doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 1 of 58 Project: IEEE P Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [The ParthusCeva Ultra Wideband PHY proposal] Date Submitted: [05 May, 2003] Source: [Michael Mc Laughlin, Vincent Ashe] Company [ParthusCeva Inc.] Address [32-34 Harcourt Street, Dublin 2, Ireland.] Voice:[ ], FAX: [-], Re: [IEEE P Alternate PHY Call For Proposals. 17 Jan 2003] Abstract:[Proposal for a a PHY] Purpose:[To allow the Task Group to evaluate the PHY proposed] Notice:This document has been prepared to assist the IEEE P It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release:The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 2 of 58 The ParthusCeva PHY Proposal

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 3 of 58 Overview of Presentation PHY packet contents Coding –DSSS Coding scheme - biorthogonal coding –Ternary spreading codes –FEC scheme - rate 4/6, 16 state convolutional coding Preamble I mplementation Overview Performance –Link margin –Test results –Data Throughput Complexity

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 4 of 58 Packet Contents

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 5 of 58 The coding scheme 64 biorthogonal signals [Proakis1] 64 signals from 32 orthogonal sequences Ternary sequences chosen for their auto-correlation properties Code constructed from binary Golay-Hadamard sequences

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 6 of 58 Creating Orthogonal Ternary Sequences Take a matrix of binary orthogonal sequences, in our case we used Golay-Hadamard sequences Add any two rows to get a ternary sequence Sum of any other two rows is orthogonal to this Continue till all the rows are used Repeat but subtracting instead of adding

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 7 of 58 Finding good Ternary Golay Hadarmard codes Large superset of orthogonal sequence sets to test Define aperiodic autocorrelation merit factor (aamf) as the ratio of the peak power of the autocorrelation function to the mean power of the offpeak values divided by the length of the code. Random walk used to find set with best aamf

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 8 of 58 Code comparison Length 32 code chosen for aamf and best matching with bit rates.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 9 of 58 Sample rate and pulse repetition frequency Signal bandwidth chosen is 3.8GHz to 7.7GHz Sampling rate chosen is 7.7Ghz 32 chips per codeword, 4 bits / symbol (6 bits less 2 for convolutional code)

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 10 of 58 FEC scheme A rate (rate 4/6) convolutional code was chosen for the FEC. [Proakis2] Very low complexity 16 state code, constraint length 2, Octal generators 27, 75, 72. Each of 16 states can transition to any other state, outputting 16 of 64 possible codewords. Provides 3dB of gain over uncoded errors at a cost of 50% higher bit rate

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 11 of 58 Rate 4/6 Convolutional coder Map every 6 bits to one of 64 biorthogonal codewords bits in +

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 12 of 58 Preamble Sequence

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 13 of 58 PAC properties Because of the perfect autocorrelation property, the channel impulse response can be obtained in the receiver by correlating with the sequence and averaging the results. Because the sequence consists of mostly 1, -1 with a small number of zeros, correlation can be economically implemented. (a length 553 PAC has 24 0’s)

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 14 of 58 Preamble properties Very good detect rate and false alarm probability. Pfa and Pmd < for CM1 to CM4 test suite at 10 metres. Different length sequences means other piconets won’t trigger detection i.e. Pfa still < for a different piconets PAC n, even at 0.3m separation. Preamble length varies from ~5  s to ~15  s depending on the bit rate. Lower bitrates use longer preambles (Longer distances need more training time)

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 15 of 58 PHY Header The PHY header is sent at an uncoded 45Mbps rate, but with no convolutional coding. It is repeated 3 times. The PHY header contents are the same as i.e. Two octets with the Data rate, number of payload bits and scrambler seed.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 16 of 58 Scrambler/Descrambler It is proposed that the PHY uses the same scrambler and descrambler as used by IEEE

