit F s = fs/fc samples per cycle. Therefore, a Biburst system uses only N/F s x 100 percent of the bits of a standard delta mod system under these conditions. For example, with F s = 64 and N = 10 (satisfying Figure 3 first order RC), The Biburst bit rate is 15.6 percent of the maximum. This savings figure refers to the average bit rate, not the bandwidth or maximum bit rate of a channel. Multiplexing multiple Biburst systems over a channel to take advantage of this savings is a possibility. 10 90"! V \ > \ Or °\ c p 1 \ r+ c* o o o en o CO o O (X) o ^ ii (A o ^ o o C\J o X CO Q) S- =3 en O O o o CM o o o O co o. w A \ o in s i\ ' t II Vt 111 -J K 1« °< s< 2 K o V) o x E° > Crt — m 3 O V) O z > < UJ 1/) o 2 I i I 1 1 I UJ X UJ _J -1 2 id o o o i- CQ 60 for first order RC filtered inputs. Since the filter used (see AD-5) is greater than first order and F s = 50 MHz/ (10 x 3.2KHz) = 1563, the audio samples are clearly without slope overload. Since the audio is sampled at 15.75 KHz, the Nyquist criterion dictates a cutoff of 7.87 KHz before aliasing occurs. To transmit a voice grade audio signal a bandwidth of 3.2 KHz is sufficient. To reduce quantization error it is desireable to encode higher frequencies as well as providing that they can be filtered out and that they will not be folded into the desired frequency range. The cutoff frequency then for the encoder filter is 15.75 - 3.2 = 12.55 KHz. Aliasing which occurs from 3.2 KHz to 12.55 KHz does not matter since those frequencies will be filtered out at the decoder end with a 3.2 KHz low pass. Figure 12 illustrates this. Since real filters do not have the abrupt cutoff characteristics of ideal ones and components for filters come in certain convenient values, the actual filters constructed were a 5.-5 KHz LPF in the encoder and a 3.2 KHz LPF in the decoder. Figures AD-5 and AD-6, pp. 36-37, circuits. The encoder LPF is acutally one-third order constant-k ir 12 section cascaded with two RC sections. The "faithless constant-k" was used because of its simplicity and the fact that the filter characteristics 21 3.2 KHz LPF 12.55 KHz f $ = 15.75 KHz ALIASING SAMPLED 15.75 KHz i LPF 3.2 KHz VOICE 1 to f s /2 '■ 12.55 KHz LPF 15.75 KHz SAMPLER 3.2 KHz LPF ► OUTPUT DEAL 5.5 KHz Ih^ ACTUAL 3.2 KHz 1-, INPUT 15.75 KHz SAMPLER 3.2 KHz LPF OUTPUT Figure 12. Audio Filtering 22 were not critical. With the two RCs , ripple in the passband is less than 1 dB and attenuation at the critical 12.55 KHz frequency is greater than 20 dB. The decoder LPF consists of a 3.2 KHz third order constant-k low pass cascaded with a low Q 7.9 KHz trap. The low Q trap flattens out the constant-k response as well as making sure that granular noise at f s /2 = 7.9 KHz is down more than 20 dB. The encoder circuit is also a pre-amp which amplifies an input from a high Z microphone to IV levels for the comparator. The decoder includes an audio amp which runs off of 5.2 V and provides a 3V output across an 8 ohm speaker (1/4 watt RMS). The decoder consists of a counter/D-A converter instead of a BSR to reduce the package count. Due to the fact the audio bursts arrive as separate burst samples there is no need for the "window" characteristic of a BSR. The sensor decoder is also a counter/D-A pair. The D-As consist of PNP transistors wired up just like a BSR except that the emitter resistors are weighted according to the binary weight of each counter output. Due to the range of currents (8:1 for a four bit counter) the resistor values had to be adjusted for varying base emitter voltage drops. The final D-A is accurate to within 10 percent of the least significant bit. 23 4.2 Video Circuits Design Considerations The video circuitry (see Figure AD-7, p. 38) consists