|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
As mentioned below, we have reason to believe that real TP3 testers, as well as real transmitters, have excess PIE-D penalties compared with ideal signals beyond basic characteristics such as rise/fall time. This points out again the reason we should target lower TP3 stressor targets than theoretical analysis of EDC capability might otherwise imply. That is, we may always be testing the receivers more severely than we intend and should make some reasonable allowance for that.
Attendance (no order)
· Lew Aronson
· David Cunningham
· Ernie Bergmann
· Piers Dawe
· Mike Dudek
· Majid Barazande-Pour
· Norm Swenson
· Attendance (done, see above)
· Build agenda (done)
· Technical discussions (see below)
· Next call (see below)
Mike Dudek moderated the call in Tom Lindsay's absence.
· Further discussion took place on the comparison metrics of SNR, TWDP, and implementation penalty (IP) with finite length EQ that Tom had presented last week. See http://grouper.ieee.org/groups/802/3/aq/public/upload/AnalysisofSNR_TWDPandfinitelengthIPvs.measuredwaveforms.pdf.
o Norm Swenson described work that he had performed to attempt to evaluate whether the TWDP degradations seen for practical waveforms was related to non-linearities or other factors. Early results showed that the TWDP was reduced by 0.2 to 0.5dB for best linear fits to a selection of the waveforms from Tom's presentation. He sent a graph of these early results to the reflector (based on Ewen 4.6dB pre-cursor stressor). He intends to present more results next week.
o Concern was again expressed that Tom's presentation showed that when normalized to OMSD only parts with significant pre-emphasis had lower TWDP than the supposedly worst case 47ps Gaussian pulse, and that we need to make sure that "reasonable practical" Tx's will pass the test. ie we may need to allocate some of the implementation penalty to the Tx if we use this normalization method.
· Lew Aronson suggested that practical TP3 stressed eye testers may overstress the Rx and said that he was going to be performing TWDP like tests on his tester.
o David Cunningham stated that he had measured Pie-D on some commercially available BERTS and found that they had 0.2 - 0.5dB penalty not all of which was related to rise/fall times. This indicates that making a perfect TP3 stressor will not be easy.
o Mike Dudek suggested that we will have to determine what to do when the waveforms have been matched as well as possible to the pulse shapes in the standard and the TWDP like test of the resulting waveform does not match the expected TWDP like result. He suggested that there are 3 choices with the noted advantages/disadvantages.
§ 1 Just state that test implementers need to minimize the difference and that the difference will result in conservative results. (similar to what is said about noise in the 10G 802.3ae standard for the stressed receiver sensitivity testing). Not recommended based on experience with the 10G stressed eye testers.
§ 2 Say that any difference in TWDP should be used as a correction factor for the stressed sensitivity. This is still probably conservative as it is likely to result in more distortion to equalize but with somewhat higher input power. It is however relatively easy to incorporate into the standard.
§ 3 Say that the waveshapes should be adjusted to create the required TWDP. Probably the fairest test for the receivers, but may be difficult to state in the standard how to do this.
o It was also suggested that it would be useful to have a name for the "TWDP like test" (ie test of TP3 signal with TWDP like code without the simulated fibers) to more easily differentiate this test.
o Tom will be back
o Vivek planning to present on correlation across EQ lengths (action from 6/23). (Hopefully this will take place. Vivek was not on the call to confirm).
o Norm Swenson planning to present more information on his analysis of the TWDP results from the various eyes.
o Lew Aronson planning to present information on TWDP like measurements on a TP3 stressed tester with comparison to theory.
o Requests for new sims
§ ~0/30/60 psec Gaussian with pre/post, long/short EQs, normalized with OMSD/OMA. Show SNR, TWDP, and IP. (Carried forward from 6/28 call)
§ Other IP mechanisms.
o Reminder this will be the last day for comments to rev 2.1
· Date: Tuesday, 12 July, 2005
· Time: 9:00 AM Pacific
· Duration: 1:30 max
· Number: 401-694-1515
· Access code: 421721#
Director Transceiver Engineering
4663 Nautilus Court South
Tel 303 530 3189 X7533.