Hi Roberto, and co-authors
Thanks again for your very detailed proposal contained in your presentation,
https://www.ieee802.org/3/df/public/22_11/rodes_3df_01a_2211.pdf
The updated version which you showed during the meeting last Wednesday is now also available on-line.
As you know I raised some questions to better understand the proposal. Unfortunately I wasn’t unable to fully understand the answers provided, probably due to the non-optimum audio
conditions in the meeting room.
Therefore I will be raising them again in this email (also in view of the updated presentation).
The first question which I raised was related to the range between OMAouter min (for max TDECQ), being 4.4 dBm, and OMAouter max, being 5 dBm, and how that would work in manufacturing,
considering that during the cu project there were requests to raise that difference for the 400GBASE-LR4-6 from 2.1 to 2.5 dB.
It was my understanding that you acknowledged that this range was tighter than desirable and that it would need to be considered if it would be possible to raise the OMAouter max
to 5.5 dBm.
This however raises further questions:
·
Is a range of 1.1 dB for OMAouter at max TDECQ sufficient for setting in manufacturing, considering that +/- 0.25 dB is generally seen as a ballpark testing/reproducibility accuracy? Or would you need to go more towards a range of 2 dB which seems to
be a ballpark minimum range of in-force IEEE optical PMDs.
·
Is a nominal Tx power of 5 dBm (OMAouter) achievable with current 200G PAM4 laser technology?
·
Is a Receiver overload of 5.5 dBm achievable, considering that it would seem important to operate in all back-to-back configurations?
·
Was is the impact on the FWM analysis by the raise in power by 0.5 dB?
One intermediate question is related to the proposal for “Difference in launch power between any two lanes” to be “3 dB” max.
Is this average launch power or OMA? In the in-force specifications for 400GBASE-FR4 and 400GBASE-LR4-6 it is in OMA.
Now back to my second question raised at last week’s meeting.
In your specification proposal on slide 6, you propose a max difference in launch power between any two lanes of 3 dB.
Shouldn’t we take that range into consideration for the analysis of FWM?
In October there was a presentation by your co-author Xiang Liu proposing to use a flat distribution of power per channel, whereas his first (and superseded) presentation assumed
a range of 3 dB. Wouldn’t we need to use that range of 3 dB for the FWM analysis?
Now, having been able to further look into the proposals in your updated presentation, I have some further questions for clarification, especially around the assessment of FWM penalty.
·
What are the precise assumptions used for this analysis looking at your slide 5 (and your October presentation
https://www.ieee802.org/3/df/public/22_10/22_1012/rodes_3df_01b_221012.pdf)?
o
Are you indeed using a lambda zero Gaussian distribution between 1306 and 1322nm? Instead of a flat distribution?
o
In your October presentation you mentioned “Uniform distribution of ZDW is unrealistic and it overestimates FWM probability”. How would the analysis turn out if we would use a less rigid approach like distribution between 1303 and 1322nm? Flat versus
Gaussian. It would be helpful to understand if we are close to running off a cliff or not.
·
Are you indeed using Gaussian distributions for Lambda Zero, AND, optical power as I seem to understand from your slide 5?
·
Do I understand correctly that you are using an OMA distribution with a range of 1.5 dB instead of the 3 dB range you propose on slide 6?
Then on the combined TDECQ + FWM penalty.
·
Are you indeed assuming that the combination never goes above 3.9 dB?
·
So are you assuming that for the channels with TDECQ close to 3.9 dB the FWM penalty will be negligible?
·
Considering that the highest CD penalty would seem to occur at the 1309.14 nm channel with 9.5 ps/nm, would you indeed expect the FWM penalty for that channel to be negligible? And similarly for the 1295.56nm channel with -28.4 ps/nm?
o
How dependent is this assumption on the fiber lambda zero?
I am sorry for the many questions raised in this email but I think it’s important to get clear answers in our attempt to create a robust and manufacturable 800G-LR4 specification
and to remove any shred of doubt.
Thanks in advance.
Kind regards,
Peter