of filter- buffers at the encoder and decoder. Again the third order constant-k filter is used, this time with a cutoff of 2.4 MHz. This is near the upper frequency possible with effective samples at 5 MHz from the Nyquist criterion. At this point however, there is significant slope overload error of the time/voltage type of the non-sample/hold ramp encoder. Future improvements could include the addition of sample/hold circuitry to eliminate this error. Nevertheless, Burstcomm I showed a useable visual result is possible even with this error. The encoder is adjusted to not encode any of the sync portion of the composite video signal since the sync is already being transmitted as negative going pulses. Specifically, two 50 MBit negative going pulses in a row signify HSync and one followed by nothing signifies VSync. Sync circuitry (also Figure AD-7) included two Schmidt trigger inputs to get clean pulses at ECL logic levels and two monostables at the output. As soon as incoming negative pulses are decoded into vertical or horizontal sync information, the correct monostable is triggered. 24 4.3 Optical Link Design Considerations To achieve picture quality similar to Burstcomm I, the optical link needs to transmit bits at a rate of 50 Mbits/sec. This dictates a minimum pulse period of 20 ns. In addition, three levels of intensity have to be sent. The following discussion will refer to Figure AD-8 on p. 40 which contains a schematic of the optical link. Bifurcated Dolan-Jenner E624 glass fiber bundles are used for a number of reasons; a large diameter (.044") at the transmitting end with consequent efficient coupling in of light, ease of mounting, two LEDs transmitting to one receiver for more total optical power, and most importantly--the fact that a few surplus E624s were floating around the lab. A figure of 45 percent transmittance is given for the 2 1 fiber bundle at around 650 nm. Dial co 521-9195 red LEDs were chosen for their relatively high power (for a visible LED), narrow beam (half power into half angle 35o), fast speed of response (1 ns rise and fall time in spec, sheet), and low capacitance (20 pF at V = 0). They are driven with current sources to maximize speed of response. An MPS 3640 PNP is used as each current source at the output of an ECL gate and provides a current swing of 10-40 ma through the LED. Measured optical power output varies from roughly 9 to 31 ywatts. The receiving end presented many more design difficulties. An initial design using cascaded cascode stages proved too unstable and also lacked sufficient speed. The final simple photodiode-amp 13 combination was derived from a circuit used by a group at GE. The amplifier consists of multiple 1/3 10116 line receivers operating between ground and -5.2V with 10K resistors for negative feedback stability. Only two of the three comparators on the first stage IC are used as the 25 use of all three cascaded leads to high frequency oscillations. The amp operates from DC to around 80 MHz with a measured rise and fall time of 6 ns and a gain of 4.7 volts/yamp small signal. The photodiode used is an HP4207 PIN photodiode. Advantages of a PIN over an avalance photodiode include decreased sensitivity to power supply noise, increased temperature stability, and lower power supply voltages. The major disadvantage is less output current for a 14 given light intensity. Since Burstcomm II was intended as a demonstra- tion with short fiber lengths, the PIN photodiode was chosen. Calculated p current responsivity at 640 nm was 3 yamp/mwatt/cm . Total optical power output swing of the LED is 22 ywatts. For .1" spacing between emitter and fiber and .022" fiber radius, the solid half angle is 12.50. Using the