Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [802.3_E4AI] COM starting values for 400 G

I have this for inputs so far.

pam

enob

SNR_Tx

Deterministic Jitter

R_LM

Random Jitter

Rx FFE pre taps

Px FFE post taps

Rx FFE total taps

Rx ffe floating tap groups

Taps per Group

Last floating tap relative to the cursor

A_dd

Sigma_r

 

4

5

31.5

0.02

0.95

0.01

8

22

30

2

6

160

6

5.5

32.5

0.022

0.97

0.011

7

28

35

2

5

120

8

6

33.5

0.023

0.98

0.012

6

14

20

2

4

100

 

Would this be  reasonable start point for channel analysis?

 

From: Bowman, Kurtis <Kurtis.Bowman@xxxxxxx>
Sent: Thursday, June 26, 2025 9:03 PM
To: richard.mellitz@xxxxxxxxx; STDS-802-3-E4AI@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_E4AI] COM starting values for 400 G

 

[AMD Official Use Only - AMD Internal Distribution Only]

 

Good to know. I’m certainly not a n expert here either 

 

Thanks,

Kurtis 

From: Richard Mellitz <richard.mellitz@xxxxxxxxx>
Sent: Friday, June 27, 2025 8:49:34 AM
To: STDS-802-3-E4AI@xxxxxxxxxxxxxxxxx <STDS-802-3-E4AI@xxxxxxxxxxxxxxxxx>
Subject: Re: [802.3_E4AI] COM starting values for 400 G

 

Caution: This message originated from an External Source. Use proper caution when opening attachments, clicking links, or responding.

 

For KR4 (NRZ)  we used Nb= 14

For KP4 (PAM4) we used Nb= 16

I can’t remember why though.

 

Which is the opposite direction of what you were suggesting below.

 

 

From: Ali Ghiasi <aghiasi@xxxxxxxxx>
Sent: Thursday, June 26, 2025 6:31 PM
To: Richard Mellitz <richard.mellitz@xxxxxxxxx>
Cc: STDS-802-3-E4AI@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_E4AI] COM starting values for 400 G

 

Hello Rich,

 

ADC ENOB is defend as = (SNDR - 1.76 dB)/6.02, furthermore SNDR and ENOB are frequency dependent.  PAM4 with higher frequency will have lower ENOB, my suggestion is the following for ENOB:

               PAM4 = 5.0 bits

               PAM6 = 5.5 bits

               PAM8 = 6.0 bits.

 

Proposed RX FFE fixed taps:

PAM4 = 30 UI

PAM6 = 25 UI

PAM8 = 20 UI

 

In DJ we doubled the number of taps, here is my suggestion for last floating taps:

PAM4 = 160 UI

PAM6=130 UI

PAM8 = 100 UI

 

 

Thanks,

Ali Ghiasi
Ghiasi Quantum LLC

 

 

 

On Jun 26, 2025, at 2:48 PM, Richard Mellitz <richard.mellitz@xxxxxxxxx> wrote:

 

Hi All,

 

I put together a stake in the ground for 400 Gb/s COM so we can better analyze channels.

This not a recommendation. It’s starting point for consensus discussion.  I’m not convinced the progression direction is correct for some the parameter makes sense. I can argue either way.

Comments and suggestions welcome. 

 

 

 

pam

enob

SNR_Tx

Deterministic Jitter
A_dd

R_LM

Random Jitter

Sigma_r

Rx FFE pre taps

Px FFE post taps

Rx ffe floating tap groups

Taps per Group

Last floating tap relative to the cursor

4

6.5

33.5

0.02

0.95

0.01

8

15

2

6

100

6

6

31.5

0.022

0.93

0.011

7

12

2

5

65

8

5.5

29.5

0.023

0.9

0.012

6

10

2

4

50

 

Rich Mellitz, Samtec

 

To unsubscribe from the STDS-802-3-E4AI list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-E4AI&A=1

 

To unsubscribe from the STDS-802-3-E4AI list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-E4AI&A=1

To unsubscribe from the STDS-802-3-E4AI list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-E4AI&A=1