| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Hi Peter, thanks for your e-mail, that helps me to better understand your question. The intent of the presentation was to provide the feasibility of 56Gbaud tester to an internal audience, and share the fact that it can work for parts characterization and model tuning. It becomes a presentation to share to 400G IEEE (you probably know as these things works…), yet there are some positive things that can allow to look forward in a positive way. Looking at your presentation now, I recognized you had a very similar tester and very similar results too.
So my slides just relate about an experiment and we’re still away from the right completeness, from any potential implementation for manufacturing, since (as you quoted into your contribution) lot
of questions has to be addressed.
・
Still the losses of the MUX and DEMUX have to take into account, and the minimum power you can get out of the TOSA has to be considered (depends on technology too).
・
Considering the 1340nm and TDP penalty question you pose in your presentation maybe would make sense to compute TDP after an equalization stage (similar to TWDP defined into 802.3aq) that
should have to be agreed, since an equalizer can in principle partially recover the dispersion penalty (in the slide 11 budget a 1.5/1dB split was proposed), for sure this was addressed into detail into some discussion already (apologies since I’m not 100%
synced with the contributions). About the positive of the experiment, the link was decently closed, if I well understood your comment about the budget I think you have to consider that the -10.1dBm shown in this link refer to the
single eye of the PAM. So my measurement of -5.7dBm at 51Gbaud (assuming 1E-4BER as limit) corresponds to -10.5dBm actually including the PAM amplitude reduction of 4.8dB.
In the experiments if one asks if we can manage 4dB+1.5dB demux loss budget (as shown in slide 7, right-bottom chart) answer is yes at 51Gbaud, but this is a side effect with respect the fact we
can use as a tester to verify model assumptions. During the call I said we can probably improve the tester if this makes sense, for sure the RX quantization noise from the real time make us overestimating the BER (for this your results were amazing
if you get a 2E-5 at -5dBm power at 56Gbaud). Unfortunately get high-end equipment (and relative calibration) is a price to pay when dealing with complete new technologies or if the bit rate exacerbate up to 50Gbaud or Gb/s, so we should take
into account this making decision for “low cost” parts (but since it’s correct to assume instrument and components manufacturers as part of the industry, that’s their business portion too).
Back to the experiment, we can expect dedicated RX DACs will be better and so this can be a further step ahead when technology will evolve (and it’s doing rapidly as you know). The TX tester SNR can be also improved, so the system will become less noisy toward a better evaluation of penalties that can affect the link. Hope this will explain my intent, I’ll keep ahead and provide you some more results when I’ll be back on 56Gbaud testing. Thanks and ciao Marco Cisco Photonics Italy (VMCT05) Via Santa Maria Molgora 48/C 28071 Vimercate (MB) ? Italy Phone +39-0392091270 From: Peter Stassar [mailto:Peter.Stassar@xxxxxxxxxx]
Dear Marco, Thanks for your detailed presentation at last Tuesday SMF adhoc meeting. As agreed I would send you this email to clarify my questions. What was your intent with this presentation, demonstrating that 112G PAM4 can work or that you can establish a solid power budget? If it were the first, I would agree that you succeeded, but if it were the latter, I would have some concerns. From the latter point of view it is a pity that you didn’t include slide 11, first slide of your back-up slides, during the adhoc meeting, because I had intended to raise a further question. This slide, taken from bhatt_3bs_01a_0714, shows a needed Rx sensitivity of about -10dBm. Our experiments (details in stassar_3bs_01_0714 in San Diego and song_3bs_01a_0514 to Norfolk) showed a receiver
sensitivity of -6.7 dBm, which is more than 3 dB “short” of your suggested specification. Would you be in a position (now or soon) to confirm a sensitivity of -10dBm by experiments? We would be very interesting to see the receiver sensitivity results of other
organizations. The presentation bhatt_3bs_01a_0714 was presented during the San Diego meeting. After this presentation many concerns were voiced and it be difficult to repeat all here. One specific question from myself was related to the proposal towards a CWDM configuration, specifically about chromatic dispersion penalty testing at the highest envisioned wavelength, being 1338nm.
At this maximum CWDM wavelength the maximum dispersion will be 6.8ps/nm, which is relatively higher than the 100GBASE-LR4 spec of +10ps/nm, so probably we will not be able to ignore the associated penalty, even for 2km SMF. In a similar way as to your presentation
I asked Vipul how to derive the suggested receiver sensitivity from the test results. In my presentation in San Diego, stassar_3bs_01_0714, I expressed that closing a power budget for a 4x112G PAM4 configuration for 2km, is very, very challenging, with either very stringent transmitter
power requirements or very stringent receiver sensitivity requirements. We have made this analysis after extensive optical testing a complete optical link including transmitter and receiver. In the end of the presentation other parties are invited to confirm
our assessments or show improved data. Thanks in advance for the further information. Kind regards, ciao, Peter Peter Stassar,
施笪安
Technical Director,
技术总监
Huawei Technologies Ltd,
华为技术有限公司 European Research Center,
欧洲研究所 Karspeldreef 4, 1101CJ Amsterdam The Netherlands Tel: +31 20 4300 832 Mob: +31 6 21146286 |