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

Re: [10GMMF] TP3 call Meeting Notes - Dec 7

Attached are some illustrations and suggestions related to the IPR sorting discussion today.  Overall, I suggest keeping all IPRs beyond a certain value of the DC group delay - peak metric for the pre and post cursors.  Whatever the conclusion, I think it is important to focus on looking at the fitting results for these bins.


-----Original Message-----
From: Lew Aronson 
Sent: Tuesday, December 14, 2004 12:37 AM
Subject: Re: [10GMMF] TP3 call Meeting Notes - Dec 7


So glad to see that you've taken off with the hard work of this process. It certainly seems like this mechanism gives IPRs with the right characteristics in each bin.

A couple of questions in advance of tomorrow's call (and for others to consider for 3 and 4).

1) Why use +20% for your metric for the postcursor sorting but -15% for pre-cursor.  Why different values? Clearly the two cases have some differences in their range of shapes, but was there a particular reason for different numbers? (i.e. to generate the same number in the bin?)

As to the differences in shapes, I found it useful to compare them on the same sheet, matching the time scales (they are different in the different plots).  It is also interesting to compare the post-cursors with the mirrored pre-cursors, again with the time scales the same.  I have attached this comparison.

2) It would be very ineteresting to know the numerical breakdown of this sorting:

a) How many IPRs in the compelte model (5000?)

b) How many in the 4.5 +/-0.5 dB bin? (this is fairly course isn't it, perhaps we want a narrower bin, but I guess we can sort out the different shape bins by narrower PIE windows easily at that stage)

c) Of the number in b), how many IPRs went into each of the pre-cursor, post-cursor and symmetric bins, and how many (if any) were left over / discarded?

3) Thoughts on next step? Fitting with different dT's to find optimum

4) Do any new ideas about the final choice selection from looking at the nature of these results.  My suggestion had been to choose those that essentially were best represented by the parameters we would put in the RX table.  That is, assuming as I did 4 peaks at fixed dT spacing, choosing those IPRs which have the best fit with that basis set.  However, perhaps some other characteristic jumps out as one which would be better or could be used along with the above to get to a final selection.

-----Original Message-----
From: Sudeep Bhoja [mailto:sbhoja@BIGBEARNETWORKS.COM]
Sent: Monday, December 13, 2004 10:59 PM
Subject: Re: [10GMMF] TP3 call Meeting Notes - Dec 7

Please see attached slides on  the TP3 stressed test
for discussion on the call tommorow.
Best Regards,

Feedback on Bhoja ISI Cases 12-14-04.pdf