cu4hdd backplane channel analysis -...
TRANSCRIPT
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 1
CU4HDD Backplane Channel Analysis
Presenter: Peter Wu, Marvell
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 2
Outline
• Analysis of 54 SAS backplane channels (www.t10.org)
• Channels are from connector to connector (TP1 <-> TP4)
• IL - Insertion loss
• ICR ( Insertion loss to crosstalk ratio )
• Simulation results using COM model ( Annex 93A 802.3bj with
updated configurations)
• Simulation results with Stateye V4.2.3
• Performance analysis for PCS coding and equalizations
• PCS coding: 8B/10B vs. 64B/66B encoding
• TX EQ– 3 tap TX FIR
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 3
COM – Channel Operating Margin ( Annex 93A) – Simulation Configurations
• 2.5G-BASE-X – 8B/10B
• Baud rate: 3.125G.
• TX EQ: No
• DFE: No
• 5G BASE-X –8B/10B
• Baud rate: 6.25G
• With/Without: TX EQ of 3-Tap FIR
• DFE: 6-Tap
• 5G BASE-R –64B/66B
• Baud rate: 5.15625G
• With/Without: TX EQ of 3-Tap FIR
• DFE: 6-Tap
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 4
0 1 2 3 4 5
-40
-35
-30
-25
-20
-15
-10
-5
0
X: 1.56
Y: -8.52
Insertion Loss
Frequency (GHz)
IL (
dB
)
X: 2.58
Y: -14.87 X: 3.13
Y: -17.24
X: 2.81
Y: -19.25
Channel Analysis of Insertion loss
• HP24, HP25, HP26 are with worst Insertion loss
• Large insertion loss deviation @ ~2.8GHz.
Worst IL for 5G with
64B/66B Encoding
Worst IL for 5G with
8B/10B Encoding
IL @ Nyquist for 5G
with 8B/10B Encoding
Scheme 2.5G
8B/10B
5G
64B/66B
5G
8B/10B
Worst IL
(dB)
8.52 14.87 19.25 Worst IL for 2.5G with
8B/10B Encoding
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 5
1 2 3 4 5 6
-30
-20
-10
0
10
20
30
40
50
60
70
X: 2.8
Y: 15.68
Insertion to Crosstalk Ratio
Frequency (GHz)
dB
X: 2.58
Y: 20.41
X: 1.56
Y: 29.27
Channel Analysis of ICR
• Some channels have ICR dip @ ~2.8GHz
Worst ICR for 5G with
64B/66B Encoding
Worst ICR for 5G with
8B/10B Encoding
Worst ICR for 2.5G
with 8B/10B Encoding
Scheme 2.5G
8B/10B
5G
64B/66B
5G
8B/10B
Worst IL
(dB)
29.27 20.41 15.68
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 6
Simulation Results with COM - 2.5GBASE-X
• 2.5Gbps with 8B/10B encoding
• 3dB COM reserved for loss from implementations
• All channels pass without DFE or TX EQ
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53
CO
M (
dB
)
Channel ID
2.5G 8B/10B
No DFE
6-Tap DFE
3dB COM Limit
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 7
Simulation Results with COM -5GBASE-R
• 5Gbps with 64B/66B encoding
• All channels pass without TX EQ
0
1
2
3
4
5
6
7
8
9
10
11
12
13
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53
CO
M (
dB
)
Channel ID
5G 64B/66B
3 Tap TX FFE
NO TX FFE
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 8
Simulation Results with COM -5GBASE-X
• 5Gbps with 8B/10B encoding
• 20% of channels fail even with TX EQ
• Non-optimal DFE shape due to 8B/10B idle patterns may result in
further performance loss ( Lo_802.3CU4HDD_01_0915)
0
1
2
3
4
5
6
7
8
9
10
11
12
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53
CO
M (
dB
)
Channel ID
5G 8B/10B
3 Tap TX FFE
NO TX FFE
3tap TX EQ
No TX EQ
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 9
Performance Analysis with COM – IL dip
• HP25 has IL dip @ around 2.8GHz and ICR dip @ around 2.8GHz
• Out of band for 64B/66B scheme (Nyquist at 2.578125GHz)
• In band for 8B/10B scheme (Nyquist at 3.125GHz)
1 2 3 4 5 6-40
-35
-30
-25
-20
-15
-10
-5
X: 2.58
Y: -11.71
Insertion Loss
Frequency (GHz)
IL (
dB
)
X: 2.83
Y: -19.18
HP25
1 2 3 4 5 6
-20
-10
0
10
20
30
40
50
X: 2.81
Y: 17.24
Insertion Crosstalk Ratio
Frequency (GHz)IC
R (
dB
)
X: 2.58
Y: 23.58
HP25
Worst IL with
64B/66B Encoding
Worst IL with
8B/10B Encoding
Worst ICR with
64B/66B Encoding
Worst ICR with
8B/10B Encoding
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 10
Simulation Results Using Stateye v4.2.3
• For additional margin - target BER 10-15
• No TX EQ added
• 6-Tap DFE
• Jitter model added
• Results of eye opening for all 54 channels
• Minimum Required Eye Opening 0.2V
• Observations: • 5GBASE-X Mode(8B/10B), 10 channels fail the required eye opening.