Dialco spec, sheet, roughly 15 per- cent of the output power goes into the fiber (ignoring coupling losses from reflection). Flux at the output of the fiber (radius .0625") then is 22 x .15/((.0625/2) 2 x it x 2.54 2 ) ywatt/cm 2 = .167 mwatt/cm 2 . The output of the photodiode (again ignoring losses between fiber and photodiode) is 3 x .167 = .5 yamp. The amp output then is 4.7 x .5 = 2.35V . Since the output is clipped at ECL logic levels (-9 V DD ) these calculations indicate that the optical link should work. 26 4.4 Burstcomm II Results Figure 13 shows a test pattern transmitted through Burstcomm II. The effects of quantization are clearly visible but the result is definitely useable. Sync was transmitted with no problems. Figure 13. Burstcomm II Video Audio presented some problems initially with the sensitive mic preamp picking up radio stations. An RF bypass capacitor on the input eliminated the problem. The amplifier on the demultiplexer causes inter- ference with the video when set to maximum volume. This problem would be eliminated by increased power supply filtering or by separating the amplifier from the demultiplexer. Figure 14 shows what a transmitted 1 KHz sine wave (unfiltered) looks like. The top trace is the input, the middle trace is the unfiltered output, and the bottom trace is the difference between the two. 27 ^•///'-^v \v Figure 14. Burstcomm II Transmission 1 KHz Sine Wave The overall audio frequency response is shown in Figure 15. It corresponds well with what was predicted from the design. The actual -3 dB points are approximately 400 Hz and 3 KHz. -4 -8 o -io -12 -14 -16 — •— - »s \ \ L \ AUDIO FREQUENCY RESPONSE \ \ \ \ \ .2 .4 .6 .8 1 2 FREQUENCY IN KHz 4 5 Figure 15. Burstcomm II Audio Response 28 5. BIBURST A demonstration of the Biburst encoder (see Figure 2e) was built and tested in the summer of 1975. It was designed to encode/ decode audio frequency signals and transmit them with 3 levels of trans- mission (positive and negative going pulses). Circuit diagrams for the encoder are in Figure AD-9. The decoder consists simply of the bidirectional shift register with a clock derived from the delayed edge of the input pulses and mode controlled by polarity of the input pulse. Note that to have no clock on the decoder, the bits must of course be transmitted in return to zero notation. Visual results (on an oscilloscope for the encoding of a single frequency) were quite positive for non-filtered output due to the nature of Biburst error. The encoded signal appears only to lag an input by a phase difference dependent on the slope of the input unlike conventional delta-mod which, in regions well away from slope overload, jumps around on both sides of the input. Filtered errors should be similar for the two systems. Figures 16 and 17 give an idea of what Biburst encoding looks like for two values of F . Even for low F , degradation of encoder performance is controlled and gradual and a useable result (even though it is attenuated) is achieved. The top trace is the input, the middle trace is the Biburst non-filtered output, and the bottom trace is the difference between the two. 29 Figure 16. F = 100 Noise Amplitude Doubled Figure 17. F = 20 Noise Amplitude Normal 30 6. CONCLUSIONS The demonstrations of Burstcomm I, Burstcomm II, and Biburst clearly show the possibility of using burst techniques in the transmission of digitally encoded analog signals. Depending on the specific applica- tion, a version of encoding using burst techniques may prove to be otpimal by taking advantage of circuit simplicity, minimal synchroniza- tion, easy intrinsic companding, or multiplexing possibilities. Ramp encoding with sample/hold circuitry on the input is generally