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 17 of 58 Typical Tx/Rx configuration Channel Matched filter (Rake Receiver) A/D 7.7GHz, 1 bit Correlator Bank Viterbi Decoder Output data at Mbps Antenna Convolutional encoder 8-240M symbols/sec Code GeneratorChip to Pulse Generator Input data at Mbps Mchips/sec Descramble Scramble Fine/ Band Reject Filter LNA Switch / Hybrid Band Pass Filter* Band Pass Filter Band Reject Filter * Can be avoided with good LNA dynamic range Single Chip Possible

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 18 of 58 Band reject Filter ** Possible RF front end configuration Total Noise Figure = 7.0dB Tx/Rx switch / hybrid BP Filter* From Tx Fine Filter** Filter NF= 2.0dB NF= 0.8dB NF= 4.0dB NF= 0.2dB (input referred) To Rx * Can be avoided with good LNA dynamic range LNA ** Depending on Local National or User requirements

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 19 of 58 Matched Filter configuration CnCn DiDi C n+N D i-N 41 4x C n+1 D i-1 C n+N+1 D i-N x 44 4 bit adder 5 bit adder …..

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 20 of 58 Matched Filter configuration Structure repeated 16 times e.g. a 500 tap filter with 4 bit coefficients would have 500 x 16 x 4 AND gates in first stage Calculates 16 outputs in parallel, each runs at (480/mps) MHz. –e.g. 120MHz for 240Mbps Multiplier is 4 AND gates. First adder stage is 4 OR gates. Very little performance loss. (0dB for CM1-3, 0.23dB for CM4). Coefficients are pre-processed to remove smallest if two clash. mps is max pulses/sample. = 960/(bit rate (Mbps))

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 21 of 58 Matched filter 560 tap filter takes 135k gates or 0.82 sq mm in 0.13  standard cell CMOS Power consumption = 120mW ( at 480Mbps ), proportional to data rate. Matched filter re-used for correlation with training sequence during training phase All simulations were carried out with this filter/correlator structure

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 22 of 58

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 23 of 58 Distance achieved for mean packet error rate of best 90% = 8%

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 24 of 58 Distance achieved for at worst packet error rate of best 90% = 8%

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 25 of 58 Mean distance for an 8% PER

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 26 of Mbps average PER

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 27 of Mbps average PER

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 28 of Mbps average PER

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 29 of 58 Adjacent Channel Interferers: Single uncoordinated piconet Tests were done with reference links using channel models 1-4, channels 1-5 with shadowing removed. Interferers used channels 6-10 of Channel models 1 to 4. To allow some error margin, the distances to the reference receivers were 5m, 2m and 1.5m. For each channel model, at each distance, the mean PER for all 100 tests was calculated. (5 x 4 x 5)

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 30 of Mbps with single adjacent interferer Single uncoordinated piconet, Reference Link 120Mbps at 5m, cm1-4 Interferer Distance (m) log 1 0 average PER 8% PER channel model 1 channel model 2 channel model 3 channel model 4

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 31 of Mbps with single adjacent interferer Single uncoordinated piconet, Reference Link 240Mbps at 2m, cm1-4 Interferer Distance (m) log10 Average PER 8% PER channel model 1 channel model 2 channel model 3 channel model 4

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 32 of Mbps with single adjacent interferer

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 33 of 58 Two adjacent channel interferers Tests were done with reference links using channel models 1-4, channels 1-5 with shadowing removed. Adjacent channel interferers used a freespace channel, but filtered by front end bandpass filter. To allow some error margin, the distances to the reference receivers were 5m, 2m and 1.5m. For each channel model, at each distance, the mean PER over the 5 channels is plotted. Test with CM1-4 as interferer showed similar results

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 34 of Mbps - Two free space interferers

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 35 of Mbps - Two free space interferers

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 36 of Mbps - Two free space interferers

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 37 of Mbps - Three free space interferers

