Scrambler Choices to Meet Emission Requirement for 1000BASE-T1 Kanata, ON, Canada September 08, 2014 Ahmad Chini achini@broadcom.com Version 1.0 IEEE 802.3bp Task Force Sept 8, 2014 Page 1
Contributors Mehmet Tazebay Mike Tu Peiqing Wang Page 2
Outline Peak emission, measured on100khz bandwidth is bounded by CISPR25 testing requirements and industry set levels. While scramblers are used to reduce peak emission, it was shown in chini_3bp_01_0714 that scramblers with higher order polynomials produce higher peak emission than some lower order ones. N-bit scramblers allow any sequence of N bits including some correlated patterns. Longer runs of correlated signal generate larger peak emission. Similarly, very short scramblers have correlation due faster repetition. A good scrambler choice should avoid both extremes. Scrambler size and mapping are optimized for minimizing peak emission given 2D-PAM3 modulation for 1000BASE-T1. The Master/Slave scramblers are verified to train DSP both in HDX and FDX operations. Page 3
TX PSD Peak Guide Line Derivation TX PSD Peak Guide Line (dbm/hz) = Peak Emission Limit (dbm/hz) Emission Transfer Function (db) Peak Emission Limit (dbm/hz) = 15 (dbuv) 157 = -142 (dbm/hz) Assumed industry limit Convert dbuv to dbm/hz given 100KHz measurement bandwidth Emission Transfer Function (db) measured and defined per tazebay_3bp_01a_0113.pdf is a measure of differential (transmitted signal) to common mode (emission) conversion using strip lines. Page 4
TX PSD Peak Guide Line Emission-TF(dB) -40-45 -50-55 -60-65 -70-75 -80-85 Emission-TF, 6dB margin over best measured cable sample -90 0 100 200 300 400 500 600 700 800 900 1000 Frequency(MHz) TX PSD Peak Guide Line -55-60 -65-70 -75-80 -85-90 -95-100 TX PSD Peak Guide Line assuming 15dBuV Peak Emission Level -105 0 100 200 300 400 500 600 700 800 900 1000 Frequency (MHz) Page 5
Measurement Set-up and Assumptions An arbitrary Waveform Generator (AWG) and a Spectrum Analyzer used to generate and measure TX PSD under various conditions. TX Voltage of 1Vpp assumed at MDI. TX PSD shaping of (0.75 + 0.25 Z -1 ) was applied. Baud rate of 750Msps assumed. Every three bits of scrambler assigned to a 2D-PAM3 symbol with gray mapping. TX PSD peak was compared against PSD Peak Guide Line for different scrambler sizes. Effect of un-scrambled parity bits in FEC is measured. Effect of 80/81 Idle code is measured. Page 6
TX PSD Peak: 17-bit Scramblers after FEC, zero data -70-75 PSD peak Guid Line PSD peak, Master PSD peak, Slave PSD rms. Master/Slave -80 PSD rms Limit PSD, dbm/hz -85-90 -95-100 50 100 150 200 250 300 350 Frequency (MHz) Master Scrambler Polynomial: X 17 +X 6 +1 Slave Scrambler Polynomial: X 17 +X 11 +1 Page 7
TX PSD Peak: 15-bit Scramblers after FEC, zero data -70-75 PSD peak Guid Line PSD peak, Master PSD peak, Slave PSD rms. Master/Slave -80 PSD rms Limit PSD, dbm/hz -85-90 -95-100 50 100 150 200 250 300 350 Frequency (MHz) Master Scrambler Polynomial: X 15 +X 4 +1 Slave Scrambler Polynomial: X 15 +X 11 +1 Page 8
TX PSD Peak: 15-bit Scramblers before FEC, zero data -70-75 PSD peak Guid Line PSD peak, Master PSD peak, Slave PSD rms, Master PSD rms, Slave -80 PSD rms Limit PSD, dbm/hz -85-90 -95-100 50 100 150 200 250 300 350 Frequency (MHz) Master Scrambler Polynomial: X 15 +X 4 +1 Slave Scrambler Polynomial: X 15 +X 11 +1 Page 9
TX PSD Peak: 15-bit Scramblers before FEC, 80/81 Idle -70-75 PSD peak Guid Line PSD peak, Master PSD peak, Slave PSD rms, Master PSD rms, Slave -80 PSD rms Limit PSD, dbm/hz -85-90 -95-100 50 100 150 200 250 300 350 Frequency (MHz) Master Scrambler Polynomial: X 15 +X 4 +1 Slave Scrambler Polynomial: X 15 +X 11 +1 80/81 Idle Pattern: 1_00001010_10001010_01001010_11001010_00101010_10101010_01101010_11101010_00011010_10010010 Page 10
Other considerations for scrambler selection TX PSD should meet defined PSD rms limit. Need to have low cross correlation between Master and Slave scramblers in order to a robust echo cancellation in full duplex operation. Need to have a good auto correlation to resolve ISI and Echo given 15m cable length. No cross talk cancellation required (Single pair communication). Computer Simulations for 1000BASE-T1 channels confirmed proper training of DSP in full duplex operations using 17-bit and 15-bit scramblers when compared to 33-bit scramblers. Page 11
Killer Packets? Probability of having a random data sequence matching the 15-bit scrambler sequence is (1/2)^ (2^15-1) which is practically zero. Probability of having a match between scrambler and data sequence in a shorter length exists for any scrambler design and need to be addressed in receiver. For example the probability of having a match of length 60 bits is 8.7e-19 which takes more than 48 years to happen with baud rate of 750MHz. A match half this length (30 bits) may happen every second! Killer packets are discussed in the past for 100BASE-TX which uses a 11- bit scrambler over two pairs. Existing 100BASE-TX receivers are designed to survive killer packets. Killer packets do not kill anymore! Page 12
Summary Various scrambler sizes and polynomials studied for 1000BASE-T1. Two 15-bit scramblers shown in this document are seen to pass peak emission requirement and able to train system in full duplex format. It is proposed that the following 15-bit polynomials are used for scrambling in Idle and Data modes. Master Scrambler Polynomial : X 15 +X 4 +1 Slave Scrambler Polynomial : X 15 +X 11 +1 The suggested 15-bit scrambler polynomials are expected to improve peak emission by up to 3dB as compared to 33-bit scramblers and even more when compared to a 58-bit scrambler. Page 13
Appendix Earlier reported measurements on 1000BASE-T and 10GBASE-T Page 14
1000BASE-T: Forced Master Idle -55-60 rms peak PSD (dbm/hz) -65-70 -75-80 -85 0 10 20 30 40 50 60 Frequency, MHz Peak PSD values measured for 1000BASE-T in forced master mode (PAM3 with 33bit scrambler) on 100KHz RBW are seen larger than rms by more than 10dB. Page 15
10GBASE-T: Test Mode 1 (PAM2 wo/thp) -65 rms peak -70 PSD (dbm/hz) -75-80 -85-90 0 50 100 150 200 250 300 350 400 Frequency, MHz Peak PSD values measured for 10GBASE-T in test mode 1 (PAM2 without THP) and on 100KHz RBW are seen larger than rms by more than 10dB. This mode uses PRBS 33 training pattern (33bit scrambler). Page 16