Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [802.3_25GSMF] Parameter changes for MPI penalty



Gary,

The simulation results were presented by Jonathan King in our ad hoc last week: http://www.ieee802.org/3/cc/public/adhoc/170614/king_01_25gsmf_061417.pdf. 

 

On pages 3- 4, the lower charts shows that there are configurations with 6 connectors that have up to 0.7 dB MPI penalty.  This is with transmitter reflectance improved from -12 to -26 dB.   If all connectors are 35 dB or better, then the penalty is 0.22 dB or less. However as soon as you introduce 26 dB connectors, the penalty is much higher.

 

In the discussion we thought that existing cable plant used for 10GBASE-LR would be used for 25GBASE-LR.  The 10GE PMD has -12 dB reflectance Tx and Rx, so there is a significant reflection penalty built into the link budget.  My quick check of the link spreadsheet showed that Prefl is ~ 0.5 dB for 10GBASE-LR and is included as part of the 3.2 dB allocation for penalties.

 

As far as I know, the 40GBASE-ER and 100GBASE-ER4 link budgets don’t have a separate allowance for MPI penalty.  It is not captured by the TDP or the SRS test, so users of those standards with large numbers of connectors are potentially susceptible to this problem.

 

Note that the PAM4 PMDs in 802.3bs and 802.3cd have a 0.5 dB MPI penalty for 10km, but for those links the connector requirements are tighter.  The 6 connector case requires all connectors to be -35 dB or better.  If we made that assumption for 25GBASE-LR, then according to Jonathan’s model the MPI penalty would only be 0.22 dB.

 

The 25GBASE-ER penalty would be lower if we were not aiming to be interoperable with 25GBASE-LR with a minimum loss of 10 dB.

 

So, the question came down to whether we support older cable plants that might have 26 dB RL connectors or we require users to make sure that only 35 dB RL connectors are used.

 

Dave Lewis

 

 

 

From: Gary Nicholl (gnicholl) [mailto:gnicholl@xxxxxxxxx]
Sent: Tuesday, June 20, 2017 3:16 PM
To: David Lewis <David.Lewis@xxxxxxxxxxxx>; STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_25GSMF] Parameter changes for MPI penalty

 

Dave,

 

I haven’t been falling this discussion carefully, but how can there be a 0.7dB MPI penalty for 25GBASE-ER ?  I thought this PMD was based on NRZ, and for NRZ the MPI penalty was minimal (compared to PAM4). I don’t believe we have accounted for 0.7dB MPI penalty in the existing 40GE and 100GE budgets in 802.3ba (or 10GE in 802.3ae for that matter).

 

Gary

 

From: David Lewis <David.Lewis@xxxxxxxxxxxx>
Reply-To: David Lewis <
David.Lewis@xxxxxxxxxxxx>
Date: Monday, June 19, 2017 at 4:47 AM
To: "
STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx" <STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx>
Subject: Re: [802.3_25GSMF] Parameter changes for MPI penalty

 

Pete,

Thank you.  I think this will be the way to do it for 25GBASE-ER.

 

What do people think about 25GBASE-LR?  Should we tighten sensitivity and SRS by 0.7 dB or tradeoff channel insertion loss versus MPI?  I think the consensus from the ad hoc call was to keep the same 6.3 dB budget currently used at 10G and 100G-LR4.

 

David

 

From: Anslow, Peter [mailto:panslow@xxxxxxxxx]
Sent: Monday, June 19, 2017 10:30 AM
To: David Lewis <David.Lewis@xxxxxxxxxxxx>; STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx
Subject: RE: Parameter changes for MPI penalty

 

David,

 

I think that if you want to adopt the principle of trading MPI penalty against channel insertion loss, then you should follow what was done for 100GBASE-DR in the P802.3cd draft:

 

cid:image001.jpg@01D2E9D8.A633DC30

 

cid:image002.jpg@01D2E9D8.A633DC30

 

Regards,

Pete Anslow | Senior Standards Advisor
43-51 Worship Street | London, EC2A 2DX, UK
Direct +44 2070 125535
|

 

From: David Lewis [mailto:David.Lewis@xxxxxxxxxxxx]
Sent: 19 June 2017 08:19
To: STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_25GSMF] Parameter changes for MPI penalty

 

Hi Kohichi,

I agree with what you say. 

 

Thinking about it, we have 2 columns for 25GBASE-ER:  the 30km column that has 3 dB of unallocated margin, and the 40km column that has 0 dB margin but is an “engineered link”.

 

Would you support the following:

  1. Add a note to the 3 dB margin in the 30km column that says the margin is to include allowance for an MPI penalty of up to 0.7 dB.
  2. Add a note to the 18 dB channel insertion loss in the 40 km column that says the 18 dB includes an allowance of up to 0.7 dB for MPI penalty.

 

I think that we need a comment that will not draw opposition in the recirculation ballot.   I would like to know whether the above will be supported or opposed.

 

Regards,

David Lewis

 

 

From: 田村公一 Kohichi Tamura [mailto:Kohichi.Tamura@xxxxxxxxxx]
Sent: Monday, June 19, 2017 1:46 AM
To: STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_25GSMF] Parameter changes for MPI penalty

 

Hi David,

For 25GBASE-ER, my feeling is that we’ve already put the sensitivity as low as it ought to be for APD receivers. The transmitter OMA is also on the high side. So, to budget for MPI, I’d prefer to trade off a small amount of channel loss, leaving the powers as they are. One would expect use cases requiring the full loss budget to be rare, so it doesn’t seem necessary to burden the transmitter or receiver specifications further, unless justification is provided from channel data.

Regards,

Kohichi

 

From: David Lewis [mailto:David.Lewis@xxxxxxxxxxxx]
Sent: Sunday, June 18, 2017 8:14 PM
To:
STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_25GSMF] Parameter changes for MPI penalty

 

Mike,

That’s my initial proposal.  However if enough people object to the 0.7 dB reduction in sensitivity & SRS, we’ll need to introduce a loss table similar to the one in 100GBASE-DR.

 

During the ad hoc call there was opposition to the idea of giving users a tradeoff table for insertion loss for the 10km application.  The point was that some users would want to operate over existing cable plant currently operating at 10 Gb/s.  If we were to accept what I proposed, then those users would still need to check on how many connectors and the RL of each one, before knowing they could use it for 25 Gb/s.  At least they would not need to measure channel loss before knowing they could re-use a 10 Gb/s cable.

 

David Lewis

 

From: Dudek, Mike [mailto:Mike.Dudek@xxxxxxxxxx]
Sent: Saturday, June 17, 2017 3:14 AM
To: David Lewis <David.Lewis@xxxxxxxxxxxx>; STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx
Subject: RE: Parameter changes for MPI penalty

 

Is your proposal to not make any allowed trade offs between number of connectors and link loss?   If so I think the numbers in the table on slide 6 would be replaced by ticks.   I do wonder with such a large difference between TDP and the allowance for penalties whether we should give people a clue what that is for.   Maybe split the row into two.   One that says “Allocation for TDP” (value of 2.7dB) and the other says “Allocation for other penalties including Multi Path Interference” (value of 0.7dB).

 

From: David Lewis [mailto:David.Lewis@xxxxxxxxxxxx]
Sent: Friday, June 16, 2017 5:17 PM
To: STDS-802-3-25GSMF@xxxxxxxxxxxxxxxxx
Subject: [802.3_25GSMF] Parameter changes for MPI penalty

 

All,

 

At the ad hoc on Wednesday I took an action to provide the changes that would be needed in the optical link parameter tables for an MPI penalty of 0.7 dB.  See the attached presentation for what I think is needed.

 

The approach is to tighten Rx sensitivity and SRS by 0.7 dB for both the -LR and -ER variants.  In addition this proposal tightens the Transmitter reflectance from a maximum of -12 dB to a maximum of -26 dB.

 

Please use the reflector for discussion of this topic.  Our next ad hoc is on Wed June 28th and will be the only one between now and when the sponsor recirculation ballot ends.

 

Regards,

David Lewis

P802.3cc Task Force Chair