|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
During the 802.3 Working Group meeting last Thursday (which I unfortunately was not able to attend), there was further discussion of the 802.3cn technical specs. Given some of the comments that were made, it appears there is a misunderstanding of Chromatic Dispersion Penalty (CDP) scaling with reach and Baud rate, and interplay with transmitter chirp characteristics. Therefore an explanation is warranted.
CDP is proportional to CD coefficient, link length, and Baud rate squared.
CDP ∝ CD * L * B2
Calculating the exact penalty is complex, however the CD * L * B2 term can be used for direct, relative comparisons, by defining it as a CDP Figure of Merit (CDP FM).
For convenience, 10km and 25Gbaud are used for normalization.
CDP FM = CD * (L/10) * (B/25)2
Table 1 below lists key transmitters specs. and CDP FM for various codes, with red highlighting concerns. The take away points are:
(Qualitative statement that 50G PAM4 “is a lot easier” than 100G PAM4 is not true at 40km. In fact the opposite it true; 400G ER8 40km (50G PAM4) is harder than 400G LR4 6km and 100G LR1 10km (both 100G PAM4). )
(This is higher than previous TX OMA values (0.3 and 0.1 dBm for 40G ER4 and 100G ER4, respectively), and is higher than other codes in Table 1. It will significantly drive transmitter cost up by likely requiring a TX amplifier. This undermines the Broad Market Potential, and is counter to the expressed need for low cost by Network Operators.)
Table 2 and 3 below explains the dynamics of the review process in the 802.3cu and 802.3cn.
In 802.3cu the proponents showed favorable data for 400GBASE-LR4 10km using TX favorable to the proposal. The opponents showed unfavorable data using TX unfavorable to the proposal.
In 802.3cu the proponents showed favorable data for 400GBASE-ER8 40km using TX favorable to the proposal. No unfavorable data was shown because no other type of TX was used, i.e. the data is incomplete. And the missing data is exactly in the area of concern. This means Technical Feasibility has not been established.
802.3cn TF needs to go back and take more measurements to establish a complete data set to demonstrate Technical Feasibility Criteria, and Broad Market Potential.
From: Chris Cole
Yes, of course. Excellent catch. Please don’t listen to what I say, but what I mean.
The chirp for the shortest wavelength is approx. -200ps/nm so positive chirp TX is required to compensate, exactly as you discuss in detail. The beauty of LWDM is that for the 1310nm longest wavelength, practically no positive chirp is seen because no SMF in the field has ZD wavelength appreciably below. That’s why when we make measurements for positive dispersion we have to go through such contortions to use artificially long fibers to simulate conditions never seen in actual operation.
You just demonstrated the kind of review that makes 802.3 great. So where were you when the ER8 spec was being written? You actually have to make these EML TOSAs. Where is the concern about +2.4dBm OMA (min) and +1dBm OMA – TDECQ (min), over temp., eight wavelengths and manufacturing spread, while operating in a sufficiently linear regime to get decent PAM4 waveforms? Could it be that deep in your heart you are wishing for Coherent to win? 😊
From: John Johnson <000007ff7d378f43-dmarc-request@xxxxxxxx>
Outside of the larger debate, I have one nit to pick. You stated:
Actually the opposite is true: you need positive chirp to counteract the effects of negative dispersion. In EA modulators, strong positive chirp is easier to achieve than negative chirp, which entails more negative bias, higher modulator insertion loss and possibly decreased linearity. Positive chirp is the “natural” operating mode of an EA modulator, and is easy to achieve with low insertion loss and good linearity. This is the reason that the CD penalty is slightly negative over a wide range of negative dispersion in the presentation by Ooe and Jackson cited below. It’s also the reason that multiple EML suppliers have presented low 10km CD penalties at 1270nm at 100Gbps (but higher CD penalty at 1330nm where negative chirp is needed). Since the 400GBASE-ER8 grid ends at ~1310nm, there is no need for negative chirped modulators for 400GBASE-ER8. At least in this regard, the 25GBd LWDM EML solution is relatively more robust than for 50GBd CWDM.
I can't comment on the capability of SiP transmitters for 400GBASE-ER8. Perhaps Brian can.
On Thu, Sep 12, 2019 at 2:08 AM Chris Cole <chris.cole@xxxxxxxxxxx> wrote:
To unsubscribe from the STDS-802-3-B10K list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-B10K&A=1