Channel Ad Hoc telecon minutes Octiber 12, 2005 802.3ap task force interim meeting in Vancouver 11/13-14 and 11/15-18 802.3 meeting in Vancouver 11/13-18 old business - comment 105 progress, resove 128, resolve 300 new business - Annex 69A, comments 259, 262, 578, 259, 299 Charles - I'm doing experiments with a real channel (Tyco channels) and real aggressor along with a sinusoidal aggressor. Rich - This is John's data and equations 23-25 need a log term. On the last page, a linear plot is shown with a log fit. Looking at a log plot is preferable to me. John - The proposed verbage is clear, should we change it? Rich - Doesn't matter to me. Charles - Rich says he wants it on a log scale. I did both plots for comparison. moore_c1_1005 shows various fitting techniques of the ICR data. These are identical to Rich's with... On page 8 I have equations that specify the range instead of a sum. The verbage specifies the range but not the sum. John - Under 69.20 it specifies the range... Charles - ..but not where you do the fitting. John - good catch Charles - Does anyone feel uncomfortable with the verbage on page 8? (no response) Accepted by voice vote without objection. I'll send a .doc to Shelto for the text. Charles - comments to 69.3...CDR stress topics... Adam - paraphrasing comment 259, ...regarding the "compliant TX" in the RX int tol test, you need to specify a worst case value for amplitude, jitter (sine modulation), and range, resolution of the TX equalizer (needs to be constrained). Charles - I'd like to treat 259, 262, 578 as a group. Adam - Rob isn't here, but this is just an extension of 259 Adam - why did you leave out 299 from that list? Charles - 299 is an amplitude topic, but we can add it to this group. Anyone wish to comment? Rich - Joe and Adam's seem to oppose. It would be easy to compare each characteristic at a plugfest. Adam - comment 299 would hurt interoperability and mission mode, so I am against it. Amplitude adjustment and jitter is available, but the signal degradation from a mistuned TX FIR is not. Charles - The BERT can't do de-emphasis unless you get a ParBERT. Rich - does that rule out the test equipment Adam - you may need to tweak the test, but it should be affordable using equipment Rich - Adding jitter stresses phase locking. Is there a correlation between noise injection and jitter. Charles - My CDR is most likely different than someone elses. Original thought was increase the aggressor amplitude, but now I'm including backplane aggressor signals and sine wave aggressor. Adam - Specialized test equipment is a problem. You have shown you can create an instrument grade source, and that's is what is required. I have not seen the link between the sine aggressor and channel jitter. The notion of picking up an off the shelf TX is ludicrous since it's not repeatable or transferrable. Charles - You are recommending we have an instrument TX that has the characteristics of the "compliant TX". Ali's comment 578 shows some numbers. Adam - We need jitter tracking and high freq jitter tolerance. I can't say how much high freq jitter to add. Charles - Ali has the most complete approach at this point. Let's take his proposal, and use a piece of test equipment to generate it. Then write it up as a proposed response to the comment 578 and 262, and a partial answer to 259. Is that a good enough starting point? Adam - let's say this is a starting point for the next 3 weeks Rich - what equipment is available? Charles - We have seen useful results with a two tap Tx equalizer. I'll write up a proposed answer to 578. Adam - I'll take an action to find out what Rob was after in 262. In CX4 asynchronous aggressors were used to cover the full frequency range of coupling. Charles - If we are using test equipment, I would propose we write an amplitude spec that is close to 800mV, but ~900mV would be preferred. Adam - I'd rather have it be 800mV since it traces back to the TX spec. Charles - Would you like to see single verbage for all the comments on this topic? Adam - To me, the sticking point is what to do with the TX FIR... Charles - How to specify the granularity of the TX equalization. Adam - Questions to ask....Can we get a BERT with equalization? Can we rig one up? Can we design the channel stress so that we can leave out the FIR stress requirement. Charles - I've seen an RX that doesn't need TX equalization. Adam - I'd like to fully exercise all these options. Charles - Amplitude issue is clear, but tap issue is muddy. Rich - Isn't this an RX test, so isn't it just a uniform lossy transmission line issue. Adam - Didn't you claim that the interference spans the distortion of the connectors. Charles - Yes, but different RX need different TX equalization. Adam - which is originally why we proposed a feedback loop. I just want to make sure we have proper coverage. Charles - a usable RX will have to work with channels that are not precisely equalized with the TX. In the presense of crosstalk, there is tradeoff necessary to require the RX to be tolerant, and the spec for tap weight to be somewhat loosely spec'd as well. Sounds like we have enough for a proposal to spec amplitude and jitter verbage for the "compliant TX" in the RX int tol test. I'll provide that verbage. Charles - Adam, I think the range of the Tx equalizer is defined in the draft. It seems only the resolution needs to be spec'd. Adam - resolution should also be constrained by the spec. Charles - the spec has the max and min step size. Should we require the RX to operate with the equalizer with a misadjusted amount. This is a way to test the receiver. Adam - if we allow the TX to have a certain granularity, it seems prudent to test the RX with an "unoptomized tap setting" Charles - I'll write up a proposal for 259, 262 and 578 with this wording. Brian Brunn - How will you define an optomized setting, then a de-optomized setting. Adam - The tester will try to make the device pass. I may not be able to be optimize with a TX that doesn't have the steps you need. Brian - How about saying it must satisfy multiple settings, for example two settings. Adam - We want to test the RX with a perturbed settingt that means a few settings different than what is optimal with an initialization feedback path. Charles - Let's finish up here, adn meet again on 11/2.