Almost; I'm not sure about #3. It seems some EDA tools are assuming that the proper units are "Volts/sec.", which is not how I read the spec., as per my previous response to you. Perhaps, we could get all the EDA tool vendors to comment? -db From: Todd Westerhoff [mailto:twesterh@xxxxxxxxxx] Sent: Wednesday, June 26, 2013 1:37 PM To: David Banas Cc: ibis-macro@xxxxxxxxxxxxx Subject: RE: On impulse and step responses. Dave, Thanks for your response. From a specification standpoint, it seems to me that you're saying the following: 1. The language in the current spec describing the impulse response matrix is imprecise and should be improved 2. As we're dealing with a discrete time application, we should spell that our more precisely when describing the impulse response 3. Given (1) and (2), the impulse responses currently being generated by EDA tools and the AMI models that use them are OK as-is and do not need to be changed. Did I get that right? Todd. Todd Westerhoff VP, Software Products Signal Integrity Software Inc. * www.sisoft.com<http://www.sisoft.com> 6 Clock Tower Place * Suite 250 * Maynard, MA 01754 (978) 461-0449 x24 * twesterh@xxxxxxxxxx<mailto:twesterh@xxxxxxxxxx> "I want to live like that" -Sidewalk Prophets From: David Banas [mailto:DBanas@xxxxxxxxxx] Sent: Wednesday, June 26, 2013 1:07 PM To: Todd Westerhoff Cc: ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx> Subject: RE: On impulse and step responses. Hi Todd, There are two things being discussed, which is probably adding to the confusion. Firstly, I'm challenging Mike's belief that the discrete time equivalent to the Dirac delta is the sequence, {<sample_rate>, 0, 0, ...}. I believe the discrete time equivalent to the Dirac delta is the sequence, {1, 0, 0, ...}. I think this one has probably been exhausted, unfortunately without resolution. Secondly, the current spec. fails to name the units, which are to be assumed for the values passed into Init(), via the impulse_matrix parameter. This is the issue, which is less academic, more practical, and more worthy of the committee's time. Here is the current relevant language, excerpted from IBIS v5.1: "impulse_matrix" points to a memory location where the collection of channel voltage impulse responses, ... The algorithmic model is expected to modify the impulse responses in place by applying a filtering behavior, for example, an equalization function, if modeled in the AMI_Init function. ... (Note that my omissions of any original text are indicated by ellipses, and any emphasis is entirely mine.) Now, the language "impulse response" is vague in signal processing parlance, as it can refer to either: 1. The continuous time "impulse response function", or 2. The discrete time "unit pulse response sequence", both of which are more precise concepts. In deciding which of the two interpretations, above, to accept, one notes the use of the language, "voltage" (i.e. - NOT "volts/sec."), as a qualifying preface to the term, "impulse response." Therefore, one could defensibly argue that `2' should be assumed, since the continuous time impulse response function must have units of "Volts/sec.", as has been pointed out now numerous times in this discussion. Further confidence in choice '2' is gained, by noting that ours is a necessarily discrete time application. (It takes place entirely within the state space of a digital computer, and the interface between the model and the rest of the system is a discrete sequence of numbers.) Finally, the language, "in place by applying a filtering behavior," suggests that the model should expect to be receiving values with units most natural to direct digital filter application to the unmodified input, which would be "Volts". -db From: Todd Westerhoff [mailto:twesterh@xxxxxxxxxx] Sent: Wednesday, June 26, 2013 7:39 AM To: David Banas Cc: ibis-macro@xxxxxxxxxxxxx<mailto:ibis-macro@xxxxxxxxxxxxx> Subject: On impulse and step responses. Dave, Can I ask you to confirm your initial question on this subject? I want to make sure I understand what we're saying before we get too deep into the math. It seems to me that you've called into question the mathematics behind impulse responses being created for current IBIS-AMI models, and therefore how IBIS-AMI models must be written to process those impulse responses. If there is indeed a problem with the math, it follows that the current standard would need to be either updated or extended, depending on how the details play out. Is that what you're saying? Todd. Todd Westerhoff VP, Software Products Signal Integrity Software Inc. * www.sisoft.com<http://www.sisoft.com> 6 Clock Tower Place * Suite 250 * Maynard, MA 01754 (978) 461-0449 x24 * twesterh@xxxxxxxxxx<mailto:twesterh@xxxxxxxxxx> "I want to live like that" -Sidewalk Prophets ________________________________ Confidentiality Notice. This message may contain information that is confidential or otherwise protected from disclosure. If you are not the intended recipient, you are hereby notified that any use, disclosure, dissemination, distribution, or copying of this message, or any attachments, is strictly prohibited. If you have received this message in error, please advise the sender by reply e-mail, and delete the message and any attachments. Thank you. ________________________________ Confidentiality Notice. This message may contain information that is confidential or otherwise protected from disclosure. If you are not the intended recipient, you are hereby notified that any use, disclosure, dissemination, distribution, or copying of this message, or any attachments, is strictly prohibited. If you have received this message in error, please advise the sender by reply e-mail, and delete the message and any attachments. Thank you.