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

Re: [STDS-802-3-400G] IEEE P802.3bs TDECQ optimizing eye time centre of eye



 

Mike,

 

Thank you very much for your kind explanation. I could understand your opinion.

 

Best regards,

 

********************************************************

Akinori Hayakawa

FUJITSU LABORATORIES LTD.

Network System Laboratory

Network Component Project

10-1, Morinosato-Wakamiya, Atsugi 243-0197, Japan

TEL : +81-46-250-8251

FAX : +81-46-250-8274

E-mail : hayakawa_aki@xxxxxxxxxxxxxx

********************************************************

 

From: Dudek, Mike [mailto:Mike.Dudek@xxxxxxxxxx]
Sent: Wednesday, September 06, 2017 6:13 AM
To: Hayakawa, Akinori/
早川 明憲; STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: RE: IEEE P802.3bs TDECQ optimizing eye time centre of eye

 

If the test system for the stressed receiver sensitivity test has a reasonably good Tx followed by the BT filter to create most of the stress then I don’t think the optimum sampling time will be significantly different from the 0.45 and 0.55 UI as the BT filter with its flat group delay creates a waveform that is symmetric in time.  The stress in the test therefore won’t be changed.   The stress will only be greater if the original stressed Tx was assymetric in time, and in that case it would have been easier for an RX with adaptive timing to pass the test with that transmitter than with the one assumed in the test (which has the BT filter as the dominant degradation).

 

From: Hayakawa, Akinori [mailto:hayakawa_aki@xxxxxxxxxxxxxx]
Sent: Monday, September 04, 2017 6:06 AM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-400G] IEEE P802.3bs TDECQ optimizing eye time centre of eye

 

 

Mike, Tony and Jonathan,

 

Thank you very much for your replies. I basically support Jonathan’s proposal but I’m just worried about SECQ measurement test.

Do you apply this change to SECQ measurement without any exception? If we adopt this method to SECQ measurement, we have to adjust SECQ measurement timing (timing position of the histograms) iteratively with keeping SECQ value of 3.4 dB and adding higher stress. This means that the final applied stress to the stressed receiver conformance test signal will be greater than the stress that was measured as SECQ of 3.4 dB at original timing of 0.45 UI and 0.55 UI. Please give me comments.

 

Thanks,

 

********************************************************

Akinori Hayakawa

FUJITSU LABORATORIES LTD.

Network System Laboratory

Network Component Project

10-1, Morinosato-Wakamiya, Atsugi 243-0197, Japan

TEL : +81-46-250-8251

FAX : +81-46-250-8274

E-mail : hayakawa_aki@xxxxxxxxxxxxxx

********************************************************

 

From: Jonathan King [mailto:jonathan.king@xxxxxxxxxxx]
Sent: Saturday, September 02, 2017 1:23 AM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-400G] IEEE P802.3bs TDECQ optimizing eye time centre of eye

 

Thanks Mike and Tony, for the succinct responses to Mr Hayakawa’s excellent question.

jonathan

 

From: Tony Zortea [mailto:Tony.Zortea@xxxxxxxxxxxxx]
Sent: Friday, September 01, 2017 8:47 AM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: [EXTERNAL]: Re: [STDS-802-3-400G] IEEE P802.3bs TDECQ optimizing eye time centre of eye

 

Yes, as I understand Jonathan’s proposal this allows the TDECQ measurement to be made with the reasonable assumption that the Rx timing is adaptive.

 

Thanks,

 

Tony

 

multiphy logo

Tony Zortea

Analog CTO

Phone +1 215.703.7003

www.multi-phy.com

 

 

 

From: Dudek, Mike [mailto:Mike.Dudek@xxxxxxxxxx]
Sent: Friday, September 1, 2017 11:32 AM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-400G] IEEE P802.3bs TDECQ optimizing eye time centre of eye

 

If you are able to make the stressed eye exactly per the recipe it won’t increase the stress in the stressed receiver conformance test signal because with the BT filter closing the eye in the TX and the software BT filter in the RX as the only data dependent degradations the optimum eye sampling and the central eye are in the same place.     In the likely event that a real stressed receiver test system isn’t that perfect the stress applied to an Rx that can adapt its timing won’t be any greater than with this perfect test equipment.   Obviously if the Rx isn’t adapting its timing then there is a problem with the Rx, as this change assumes all Rx’s can adapt.

 

From: Hayakawa, Akinori [mailto:hayakawa_aki@xxxxxxxxxxxxxx]
Sent: Thursday, August 31, 2017 11:25 PM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: Re: [STDS-802-3-400G] IEEE P802.3bs TDECQ optimizing eye time centre of eye

 

 

Hi, Jonathan,

 

Do you suggest to apply this change to not only TDECQ measurement but also SECQ measurement? If so, I think this change will give higher stress to the stressed receiver conformance test signal. Is my understanding correct?

 

Best regards,

 

********************************************************

Akinori Hayakawa

FUJITSU LABORATORIES LTD.

Network System Laboratory

Network Component Project

10-1, Morinosato-Wakamiya, Atsugi 243-0197, Japan

TEL : +81-46-250-8251

FAX : +81-46-250-8274

E-mail : hayakawa_aki@xxxxxxxxxxxxxx

********************************************************

 

From: Jonathan King [mailto:jonathan.king@xxxxxxxxxxx]
Sent: Friday, September 01, 2017 3:17 AM
To: STDS-802-3-400G@xxxxxxxxxxxxxxxxx
Subject: [STDS-802-3-400G] IEEE P802.3bs TDECQ optimizing eye time centre of eye

 

Dear colleagues,

Please find attached a short presentation which I hope to review in the 802.3bs face to face meeting in Charlotte, and which describes a proposed change to the TDECQ method which will allow TDECQ to be calculated at the optimum eye timing point, by allowing the histogram time positions to be adjusted for optimum TDECQ.  This change would reflect the capabilities of real world  implementations of receivers with T spaced and T /2 spaced equalizers.

 

Without this change, the TDECQ definition in the current draft (802.3 D3.3) means that transmitters with equalized eyes which are offset from the time centre of the eye are penalized with higher TDECQ, which would contribute to the decorrelation of Tx TDECQ and measured link performance.

 

I believe the change is absolutely necessary to prevent unnecessary transmitter TDECQ failures. 

 

If you are willing to be a supporter of this proposed change please send me an e-mail to let me know your name and affiliation.

Feel free to ask questions of course!

 

Best wishes

Jonathan King