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

RE: XAUI channel group delay




Hi Dawson

I don't quite understand this. 

To me it seems reasonable that the maximum pp group delay MUST be less that
1 bit period (ie. 320 ps for a 3.125 Gb/s). At this limit the ISI will be
severe. 

The impact of the ISI is dependent on the energy that leaks into the
adjacent bits periods. A flat group delay corresponds to a linear phase,
which means that all frequency components of the signal arrive at the far
end simultaneously, and thereby all the signal energy stayes within its bit
period. 

This means that large group delay variation may be allowed if the signal
energy is sufficiently low in that frequency range.

So, strictly speaking, one may allow larger peaks in the group delay
provided that the integrated signal energy over the particular frequency
range around that peak, is less than a few procent of the total signal
energy. This may be what is included by specifying the span (but this
assusmes a flat signal spectrum).

The ISI (eye closure method) may give a more correct picture of the impact
from group delay variations, as this is a time domain measurement (where the
integration of the signal energy convolved with the group delay) and
actually measures directly the signal that the receiver will see. The group
delay is only a handy way to look at the behavior in the frequency domain
but may not give the correct picture in the time domain.

One has to consider this convolution with the signal spectrum and be more
specific about the definitions. 


Benny
--------------------------------------------
  llllllll   ii     llllllll     llllll
ll                ll           ll      ll
ll    llll   ll   ll    llll   llllllllll
ll      ll   ll   ll      ll   ll      ll
  llllllll   ll     llllllll   ll      ll

GIGA, an Intel company
Benny Christensen, M.Sc.E.E, Ph.D.
Mileparken 22, DK-2740 Skovlunde, Denmark
Tel: +45 7010 1062, Fax: +45 7010 1063
e-mail: benny.christensen@intel.com, http://www.giga.dk








> -----Original Message-----
> From: Kesling, Dawson W [mailto:dawson.w.kesling@intel.com]
> Sent: 21. februar 2001 04:06
> To: Serial PMD reflector (E-mail)
> Subject: XAUI channel group delay
> 
> 
> The XAUI compliance channel group delay limit of 160 ps p-p 
> (adopted in
> Irvine and included in D2.1) appears to be impractically 
> restrictive. A plot
> of several compliance channels is attached to illustrate. The 
> best channel
> measures 628 ps p-p. The best channel having connectors 
> measures 1425 ps
> p-p! (Apertures close to the 3% of span allowed by the draft 
> were used.) 
> 
> I see a few options to fix this in the next draft:
> 
> 1. Loosen the limit to about 2 ns p-p to accommodate valid 
> channel data. But
> such a loose limit will have little value in preventing manipulated
> channels, which was the original purpose of the limit.
> 
> 2. Develop a more complex limit to make distinction between 
> valid channels
> and inlikely ones. This will take a committed person or group 
> of persons to
> analyze the problem, gather more data and propose the 
> solution. It also
> complicates compliance testing. The benefit in preventing manipulated
> channels is not worth this effort in my opinion, particularly 
> in view of
> likely interoperability tests and other practical 
> requirements that will
> exist for XAUI component performance.
> 
> 3. Eliminate the group delay limit. Given the complexity of preventing
> manipulation and the doubtful benefit of testing with a 
> manipulated channel,
> I no longer see a need for this limit. I propose we remove it.
> 
> Thoughts?
> -Dawson
> 
>  <<gd.ZIP>>  <<mag.ZIP>> 
>