the best burst system in terms of bandwidth and could be optimal if the added complexity of the sample/hold circuitry is permissible. Against standard PCM it offers synchronization advantages versus higher bandwidth requirements. The absolute simplest system is the original ramp encoder but it requires high bandwidth if its form of slope overload error is to be avoided. If bandwidth is sufficient or the error can be tolerated to some extent, it does provide the cheapest encoding method. If companding is useful (due to amplitude distribution characteris- tics of a particular input), a bidirectional delta-mod like system could be optimal. It offers minimum sample bit rates for certain inputs (see Figure 3), those with strongly limited maximum slopes. The major dis- advantage is that bidirectional shift registers versus sample/hold circuitry for the s/h ramp encoder are needed. For ultimate lack of synchronization (no decoder clock) and reduced average sample bdt rate, Biburst stands out. Multiplexing possibilities, no sync, and easy companding must be weighed against multiple levels of transmission and increased circuit complexity. 31 APPENDIX CIRCUIT AND TIMING DIAGRAMS 32 - • UJ Ui 4 u o -8 o So-r-J IS o B -».*, T7 o to A Z UJ V) UJ o o- JK IO rw IO is- Dn »-3>J UJ > w *2 *> - uj »- CO o ? Q. ■■ >- M >- 2 33 ii u> ID o> to II CM I O - >- >- >- -J co CO CO (7) o X X > > CM 3 O O Q GO < z UJ ceuj coco z< UJZ COQJ UJ O-l UJ CD Q< >5 UJ CO UJ (Z 34 a; X 01 0) Q o o ■M - 3 GO CO I - < s- 3 35 1 r a> m s- --j >--J lo WID (03 oo 10. >Q- UJO Q UJ UJ UJ ■^uj O* ;=:o o r UJCD Q< aruj z< UIZ (/)UJ 36 •«- > m 0) U3 —I I I t-H CM E s i- Q. > N in I 3 O > I CO o_ s_ CD -o s- o Q- ro -o S- — h- ro N N ni nz ^ ^ O* CM • • r-*» CO «3- Lf) o. E S- ■•-> -a 3 i. CD -o o o 0) Q a <: a s- en (0 00 38 VIDEO INPUT 10/i H _J20 ■7 r— II — "^ — Ufi 100 ■> 820 > I f ... n- 2K {%PS 6318 TO COMPARATOR -5.2V -5.2V Multiplexer Video Filter/Buffer BSR CURRENT SUM 10/t H =±=820 =±=820 £75 I 4 TO MONITOR PS 6518 -5.2V Demultiplexer Video Filter/Buffer H OR V SYNC INPUT ECL LEVEL SYNC -5.2V Multiplexer Sync Shaper Figure AD-7. Video and Sync Circuitry 39 o c/) K 3 r+ w ■VwV m V) a. X£ > ■-I o ff) m 10 z M ro i-i ►- m 10 o. O > (VI in I 3 U t- •r* O o e >> 00 ai x (U u - V) ( 1 - -J L-J Id *1 f •» LJ m q. 4.0) 1 1 O -J O 3 M a. O o f- N * «■ : _ AAA ^ 5' - A A* ^ > VVV-^-pj in in 1 O O N N m * 00 «■ O IO O IO o 10 o » x o ► * a: to Z> C Q. O o CM u CVJ O Z z >- c o o o 2 a> 40 s A i!« o Q. O CO I o Q) i- 3 41 s- ■o o o S- -Q •r- CO • en Q s- en O K 3 2 < ± 42 LIST OF REFERENCES Poppelbaum, W. J., Appendix I to "A Practicability Program in Stochastic Processing," Department of Computer Science, University of Illinois, March, 1974. Mohan, P. L. , "The Application of Burst Processing to Digital FM Receivers," Report UIUCDCS-R-76-780, Department of Computer Science, University of Illinois, January, 1976. 3 Pleva, R. M. , "A Microprocessor-Controlled Interface for Burst Pro- cessing," Report UIUCDCS-R-76-812, Department of Computer Science, University of Illinois, July, 1976. 4 Taylor, G. L. , "An Analysis of Burst Encoding Methods," Report UIUCDCS-R-75-770, Department of Computer Science, University of Illinois, December, 1975. 5 0'Neal, J. B. Jr., "Delta Mod Quantization Noise," The Bell System Technical Journal 45 No. 1 (January 1966), pp. 117-141. Protonotarious, E. N. , "Slope Overload Noise in DPCM Systems," The Bell System Technical Journal 46 No. 9 (November 1967), pp. 2119-2161. Van De Weg, H. , "Quantization Noise of Delta Modulation," Phillips Research Reports 8 No. 5 (October, 1953), pp. 367-385. o Bennett, W. R. , "Spectra of Quantized Signals," The Bell System Technical Journal 27 No. 3 (July 1948), pp. 446-472. g Goodman, D. J., "Delta Modulation," The Bell System Technical Journal 48 No. 5 (May, 1969), pp. 1197-1218. Taub, H. , Principles of Communication Systems , McGraw-Hill Book Co., New York, 1971. Faiman, M. , ed. , "IEL Quarterly Technical Progress Report," Department of Computer Science, University of Illinois, May-June 1975. 