• 5GBASE-R Mode(64/66B) and 2.5GBASE-X(8B/10B) all channels pass
• The results align with that of COM analysis
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 11
Simulation Results Using Stateye – 5G mode
Channels fail
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 12
Simulation Results Using Stateye – 2.5G mode
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 13
Conclusions
• 2.5Gbps
• ICR are good for all channels.
• All channels pass without TX EQ or DFE.
• 8B/10B is feasible
• 5Gbps
• All channels pass with 64B/66B encoding and no TX EQ
required
• Some channels fail with 8B/10B encoding
• 64B/66B is a better choice than 8B/10B for both performance
and implementation considerations
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 15
List of Channels : 54 channels
15
channel ID channel XTALK
1 'HP01' 'HP19'
2 'HP01' 'HP15+HP16+2HP17+2HP18'
3 'HP02' 'HP19'
4 'HP02' 'HP15+HP16+2HP17+2HP18'
5 'HP03' 'HP19'
6 'HP03' 'HP15+HP16+2HP17+2HP18'
7 'HP04' 'HP19'
8 'HP04' 'HP15+HP16+2HP17+2HP18'
9 'HP05' 'HP19'
10 'HP05' 'HP15+HP16+2HP17+2HP18'
11 'HP06' 'HP19'
12 'HP06' 'HP15+HP16+2HP17+2HP18'
13 'HP07' 'HP19'
14 'HP07' 'HP15+HP16+2HP17+2HP18'
15 'HP08' 'HP19'
16 'HP08' 'HP15+HP16+2HP17+2HP18'
17 'HP09' 'HP19'
18 'HP09' 'HP15+HP16+2HP17+2HP18'
19 'HP10' 'HP19'
20 'HP10' 'HP15+HP16+2HP17+2HP18'
21 'HP11' 'HP19'
IEEE 802.3 CU4HDDSG – Nov 2015 Plenary Meeting, Dallas, TX 16
List of Channels
22 'HP11' 'HP15+HP16+2HP17+2HP18'
23 'HP24' 'HP19'
24 'HP24' 'HP15+HP16+2HP17+2HP18'
25 'HP25' 'HP19'
26 'HP25' 'HP15+HP16+2HP17+2HP18'
27 'HP26' 'HP19'
28 'HP26' 'HP15+HP16+2HP17+2HP18'
29 'long_board_to_drive_oldconn' 'long_board_to_drive_oldconn_next'
30 'short_board_to_drive_oldconn' 'short_board_to_drive_oldconn_next'
31 'long_board_to_board' 'long_board_to_board_FEXT'
32 'short_board_to_board' 'short_board_to_board_FEXT'
33 'b1_thu' 'b1_next'
34 'b2_thu' 'b2_next'
35 'c1_thu' 'c1_next'
36 'c2_thu' 'c2_next'
37 'd1_thu' 'd1_next'
38 'd1_thu' 'd1_lcc'
39 'd2_thu' 'd2_next_hdd'
40 'd2_thu' 'd2_next_lcc'
41 'a2_thu' 'a2_next'
42 'a2_thu' 'a2_lcc'
43 'Intel_HDD_BP_C_MB_03_thru' 'Intel_HDD_BP_C_MB_03_FEXT'
44 'Intel_HDD_BP_C_MB_04_thru' 'Intel_HDD_BP_C_MB_04_FEXT'
45 'Intel_HDD_SC_MB_11' 'Intel_HDD_SC_MB_11_FEXT'
46 'Intel_HDD_SC_MB_12' 'Intel_HDD_SC_MB_12_FEXT'
47 'Intel_MB_C_BP_HDD_01_thru' 'Intel_MB_C_BP_HDD_01_FEXT'
48 'Intel_MB_C_BP_HDD_02_thru' 'Intel_MB_C_BP_HDD_02_FEXT'
49 'Intel_MB_LC_HDD_05' 'Intel_MB_LC_HDD_05_FEXT'
50 'Intel_MB_LC_HDD_06' 'Intel_MB_LC_HDD_06_FEXT'
51 'Intel_MB_LC_HDD_07' 'Intel_MB_LC_HDD_07_FEXT'
52 'Intel_MB_LC_HDD_08' 'Intel_MB_LC_HDD_08_FEXT'
53 'Intel_MB_SC_HDD_09' 'Intel_MB_SC_HDD_09_FEXT'
54 'Intel_MB_SC_HDD_10' 'Intel_MB_SC_HDD_10_FEXT'