[ibis-macro] Re: Adaptive Equalization in Statistical Simulations
- From: Gregory R Edlund <gedlund@xxxxxxxxxx>
- To: ibis-macro@xxxxxxxxxxxxx
- Date: Tue, 25 Mar 2014 14:52:34 -0500
Mike, Todd, Arpad, Fangyi,
Thank you for the responses. It seems you are making the same points:
1. Statistical simulation using AMI models DOES support adaptive
equalization ONCE at the beginning of the simulation to optimize the RX to
the channel.
2. If you want the RX to adapt to the bit pattern, you need to perform a
time-domain simulation using the GetWave function.
(I like to repeat what I heard to make sure I got it right.)
The only paper I have on statistical simulation is the one Sanders, Resso,
and D'Ambrosia wrote back in 2004. Does anyone know of a more recent paper
that might highlight advances made since then?
Greg Edlund
Senior Engineer
Signal Integrity and System Timing
IBM Systems & Technology Group
3605 Hwy. 52 N Bldg 050-3
Rochester, MN 55901
From: Mike Steinberger <msteinb@xxxxxxxxxx>
To: ibis-macro@xxxxxxxxxxxxx,
Date: 03/25/2014 02:15 PM
Subject: [ibis-macro] Re: Adaptive Equalization in Statistical
Simulations
Sent by: ibis-macro-bounce@xxxxxxxxxxxxx
Greg-
Many, if not most AMI receiver models perform adaptive equalization in
statistical analysis, in the sense that they're able to identify and adopt
the optimum configuration before calculating the output impulse response
(s). I've written a number of such models myself.
The implementation of adaptive equalization in statistical analysis is
often very different from the implementation in time domain simulation or
in the actual device, however. The model developer must therefore have a
deep understanding of the algorithm in the actual device, and must be able
to apply some creativity. Fortunately, there are numerous model developers
out there with sufficient skills for the job.
Mike Steinberger
On 03/25/2014 01:09 PM, Gregory R Edlund wrote:
It is my understanding that adaptive equalization is not possible in
AMI statistical simulations. To model adaptive equalization, I need
to first run a short transient simulation, load the RX tap settings
from that simulation into the model-specific parameters, and then run
a statistical simulation.
Question: what is it about modifying the impulse response that
prevents RX adaptive equalization in statistical simulations?
Thanks in advance for your time.
Greg Edlund
Senior Engineer
Signal Integrity and System Timing
IBM Systems & Technology Group
3605 Hwy. 52 N Bldg 050-3
Rochester, MN 55901
Other related posts:
