[ibis-macro] Re: Question about "impulse response"

  • From: Mike Steinberger <msteinb@xxxxxxxxxx>
  • To: ibis-macro@xxxxxxxxxxxxx
  • Date: Thu, 2 Nov 2017 09:02:16 -0500


If you want an NRZ signal with unit amplitude, the symbol voltages need to be -0.5 and +0.5. The symbol voltages can then be multiplied by the signal amplitude in the system being modeled to correctly represent the signal in that system. Choosing -1.0 and +1.0 is simply an error to begin with, unless one then multiplies by half the signal amplitude in the system being modeled.

My 2C.
Mike Steinberger

On 11/02/2017 12:27 AM, Muranyi, Arpad wrote:

Todd, Walter,

I understand that the impulse response represent the analog channel, including the Tx and Rx buffers.

I also understand that the voltage swing the Tx can drive into the channel is also captured by this

channel response, and for that reason we are not talking about “unit” *** responses.

Let me try to explain my question in different words:

First, let’s assume for the sake of this conversation that the channel impulse response is generated

by taking the derivative of the channel step response waveform. With this assumption we are

shifting the question to what is the definition of the channel’s step response. Let’s assume that

a step response is going from logic ‘0’ to logic ‘1’. The IBIS Tx buffer has a digital stimulus internal

to the simulators which does just that.

Using that stimulus, the non-inverting output of Tx goes ‘0’ to ‘1’ and the inverting output goes

‘1’ to ‘0’. However, in IBIS-AMI we are dealing with the difference waveform at the Rx input, in

which case the states would be ‘-1’ and ‘+1’. (We all know that the difference waveform is usually

twice the magnitude of the two individual output waveforms of a differential buffer). So if the

states of the difference step waveform obtained at the Rx correspond to ‘-1’ and ‘+1’, while the

“text book” definition of a step response is ‘0’ to ‘1’, than we should theoretically divide the channel’s

difference step response waveform seen at the Rx input pins by 2 in order to arrive to a “text book”

channel step response (which corresponds to the states ‘0’ and ‘1’) before we differentiate that

step response to generate the channel’s “impulse response”.

But I am afraid that this is not what all EDA vendors do when they generate the impulse response

for the AMI Tx Init function. I suspect (this needs to be verified) that most simulators just take the

channel step response difference waveform at the Rx and take the derivative of that to turn it into

a channel impulse response for the Tx AMI Init function.

This is why I wanted to find a definition in the IBIS-AMI specification for what “impulse response”

is, but I didn’t find the answer to my question. And since I don’t like customers saying things like

“but other simulators do this or that”, I would like to make sure that we, EDA vendors, are all in

agreement on this question.

This might also have an effect on how we deal with single ended channel response waveforms.

Theoretically those should not be divided by 2, but then how would the EDA tool know when is

the waveform single ended or differential in order to decide when the channel response should

be divided by 2 or not... Or, should this be decided by looking at the [Diff Pin] keyword and check

whether the [Model] is differential or not?




*From:*Todd Westerhoff [mailto:twesterh@xxxxxxxxxx]
*Sent:* Thursday, October 26, 2017 9:48 AM
*To:* wkatz@xxxxxxxxxx; Muranyi, Arpad <Arpad_Muranyi@xxxxxxxxxx>; ibis-macro@xxxxxxxxxxxxx
*Subject:* RE: [ibis-macro] Re: Question about "impulse response"

Arpad & Walter,

I think the thing that we need to be really clear about in IBIS-AMI is that the impulse response represents the analog channel, **NOT** the passive channel. I’ll admit to only having done a quick search, but the only two places where I find that point made are at the start of and the start of (both on page 177). Both of these are indirect statements; they require that you understand what the analog channel is and how it’s different from the passive channel.

So – to the original question, I would say it’s “under-defined” and could use a bit of beefing up.


*Todd Westerhoff*

VP, Semiconductor Relations

Signal Integrity Software Inc. • www.sisoft.com

6 Clock Tower Place • Suite 250 • Maynard, MA 01754

(978) 461-0449 <978-461-0449> x124  • twesterh@xxxxxxxxxx

*“*I want to live like that”

                                             -Sidewalk Prophets

*From:*ibis-macro-bounce@xxxxxxxxxxxxx <mailto:ibis-macro-bounce@xxxxxxxxxxxxx> [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] *On Behalf Of *Walter Katz
*Sent:* Wednesday, October 25, 2017 9:51 PM
*To:* Arpad_Muranyi@xxxxxxxxxx <mailto:Arpad_Muranyi@xxxxxxxxxx>; ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx>
*Subject:* [ibis-macro] Re: Question about "impulse response"




The impulse response of a channel has a clear definition in the industry. Why does IBIS need one?


Walter Katz

wkatz@xxxxxxxxxx <mailto:wkatz@xxxxxxxxxx>

978.461-0449 x 133

Mobile 303.335-6156

*From:*ibis-macro-bounce@xxxxxxxxxxxxx <mailto:ibis-macro-bounce@xxxxxxxxxxxxx> [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] *On Behalf Of *Muranyi, Arpad
*Sent:* Wednesday, October 25, 2017 9:01 PM
*To:* ibis-macro@xxxxxxxxxxxxx <mailto:ibis-macro@xxxxxxxxxxxxx>
*Subject:* [ibis-macro] Question about "impulse response"

Hello Everyone,

I would like to find out what the definition of “impulse response” is in

the AMI specification.  I thought we had something about that in the

spec, but I can’t find anything now that I am looking for it.  Am I just

not seeing the forest from the trees, or is it really not defined?

I thought we did have something along the lines of “derivative of a

step response”, but I can’t find anything…




Other related posts: