David
I would not say ubiquitous. Single purpose driven silicon may very well
have dc blocking capacitors on-die, but most general purpose silicon IP do
not. For example, any silicon capable of both 100G BaseKR and PCIE must
necessarily not have on-die receiver capacitors, as that would be in
violation of the PCIe specification.
A device created for module applications alone may very well have
capacitors built into the Tx and Rx I/Os, as module specifications require
the capacitors be be carried on module.
However, Walter's point is quite aptly stated. AMI modeling does not
generally care about the DC solution. A DC short will work quite well.
The error at 50 MHz for a 100 nF capacitance, when modeled as a simple
short, is 0.3%. That is essentially negligible. This is lower than all
other systemic modeling errors in the system.
As a point of reference, most serdes systems are quite sufficiently modeled
with s-parameters that have 50 MHz step size. The exception to this are
very long low loss cable systems, where the step size needs to be reduce to
10 MHz, to avoid time domain alising.
regards
Scott
Scott McMorrow
R&D Consultant
Teraspeed Consulting - A Division of Samtec
16 Stormy Brook Rd
Falmouth, ME 04105
(401) 284-1827 Business
http://www.teraspeed.com
On Sun, Apr 3, 2016 at 11:09 AM, David Banas <capn.freako@xxxxxxxxx> wrote:
Hi Walter,
Thanks for the response.
Please, see below, in-line.
Thanks,
-db
On Apr 3, 2016, at 7:51 AM, Walter Katz <wkatz@xxxxxxxxxx> wrote:
David,Yes, that’s correct. I’m talking about a series capacitor connected
I assume you are talking about an on-die blocking cap, as opposed to a
blocking cap in the package or board interconnect.
between the pad and buffer input, with the assumption that the effect of
the [GND Clamp] data is applied at the buffer input. Said differently: a
capacitor, which will serve to isolate the pad from any d.c. bias in the
[GND Clamp] data.
1.) I don’t want to invoke S-parameters, as I’m comfortable with a lumped
For an on-die blocking cap, one can to do this is an on-die s-parameter,
or just assume it does not exists since a blocking cap is a short at
frequencies of interest in AMI modeling.
element representation of this component.
2.) Won’t omitting it expose me to potential channel mis-characterization,
in those tools, which attempt an explicit d.c. solution?
either
The new interconnect BIRD will allow you to define blocking caps in
the on-die interconnect, or package interconnect.Okay, thanks. I’ll familiarize myself with that effort. However, this is
something that is present in nearly all Rx silicon. So, I’m a little
worried that I’ve confused the issue. (Perhaps, I’m not asking my question
very clearly?) I guess I was expecting a one-liner, like:
C_block 100nF NA NA
would be possible, given the ubiquitous nature of d.c. blockers in Rx
silicon.
If you’re talking about tool-specific pseudo-comment style augmentation,
I assume that there are other EDA tools besides ours that allow the user
to define package and on-die subckts that are either s-parameter or SPICE
(IBIS-ISS) subckts that contain blocking caps.
I’d like to try to avoid that, in order to avoid having to maintain several
different versions of the model. (I would be really dismayed, if I had to
resort to this just to get a d.c. blocker in place, since it’s such a
simple and ubiquitous component in receivers.)
I don’t think so; at least, not completely. But, I do appreciate the
Does this answer your question?
reply! :)
I suspect that (once again) I’m just not expressing my desire very
clearly. :(
Walter
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of David Banas
Sent: Sunday, April 3, 2016 10:14 AM
To: Bob Ross <bob@xxxxxxxxxxxxxxxxx>
Cc: ibis-macro@xxxxxxxxxxxxx
Subject: [ibis-macro] Re: Terminator still illegal for AMI?
Thanks, Bob!
What is, currently, the preferred way of representing a d.c. blocking
capacitor in the signal path of a Rx model, assuming I'm okay with a
lumped element representation?
Thanks,
-db
On Apr 3, 2016, at 7:04 AM, Bob Ross <bob@xxxxxxxxxxxxxxxxx> wrote:
Hi David,a
Yes, Terminator is illegal. According to the [Algorithmic Model]/[End
Algorithmic Model] Usage Rules (Section 10.1):
Usage Rules: The [Algorithmic Model] keyword must be positioned within
[Model] section and it may appear only once for each [Model] keyword
in a .ibs file. It is not permitted under the [Submodel] keyword or
in [Model]s which are of Model_type Terminator, Series or Series_switch.
Bob
-----Original Message-----
From: ibis-macro-bounce@xxxxxxxxxxxxx
[mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of David Banas
Sent: Sunday, April 03, 2016 6:38 AM
To: ibis-macro@xxxxxxxxxxxxx
Subject: [ibis-macro] Terminator still illegal for AMI?
Hi all,
Is Terminator still an illegal model type for AMI Rx models?
Thanks,
-db