12 Lubkin, J. Y. , Filter Systems and Design , Addi son-Wesley, Reading, Mass. , 1970. 13 Schmid, H. , "Fiber-Optic Data Transmission," Electronics 49 No. 18 (September 2, 1976), pp. 94-99. Compton, R. D. , ed. , "Fiber Op No. 6 (June, 1975), pp. 14-24. Compton, R. D. , ed. , "Fiber Optics," Electro-Optical System Design 7 SECURITY CLASSIFICATION OF THIS PAGE (Whit Data Bntarad) REPORT DOCUMENTATION PAGE READ INSTRUCTIONS BEFORE COMPLETING FORM I REPORT NUMBER UIUCDCS-R-77-838 2. GOVT ACCESSION NO 3. RECIPIENT'S CATALOG NUMBER I. TITLE (and Subtitle) TRANSMISSION OF ANALOG SIGNALS USING BURST TECHNIQUES S. TYPE OF REPORT * PERIOD COVERED M.S. Thesis/January 1977 *. PERFORMING ORG. REPORT NUMBER UIUCDCS-R-77-838 7. author^; Martin Wolff •. contract or grant number*-*; N00014-75-C-0982 9 PERFORMING ORGANIZATION NAME AND ADDRESS Department of Computer Science University of Illinois at Urbana-Champaign Urbana, Illinois 61081 10. PROGRAM ELEMENT. PROJECT, TASK AREA « WORK UNIT NUMBERS II. CONTROLLING OFFICE NAME AND ADDRESS Office of Naval Research Code 437 Arlington, VA 22217 12. REPORT DATE January 1977 19. NUMBER OF PAGES AS- U MONITORING AGENCY NAME 4 ADDRESS?!/ dlltatant /root Controlling Oifiem) IS. SECURITY CLASS, (ol thla rmpori) Release Unlimited 15a. DECLASSIFICATION/ DOWNGRADING SCHEDULE 16. DISTRIBUTION STATEMENT (ol this Report; DISTRIBUTION UNLIMITED 17. DISTRIBUTION STATEMENT (ol Ota abttract anlarad In Block 20, II dlllmrmnt from Raport) 18. SUPPLEMENTARY NOTES 19. KEY WORDS fConllnut on ravaraa alda II nacaaaaty and Idantlty by block numbar) Burst Processing, Burstcomm, Biburst, Block Sum Register, Delta Modulation, Slope Overload 20. ABSTRACT (Contlnua on ravaraa aid* II nacaaaary and Identity by block numbar) Burst techniques are considered as possible alternatives to conventional ones in the realm of digital transmission of analog signals. Three demon- stration systems using various, Burst techniques are discussed in detail. DD 1 jan 73 1473 EDITION OF 1 NOV SS IS OBSOLETE S/N 0102-014-*601 | Release Unlimited SECURITY CLASSIFICATION OF THIS PAGE (Whan Data Kniarad) BLIOGRAPHIC DATA EET 1. Report No. UIUCDCS-R-77-838 3. Recipient's Accession No. Title and Subtitle TRANSMISSION OF ANALOG SIGNALS USING BURST TECHNIQUES 5. Report Date January 1977 Auth ^RTIN WOLFF 8. Performing Organization Rept. No UIUCDCS-R-77-838 Performing Organization Name and Address Department of Computer Science University of Illinois at Urbana-Champaign Urbana, Illinois 61081 10. Project/Task/Work Unit No. 11. Contract/Grant No. N00014-75-C-0982 Sponsoring Organization Name and Address Office of Naval Research Code 437 Arlington, VA 22217 13. Type of Report & Period Covered NLS. Thesis 14. Supplementary Notes Abstracts Burst techniques are considered as possible alternatives to conventional ones in the realm of digital transmission of analog signals. Three demon- stration systems using various Burst techniques are discussed in detail. Key Words and Document Analysis. 17a. Descriptors Burst Processing Eurstcomm Biburst Block Sum Register Delta Modulation Slope Overload 7 Identifiers /Open-Ended Te 7 COSAT1 Field/Group 8 variability Statement Release Unlimited 19. Security Class (This Report) UNCLASSIFIED 20. Security Class (This Page UNCLASSIFIED 21. No. of Pages 48 22. Price ' NTIS-35 (10-70) USCOMM-DC 40329-P71 .MA8 iwf H 2 1879 oUNDfi? m