Hossein,
Great, thanks for the clarification.
I’ll incorporate the change to use cursor location instead of the max for clip_method = Fast.
And then create a Merge Request.
Thanks,
Adam
|
From: Hossein Shakiba <hossein.shakiba@xxxxxxxxxx>
Sent: Wednesday, June 18, 2025 1:05 PM
To: Adam Gregory <adam.gregory@xxxxxxxxxx>; Kent Lusted <lusted@xxxxxxxxxxxx>; Richard Mellitz <Richard.Mellitz@xxxxxxxxxx>; howard.heck@xxxxxx; Adam Healey (adam.healey@xxxxxxxxxxxx) <adam.healey@xxxxxxxxxxxx>; stds-802-3-com <stds-802-3-com@xxxxxxxxxxxxxxxxx>;
802d3.mtg@xxxxxxxxx
Cc: Adee Ran (Cisco) <aran@xxxxxxxxx>
Subject: RE: [COM] IEEE 802.3 COM ad hoc teleconference meeting
|
Caution: This email originated from outside the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
|
Hi Adam,
Thanks for your follow-up and further looking into this.
My intention is what you explained. That is, for calculating quantization noise at this step of the code, pulse response with TxFFE and CTLE, but
without RxFFE is needed for all 4 paths you listed. I quickly looked at the latest changes and used the new single .m file to run a sample case and they look fine.
I am currently running more cases.
In response to your last question, sampling the pulse response at the cursor location in the preferred intention. The reason I used sampling at the
max was that under that condition cursor location was not available. This has changed with the new changes, so using cursor location under all condition is the better approach.
As a result, I agree with further modifying get_PSDs function to remove lines 5005-5006 from the single .m file and uncomment line 5010. In my current
runs I have applied this modification too.
Again, thanks for your review and suggestions.
Regards,
Hossein
From: Adam Gregory [mailto:adam.gregory@xxxxxxxxxx]
Sent: June 17, 2025 4:35 PM
To: Kent Lusted <lusted@xxxxxxxxxxxx>; Richard Mellitz <Richard.Mellitz@xxxxxxxxxx>;
howard.heck@xxxxxx; Hossein Shakiba <hossein.shakiba@xxxxxxxxxx>; Adam Healey (adam.healey@xxxxxxxxxxxx) <adam.healey@xxxxxxxxxxxx>;
stds-802-3-com <stds-802-3-com@xxxxxxxxxxxxxxxxx>;
802d3.mtg@xxxxxxxxx
Cc: Adee Ran (Cisco) <aran@xxxxxxxxx>
Subject: RE: [COM] IEEE 802.3 COM ad hoc teleconference meeting
Hossein,
It was hard to explain what I was getting at today in the meeting, so I just made a code branch to discuss the question.
Link to test branch:
https://opensource.ieee.org/adam.gregory/com_code/-/tree/Quantization_Pulse?ref_type=heads
Link to single .m file:
This branch is created by taking the changes you make for commit request 4p9_1 and then making an additional update on top of it.
The main question I have is:
Should the section in get_PSDs that computes quantization noise operate on the pulse response that contains TxFFE and CTLE?
If the answer is yes, there are a few inconsistencies I have made modifications for.
If the answer is no, perhaps you could help explain the reason for the different type of pulse response used for different methods.
You have 4 paths to compute Quantization Noise:
- Clip Method = Slow, OP.COMPUTE_COM = 0
- Clip Method = Slow, OP.COMPUTE_COM = 1
- Clip Method = Fast, OP.COMPUTE_COM = 0
- Clip Method = Fast, OP.COMPUTE_COM = 1
I noticed the following things:
- Path #2 (clip=slow, Compute_COM=1), the pulse response used had TxFFE, CTLE, and RxFFE included. This is because the field “ctle_imp_response” in Apply_EQ has been created
with RxFFE included. I don’t think the intention was to ever have RxFFE included when computing quantization noise.
- Path #3 and #4 ( clip = fast), the pulse response only included CTLE. TxFFE was omitted. In the usual case where TxFFE = 1, this will not make a difference, but it seems
like the TxFFE should have been added here since it was added to the Slow method.
I changed the behavior to what I think is intended by making the following updates
- Add “pulse_response_w_CFT_TXFFE_noRxFFE” field to chdata in Apply_EQ so that it is available when OP.COMPUTE_COM = 0 or 1
- Change the Quantization Noise section in get_PSDs to use pulse_response_w_CFT_TXFFE_noRxFFE so you are guaranteed to always have only CTLE and TxFFE included.
There was another question which I am not sure about:
I noticed that the Fast method samples the pulse response at the max instead of at the cursor location. Is this intended or should it instead use the sampled pulse response that is used for the Slow Method?
Thanks,
Adam
-----Original Appointment-----
|
From: Kent Lusted <lusted@xxxxxxxxxxxx>
Sent: Thursday, May 22, 2025 4:13 PM
To: Kent Lusted; Richard Mellitz; Adam Gregory;
howard.heck@xxxxxx; 'Hossein Shakiba'; Adam Healey (adam.healey@xxxxxxxxxxxx); stds-802-3-com;
802d3.mtg@xxxxxxxxx
Cc: Adee Ran (Cisco)
Subject: [COM] IEEE 802.3 COM ad hoc teleconference meeting
When: Tuesday, June 17, 2025 7:00 AM-8:00 AM (UTC-08:00) Pacific Time (US & Canada).
Where:
https://synopsys.zoom.us/j/92236452845?pwd=7bctzMuZq5AFCFaVXrTE3Jl0xEFbl0.1&from=addon
|
Caution: This email originated from outside the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe.
|
I want to remind all teleconference meeting participants to review the following documents prior to participation in an IEEE 802.3 meeting teleconference:
-
IEEE SA patent policy
-
IEEE SA Copyright Policy
-
IEEE SA Participation Policy and Code of Ethics
All of these policies may be found at
http://ieee802.org/3/policies.html
|
|||||||||||||||||||||||||||||||||||||||||||
To unsubscribe from the STDS-802-3-COM list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-COM&A=1