Hi Gerrit,
Thank you for the presentation and our apologies as we couldn’t join the call due to a conflicting appointment.
To clarify the reason for the high frequency dual slope :
At that time, input from the PHY designers seemed to have been to have a most stringent Return Loss mask. The first break point and slope were decided based on this. Since the gentle slope was not sustainable at very high frequencies, a
second break point was added to get a steeper rise.
In short, the very HF mask was pretty much tracking what was assumed to be technically feasible at that time (with some margin). Our previous presentations are referenced here-
Ref:1,
Ref:2,
Ref:3 and
Ref:4. The existing mask was as presented in 3 and adopted in 4 with a lower high frequency cutoff (4000 vs 5500 MHz).
However, as was pointed out in later contributions, PHYs might require additional ESD protection which adds additional parasitic capacitance and impacts the HF RL.
There were masks shown
here and
here, but I don’t believe these were proposed for adoption (I am not sure of this as this was discussed during the Long Beach interim which I couldn’t attend).
As you correctly pointed out, the existing mask is very stringent and if the PHY designers agree on this, we certainly support a relaxed mask. Please see our latest contribution
here.
Regarding having a separate mask for different speeds within NGAUTO PHYs:
We must note that once designed, the MDI would not change and hence if a certain MDI is designed for a lower data speed say 2.5G, it may not be able to support dynamically increasing the data speed.
This is not necessarily a bad thing, we just need to be aware of this.
Having said this, it is correct that MDI RL is primarily determined by the requirements of the PHY and this spec should be driven by the PHY designers.
I have also attached a modified version of our presentation. An additional reference is added in this which indicates the
impact of relaxing the lower corner frequency of MDI Return Loss on data transmission.
Thank you again and looking forward to the discussions in Vienna!
With Regards,
Gitesh Bhagwat
Sr. Factory Applications Engineer
PoDL Products
Office (805) 560-7669
Websites analog.com, linear.com
Linear Technology is now part of Analog Devices. Learn
more.
-----Original Message-----
From: Gerrit den Besten <gerrit.den.besten@xxxxxxx>
Sent: Wednesday, July 10, 2019 9:26 AM
To: STDS-802-3-NGAUTO@xxxxxxxxxxxxxxxxx
Subject: [802.3_NGAUTO] MDI Return Loss
[External]
I presented material in today's ad-hoc call on modifications to the MDI return loss.
http://www.ieee802.org/3/ch/public/adhoc/DenBesten_3ch_01_0719.pdf
Key messages: 1) make the MDI RL limit scale with speed to avoid overspecification of the lower rates and 2) modify/relax the high-frequency limit to improve feasibility without compromising receiver margin.
An important question that was asked is where the high-frequency dual-slope MDI return loss curve in the current draft stems from. As far as I know it came from this contribution in Spokane:
http://www.ieee802.org/3/ch/public/sep18/bhagwat_3ch_01a_0918.pdf
The MDI return loss limit seems to be constructed as a modified link return loss limit to accommodate for a higher return loss above Nyquist. However the mechanisms for both are different. The link segment return loss and MDI return
loss are typically quite a bit different. As other contributions have pointed out the limit in the current draft is very strict so it seems we have to do something about it.
Looking forward to hear your feedback.
See you soon in Vienna.
Best regards,
Gerrit den Besten
Fellow, NXP Semiconductors, NL
________________________________________________________________________
To unsubscribe from the STDS-802-3-NGAUTO list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-NGAUTO&A=1
To unsubscribe from the STDS-802-3-NGAUTO list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-NGAUTO&A=1