The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> High-Speed I/O Design >> group delay of a backplane channel
https://designers-guide.org/forum/YaBB.pl?num=1355205469

Message started by ywguo on Dec 10^th, 2012, 9:57pm

Title: group delay of a backplane channel
Post by ywguo on Dec 10^th, 2012, 9:57pm

Hi Guys,

I study the group delay of channel. The ac analysis gives the following the amplitude loss and group delay. The group delay has sharp increase at ~500kHz and sharp decrease near 700MHz.

What causes the group delay looks like that shape? What is the effect on jitter/eye if the datarate is below 600Mbps and above 1Gbps respectively?

Any comments are appreciated.
Yawei

Title: Re: group delay of a backplane channel
Post by raja.cedt on Dec 11^th, 2012, 12:31am

hello Yawei,
-ve group delay is little bit new to me, i know for some filters it will be fine, but i haven't seen any real channels with this behavior. Please double check once. Double check with s parameters.

Regarding your group delay impact on link performance, normally if ripple in the group delay is less than 10% of bit period then there is not much impact (according to sakinger book), in reality if you limit less than 6% it will be really good and you don't see any data dependent jitter. In your plot ripple is around 1ps, which is acceptable even at 5Gbps (here tb=200ps, so 10% ripple 20ps ).

Thanks,
Raj.

Title: Re: group delay of a backplane channel
Post by loose-electron on Jan 16^th, 2013, 8:37pm

Delay in a backplane will not model nicely as a linear model, with a simple set of parameters.
To get something close to reality you need to use a distributed parameter model.

You care to show us the model that you are using?

Title: Re: group delay of a backplane channel
Post by andyseu on Sep 25^th, 2013, 1:28am

hi raja

i have a question , you said " In your plot ripple is around 1ps". but,according to the plot, the ripple is about 1ns. why you said the ripple is 1ps. why have 10e-3 difference. i have saw some paper, all have 10e-3 difference. i can't understand , or can you offer me some document which i can get the answer from . thank you

raja.cedt wrote on Dec 11^th, 2012, 12:31am:

Title: Re: group delay of a backplane channel
Post by raja.cedt on Sep 27^th, 2013, 6:03am

hello,
it's a typo, sorry for that. How cum some other documents are doing the same mistake? Please forward me those docs. Regarding -ve group delay, group delay will be the physical delay only when you when you have constant magnitude around that frequency spectrum.

Thanks,
Raj.

Title: Re: group delay of a backplane channel
Post by andyseu on Sep 28^th, 2013, 6:23pm

dear
"The group delay variation for the 100cm V cable is 435 ps (which corresponds to 230 GHz when using the 10% rule) up to 50GHz hence there is no need for phase equalization at data rates below 50 GHz", why 435ps corresponts to 230GHz, i think 435ps shoud correspont to 0.23GHz(1/435ps*0.1). it have 10e-3 difference. the document is in the wabside"http://www.eecg.utoronto.ca/~sorinv/theses/andreea-balteanu_MASc_thesis.pdf", page 9 .
thank you

raja.cedt wrote on Sep 27^th, 2013, 6:03am: