| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Hi Piers, It’s not just about what the receivers can do theoretically, but also what the respective transmitters can do, and what the cost benefit is for squeezing one more than the
other to achieve a particular link budget. For receivers, absorption efficiency changes with wavelength and depends on material systems, light collection efficiency for MMF vs SMF photodiodes is different (MMF is usually
a bit better). In addition, recent MMF photodiode designs use different structures to improve light absorption and sensitivity. The net result is that the range of receiver sensitivity for single channel MMF vs SMF receivers, for the same bit rate and modulation format, is similar. The spread in sensitivity
for each is greater than the offset of the averages of the population. Best wishes jonathan From: Piers Dawe [mailto:piersd@xxxxxxxxxxxx]
Peter, In
dawe_062718_02_3cd_adhoc I wrote: After converting optical power to photocurrent, this is equivalent to 50GBASE-LR (10 km) and 1.5 dB harder for the receiver than 50GBASE-FR (2 km) So let's do that. For a quantum efficiency near 1, photocurrent/power = wavelength/1240 (see e.g.
https://en.wikipedia.org/wiki/Responsivity ) Rearranging, photocurrent = power*wavelength/1240 For the nominal wavelengths of 850 nm and 1310 nm, that's photocurrent = power*wavelength/1240 photocurrent = power*0.686 or 1.056, In decibels: or dBm(power)-1.64 or dBm(power)+0.24 10 log10(1.056/0.686) = 10 log10(1310/850) = 1.88 dBo Long wavelength photodiodes' quantum efficiencies can be pretty good. In power (dBm): 50GBASE-SR: D1.0 to D3.0 URS = –7 D3.1 RS = max(–6, SECQ – 7.9) D3.2 RS = max(–6, SECQ – 7.9) D3.3 RS = max(–6.5, SECQ – 7.9) Proposed correction: max(–6.1, SECQ – 7.5) 50GBASE-FR: RS = max(–6.9, SECQ – 8.3) 50GBASE-LR: RS = max(–8.4, SECQ – 9.8) but in photocurrent (optical dBmA): 50GBASE-SR: D1.0 to D3.0 URS =
–8.64 D3.1 RS = max(–7.64, SECQ
– 9.54) D3.2 RS = max(–7.64, SECQ
– 9.54) D3.3 RS = max(–8.14, SECQ
– 9.54) Proposed correction: max(–7.74, SECQ
– 9.14) 50GBASE-FR: RS = max(–6.66, SECQ
– 8.06) 50GBASE-LR: RS = max(–8.16, SECQ
– 9.56) As I said, for similar SECQ, the sensitivity needed for 50GBASE-SR is almost the same as 50GBASE-LR and 1.5 dB harder than for 50GBASE-FR. Even
after the correction, it's still only 0.4 dB easier than for 50GBASE-LR, 1.1 dB harder than for 50GBASE-FR, and a little harder than 50GBASE-SR in D1.0 to D3.0. stassar_3cd_01_0717 showed a healthy sensitivity margin with reasonable eyes, so we could agree to tighten the receiver sensitivity a little for
an agreed reason. But in this SR case, the trigger was "Changes to other optical specs when TDECQ spec limit is adjusted"
king_3cd_01_0518 which didn't show information about sensitivity, nor did it give a reason for tightening the sensitivity.
Of course there are differences between a low power short reach MMF product and an SMF product. On the one hand, the SMF receiver may have significant
coupling losses. On the other hand, there is little to no motivation to reduce the MMF Tx optical power by this amount, but reducing the thermal power by removing receiver complexity for a high-density short-reach product would be attractive. But the big difference is: FR and LR have a max. TDECQ of 3.2 and 3.4 dB, so a max. residual penalty, TDECQ – 10log10(Ceq) of 4.2 and 4.4 dB, while
SR has a max TDECQ of 4.5, so a max. residual penalty of 5.5 dB, far higher than any previous optical PMD. As slides 4-5 of "Completing the family of TDECQ-related specifications"
dawe_062718_01a_3cd_adhoc show, this would put an extra burden on the receiver back end, that would need extra front end sensitivity to tolerate. Best regards, Piers From: Peter Stassar [mailto:Peter.Stassar@xxxxxxxxxx]
Hi all, I have some comments to Piers’s presentation dawe_062718_02_3cd_adhoc provided at yesterday’s cd ad hoc call. Piers stated on slide 4
l
In D3.3, the implied unstressed sensitivity for 50GBASE-SR (100 m) is about -7.3 dBm
l
After converting optical power to photocurrent, this is equivalent to 50GBASE-LR (10 km) and 1.5 dB harder for the receiver
than 50GBASE-FR (2 km)
l
As well as much higher stress levels (higher TDECQ and residual penalty
l
This is the wrong way round! The very short reach PMD should be easier and lower power I would like to refresh everybody’s mind on a presentation I gave at the July 2017 meeting in Berlin: http://www.ieee802.org/3/cd/public/July17/stassar_3cd_01_0717.pdf It demonstrated how much margin we have for the 50G FR/LR Rx and that we had the luxury to choose someone in the middle between reasonable
Tx and Rx levels. We demonstrated Rx sensitivities (probably SECQ non zero) of better than -12 dBm, whereas the requirement for SECQ < 1.4 dB would be
-6.9 dBm for 50GBASE-FR and -8.4 dBm for 50GBASE-LR. If you compare this with -6.5 dBm for 50GBASE-SR then there really is no issue and one would probably just make another optimization
because of VCSEL Tx power levels being at another optimum point than for SMF Tx. Furthermore the requirements in Clauses 138 – 140 are: 50GBASE-SR: RS = max(–6.5, SECQ – 7.9) (dB) 50GBASE-FR: RS = max(–6.9, SECQ – 8.3) (dB) 50GBASE-LR: RS = max(–8.4, SECQ – 9.8) (dB) I still read that SR has most relaxed Rx sensitivity, then FR 0.4 dB better and LR again 1.5 dB better. Kind regards, 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 Herikerbergweg 36, 1101 CM Amsterdam The Netherlands Tel: +31 20 4300 832 Mob: +31 6 21146286 To unsubscribe from the STDS-802-3-50G list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-50G&A=1
To unsubscribe from the STDS-802-3-50G list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-50G&A=1 To unsubscribe from the STDS-802-3-50G list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-50G&A=1 |