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 38 of 58 More interferers What matters is the total interference power, very little effect due to delay spread of the interfering channels. 2 interferers have 3dB more power than 1 which translates to 50% worse distance performance. 3 interferers have 1.76dB more power than 2 which translates to 22% worse distance performance. 4 interferers have 1.76dB more power than 3 which translates to 15% worse distance performance.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 39 of 58 Co-channel interference Different piconets use exactly the same data mode codes as each other. Separation is achieved because –a) a different piconet will have a different impulse response and thus will not correlate with the matched filter which has been trained for the piconet of interest. –b) Codes won’t be synchronised Co-channel data mode interference is exactly the same as adjacent channel interference. Training to the preamble will be affected more markedly by co-channel interference. Difficult to simulate.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 40 of 58 Co existence Out of band signals ( 10.6GHz) are always filtered out. Any desired in band energy can be filtered out, with minimal effect on performance because the whole band is used to transfer data. Only adverse effect is the transmit power reduction (e.g. Dropping 400MHz for a loses <0.5dB)

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 41 of 58 Interference and susceptibility As for co existence, out of band signal are always filtered out. Again, any desired in band energy can be filtered out, with minimal effect on performance because the whole band is used to transfer data. Only adverse effect is the receive power reduction (e.g. Dropping 400MHz for a loses <0.5dB), its just a part of the channel.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 42 of 58 PHY-SAP Data Throughput At higher bit rates, a 1024 byte frame is very short. The channel will be stationery for more than one frame so it is possible to send multiple frames for each preamble.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 43 of 58 Scalable solution Possible to improve distance achieved by increasing coder complexity. (Decoder used here <20k gates) Mbps. Gate count depends on maximum bit rate and power consumption of baseband PHY is proportional to bit rate Short range solution possible with much smaller matched filter gate count/consumption for smaller delay spread. e.g. if poorer performance acceptable on CM4 type channels

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 44 of 58 Complexity - Area/Gate count, Power consumption These figure are for a standard cell library implementation in 0.13µm CMOS

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 45 of 58 Self evaluation : General Criteria

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 46 of 58 Self evaluation : PHY protocol

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 47 of 58 Self evaluation : MAC enhancements

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 48 of 58 Summary of advantages Ternary spreading codes –Better auto-correlation properties Perfect PAC training sequence Simple RF section –1 bit A/D converter –No AGC required –No mixers required Long rake possible - near multipath immunity –4 bit coefficients –1 bit data –no multipliers Cost and Power very similar to Bluetooth

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 49 of 58 Backup Slides

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 50 of 58 Ternary orthogonal sequences From any base set of 32 orthogonal binary signals, can generate 32 C 16 sets of 32 orthogonal ternary sequences. Generate by adding and subtracting any 16 pairs. Generally, if the base set has good correlation properties, so will a generated set.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 51 of 58 Good base binary set Base set is a set of binary Golay-Hadamard sequences Take a binary Golay complementary pair. s1 16 =[ ]; s2 16 =[ ]; if A=circulant(s1 16 ) and B=circulant(s2 16 ) and G32= A B B T -A T then G32 is a Hadamard matrix. [Seberry] This type has particularly good correlation properties[Seberry] Detector can use the Fast Hadamard Transform

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 52 of 58 Creating Orthogonal Ternary Sequences Take a matrix of binary orthogonal sequences Add any two rows to get a ternary sequence Sum of any other two rows is orthogonal to this Continue till all the rows are used Repeat but subtracting instead of adding

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 53 of 58 Orthogonal Ternary Example E.g pairing 1 with 3 and 2 with 4 gives this orthogonal matrix

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 54 of 58 Finding good Ternary Golay Hadarmard codes Large superset of orthogonal sequence sets to test Define aperiodic autocorrelation merit factor (aamf) as the ratio of the peak power of the autocorrelation function to the mean power of the offpeak values divided by the length of the code. Random walk used to find set with best aamf

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 55 of 58 Code comparison Length 32 code chosen for aamf and best matching with bit rates.

