status update - sdd21 & sdd11/22 model development · january 2005 1 status update - sdd21...
TRANSCRIPT
IEEE Vancouver Interim Meeting January 2005 1
Status Update - SDD21 & SDD11/22 Model Development
John DAmbrosia, Tyco ElectronicsMatt Hendrick, Intel
IEEE Vancouver Interim Meeting January 2005 2
Acknowedgements
Rich Mellitz, IntelSteve Krooswyk, IntelMike Altmann, IntelYves Braem, Tyco ElectronicsMike Resso, AgilentJoe Abler, IBM
IEEE Vancouver Interim Meeting January 2005 3
IntroductionReference dambrosia_c1_0105
Recommended using proposed informative TP1 –TP4 SDD21 Channel Model MaskRecommended eliminating informative TP1 – TP4 SDD11 / 22 model masks
Concern regarding this recommendation expressedRecommended eliminating informative TP1 – TP4 SDD21 Group Delay Variation Mask
Reference Rx models as defined in http://ieee802.org/3/ap/public/channel_model/mellitz_m1_0105.pdf
IEEE Vancouver Interim Meeting January 2005 4
Proposed Informative Model SDD21, TP1 – TP4
Attenuation Limit
f2f1
Channel SDD21 response
HF_∆
LF_∆ RMS fit line
HF_Slope(dB/decade)
LF_∆ TBD[dB]HF_∆ TBD[dB]BS_HF_∆ TBD[dB]F1 TBD(1GHz)F2 TBD(factor*Nyquist)Eng_Limit_dB TBD(-50dB)HF_slope TBD(40 dB/decade)RMS_Fit_line > SDD21 spec line
Eng_Limit_dB
NONO
NONO
YESYESBS_HF_∆
]3^*42^*3*2)2/1(^*1[*)(10log*20 fbfbfbfbenLimitAttenuatio −++−= f = 50 MHz to 15 GHz b1 = 2.25e-5 b2 = 1.20e-10 b3 = 3.50e-20 b4 = 1.25e-30
IEEE Vancouver Interim Meeting January 2005 5
Review of Correlation Analysis• dambrosia_c1_0105
recommended eliminating TP1 – TP4 Informative masks for SDD11/22 and Group Delay Variation.
• Concern expressed to deletion of informative TP1 - TP4 SDD11 /22 mask.
IEEE Vancouver Interim Meeting January 2005 7
Effect due to 10dB packages
len_brd2 – length from connector on Rx line card to AC coupling Caplen_brd2a – length from AC coupling cap to RxUpdate from dambrosia_c1_0105.pdf10 dB packages had an impact.Reference http://ieee802.org/3/ap/public/channel_model/mellitz_m1_0105.pdf
IEEE Vancouver Interim Meeting January 2005 8
Cases #5 and #6 Frequency Characterization(dambrosia_02_0105)
IEEE Vancouver Interim Meeting January 2005 9
Frequency Characterization –Case #5 Cascaded with Rx Packaging
Rx models as defined in http://ieee802.org/3/ap/public/channel_model/mellitz_m1_0105.pdf
IEEE Vancouver Interim Meeting January 2005 10
Frequency Characterization –Case #6 Cascaded with Rx Packaging
Rx models as defined in http://ieee802.org/3/ap/public/channel_model/mellitz_m1_0105.pdf
IEEE Vancouver Interim Meeting January 2005 11
We’ve Seen This Before –A Channel the 5G Part Didn’t Go 10G
6000_Link3_Left_Channel0
Approximation via internal device diagnostics. Includes AC coupling cap and (5G optimized) device packaging effects
Channel measurements using SMA Line cards with no AC coupling Caps
Proposed Model
Reference – “PAM-4 Link Analysis, ”http://grouper.ieee.org/groups/802/3/ap/public/may04/dambrosia_01_0504.pdf”
IEEE Vancouver Interim Meeting January 2005 12
10 Gbps Pulse Response Characterization –Case #5 Cascaded with Rx Packaging
Rx models as defined in http://ieee802.org/3/ap/public/channel_model/mellitz_m1_0105.pdf
IEEE Vancouver Interim Meeting January 2005 13
10 Gbps Pulse Response Characterization –Case #6 Cascaded with Rx Packaging
Rx models as defined in http://ieee802.org/3/ap/public/channel_model/mellitz_m1_0105.pdf
IEEE Vancouver Interim Meeting January 2005 14
Per IBM - Additional Simulations for Cases #5 and #6 (dambrosia_02_0105)
22.1%20.8%33.0%32.9%4. No Packaging, No IC, No Xtalk
19.9%16.5%27.0%27.4%3. No Packaging, No IC
<0 (BER floor at E-11)<0 (BER floor at E-8)17.1%15.7%2. No Packaging
5.5%0% (BER floor at E-12)22.2%18.9%1. Original results
Case 6 FFE3/DFE5
Case 6 FFE3/DFE3
Case 5 FFE3/DFE5
Case 5 FFE3/DFE3
Simulation 1 – Original simulations, as specified in abler_01_00904.pdf
Simulation 2 – As stated in Simulation #1, except packaging removed
Simulation 3 – As stated in Simulation #2, except IC Model (ESD and load structures) replaced with perfect 50Ω terminations.
Simulation 4 – As stated in Simulation #3, except all crosstalk removed
IEEE Vancouver Interim Meeting January 2005 15
Length Analysis in Real Backplanes
0 500 1000 1500 2000 2500 3000 3500 4000
Backplane 1
Backplane 2
Backplane 3
Backplane 4
Backplane 5
Backplane 6
Backplane 7
Backplane 8
Backplane 9
Backplane 10
Backplane 11
Diff
eren
t bac
kpla
nes
Number of Differential Pairs
41-47"36-40.99"31-35.99"26-30.99"21-25.99"16-20.99"11-15.99"6-10.99"Under 6"
“The longest traces in a backplane must be distributed throughout the entire stackup. Selective use of materials on certain layers will not provide enough routing channels for these tracks. This issue grows as you move from a central switching architecture to a mesh architecture. “
Robert JardonJardon EngineeringDirector of Technology
6" to 10.99"30%
Under 6"45%
11 - 15.99"19%
16 - 20.99"6%
IEEE Vancouver Interim Meeting January 2005 16
Pulse Reponse of All Tyco Channels
Note – The two most difficult channels had the greatest peak
Let’s look at Cases #5 and #6
IEEE Vancouver Interim Meeting January 2005 17
ConclusionsParameters used in analysis based on proposed informative TP1 – TP4 SDD21 model are strong indicators of settling time.Need to review parameters in relation to efforts of Signaling AdHoc TP1 – TP4 SDD11 / 22 is meaningful, but not by itself. It works in conjunction with -
Tx Launched SignalChannel (TP1 / TP4) throughputChannel (TP1 / TP4) return lossDevice return lossPackage and IC (ESD / termination) effects
Different channels exist in a backplane environment and each results in different frequency content delivery to the receiver resulting in different reflection characteristics.
IEEE Vancouver Interim Meeting January 2005 18
RecommendationsAdopt proposed TP1/TP4 Informative SDD21 model template
Specific recommendations for each value needs to be completed upon Signal Ad Hoc analysis
Eliminate proposed TP1/TP4 Informative SDD11 / 22 mask
Leave to normative analysis with inclusion of packaging effects
Eliminate TP1/TP4 Informative Group Delay template