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 56 of 58 Ternary Orthogonal Length 32 Code Set

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 57 of 58 Matlab Code to generate PAC sequences % function phi=ipatov(nu,multiplier,mul2); % % Generate a length nu, ternary perfect periodic autocorrelation sequence % using Singer Cyclic Difference Sets e.g. (553, 24, 1) % function phi=ipatov(nu,multiplier,mul2); if nargin==1 multiplier=1; mul2=-1; end % multipliers 1,-1 are most commonly good if nargin==2 mul2=-1; end % multipliers 1,-1 are most commonly good phi=0; if gcd(nu,multiplier)>1; % must not be a common divisor of nu return; end if gcd(nu,mul2)>1; % must not be a common divisor of nu return; end switch nu case 7 % nu=7;k=3;lamda=1; D=[1 2 4 ]; case 13 nu=13;k=4;lamda=1; D=1+[ ]; case 21 nu=21;k=5;lamda=1; D=[3,6,7,12,14]; case 31 nu=31;k=6;lamda=1; D=[ ]; case 57 nu=57;k=8;lamda=1; D=[ ]; case 63 % multipliers 1,5 gives perfect ternary nu=63;k=31;lamda=15; D=1+[ ]; case 73 nu=73;k=9;lamda=1; D=[1, 2, 4, 8, 16, 32, 37, 55, 64]; % P73 case 91 nu=91;k=10;lamda=1; D= [ ]; % case 133 nu=133;k=12;lamda=1; D=[ ]; case 183 nu=183;k=14;lamda=1; D=[ ]; case 273 nu=272;k=17;lamda=1; D=[ ]; case 307 nu=307;k=18;lamda=1; D=[ ]; case 341 nu=341;k=85;lamda=21; % 1,5 gives a perfect ternary sequence D= [ ]; case 364 nu=364;k=121;lamda=40; % 1,5 gives a perfect ternary sequence D=[ ]; case 381 nu=381;k=20;lamda=1; D=1+[ ]; case 511 % 1,3 gives a perfect ternary sequence nu=511;k=255;lamda=127; D=[ ]; case 553 nu=553;k=24;lamda=1; D=[ ]; case 651 nu=651;k=26;lamda=1; D=[ ]; case 757 nu=757;k=28;lamda=1; D=[ ]; case 781 nu=781;k=156;lamda=31; % 1,2 gives a perfect ternary sequence D=1+[ ]; case 871 nu=871;k=30;lamda=1; D=[ ]; case 993 nu=993;k=32;lamda=1; D=1+[ ]; case 1057 nu=1057;k=33;lamda=1; D=[ ]; case 1407 nu=1407;k=38;lamda=1; D=1+[ ]; case 1723 nu=1723;k=42;lamda=1; D=[ ]; otherwise return end % end switch D=sort(mod(D*multiplier,nu)); % transform to new difference set. while any(D==0) D=mod(D+1,nu); end Dhat=sort(mod(D*mul2,nu)); % transform to another new difference set while any(Dhat==0) Dhat=mod(Dhat+1,nu); end ; Xd=zeros(1,nu); Xdhat=Xd; Xd(D)=1; Xdhat(Dhat)=1; phi=xcorr([Xd Xd],[Xdhat])-lamda; phi=round(phi(2*nu+1:2*nu+nu)); % phi is the ternary sequence if nargout==0 plot(xcorr(phi,[phi phi phi])) end

doc.: IEEE /123r3 PHY proposal May-2003 Michael Mc Laughlin, ParthusCevaSlide 58 of 58 References [Proakis1] John G. Proakis, Digital Communications 2nd edition. McGraw Hill. pp [Proakis2] John G. Proakis, Digital Communications 2nd edition. McGraw Hill. pp [Seberry et al] J. Seberry, B.J. Wysocki and T.A. Wysocki, Golay Sequences for DS CDMA Applications, University of Wollongong [Ipatov] V. P. Ipatov, “Ternary sequences with ideal autocorrelation properties” Radio Eng. Electron. Phys., vol. 24, pp , Oct [ H ø holdt et al] Tom Høholdt and Jørn Justesen, “Ternary sequences with Perfect Periodic Autocorrelation”, IEEE Transactions on information theory.