MikeI'd agree with your first statement. However, the paragraph in question is a mess and mixes up device modeling and channel modeling without being clear about either.
As for non-LTI systems, we model them all the time. Switches, repeaters, adaptive equalized re-drivers, and re-timers are used in the systems we see. All exhibit non-linear and time-varying behavior.
Scott McMorrow Teraspeed Consulting Group LLC 121 North River Drive Narragansett, RI 02882 (401) 284-1827 Business (401) 284-1840 Fax http://www.teraspeed.com Teraspeed® is the registered service mark of Teraspeed Consulting Group LLC On 8/26/2010 12:01 PM, Mike Steinberger wrote:
It seems to me that the model is required to be LTI, but that which is being modeled is assumed to be LTI.If some day we analyze a system for which the buffer amplifiers and/or channel is not LTI (say, the vibrating cable in a disk drive), we're not going to give up. We 're going to do the best we can with an LTI model which is admittedly only an approximation of the actual system.Mike S. On 08/26/2010 10:47 AM, Scott McMorrow wrote:"The combination of the transmitter's analog back-end, the serial channel and the receiver's analog front-end *are assumed to be* linear and time invariant."Shouldn't this be changed to *are required to be* Scott McMorrow Teraspeed Consulting Group LLC 121 North River Drive Narragansett, RI 02882 (401) 284-1827 Business (401) 284-1840 Fax http://www.teraspeed.com Teraspeed® is the registered service mark of Teraspeed Consulting Group LLC On 8/26/2010 6:41 AM, Walter Katz wrote:Arpad, I suggest that the following be changed (page 139): | This proposal breaks SERDES device modeling into two parts - electrical | and algorithmic. The combination of the transmitter's analog back-end, the | serial channel and the receiver's analog front-end are assumed to be linear | and time invariant. There is no limitation that the equalization has to be | linear and time invariant. The "analog" portion of the channel is | characterized by means of an impulse response leveraging the pre-existing | IBIS standard for device models. To | This proposal breaks SERDES device modeling into two parts - electrical | and algorithmic. The combination of the transmitter's analog back-end, the | serial channel and the receiver's analog front-end are assumed to be linear | and time invariant. The algorithmic model of a Tx model represents the | signal processing that is performed on the stimulus or input to the Tx | model. This signal processing is also know as equalization or filtering. | The "analog" portion of the channel is characterized by means of an impulse | response leveraging the pre-existing IBIS standard for device models. There | is no limitation that the equalization has to be linear and time invariant. | If the equalization can be represented as linear and time invariant then it | can be represented mathematically as an impulse response of the filter. I would then change "convolved with the Rx filter" to "convolved with the impulse response of the filter". Walter Walter Katz 303.449-2308 Mobile 303.883-2120 wkatz@xxxxxxxxxx www.sisoft.com -----Original Message----- From:ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx]On Behalf Of Muranyi, Arpad Sent: Monday, August 16, 2010 2:48 PM To: IBIS-ATM Subject: [ibis-macro] Re: AMI Flow BIRD - Editorial Comments. Hello everyone, I read through the latest editorial version from Ken, and I made a few of my own, mostly formatting or grammar (see attached file). However, I have a few bigger comments for which I didn't make any changes. 1) "Step 6b" says "convolved with the Rx filter". This bothers me because we do not say where the "Rx filter" could possibly come from. A novice user or model maker my pull their hair out looking for something in the flow or model definitions to find where they can get the Rx filter from. I think we should give a few hints on how this can be obtained and not let the reader figure it out on their own. 2) "Step 7" says "The output of step 6 becomes the simulation waveform output at | the Rx decision point, and optionally also returns clock ticks..." This is bad English. It is not clear at all what is it that returns the clock ticks... This sentence should be rewritten. 3) Note #1 after Step 6b says: "One possible option is that the user chooses not to employ the Rx optimization functionality...". Could someone please explain to me how the user can do this, when the optimizer in the Rx Init function cannot be turned off if the model maker does not provide a Boolean for that? I think the user simply does not have a choice in that case. We should make a suggestion that for this case the model maker should provide a Model Specific parameter that allows the user to turn the optimization on/off. But since this combination depends in what the Tx does, and the Tx and Rx models could come from two different vendors, it may make more sense to say that in case the Rx has an optimizer, the Rx model should have (strong should) a switch to turn it on or off. 4) Note #2 says that there is a problem case when Tx GetWave_Exists = true Tx Init_Returns_Impulse = true and Rx performs optimization. Then it goes on and says "Tx AMI_GetWave function | provides redundant Tx equalization, resulting in a "double counting" of the | Tx equalization effects. One option to address this... " but doesn't explain why this redundancy is a problem. If I am not mistaken, this redundancy is only a problem when Rx GetWave does not exist, because in that case the tool has to convolve the output from Step 5 with h_REI(t) which is not available directly. But note that this is only necessary when Rx GetWave is not there, so the list of the conditions at the beginning of this note should include Tx GetWave_Exists = false (Actually the real reason in this problem case is not revolving around Rx performing optimization, but the missing Rx GetWave)... 5) The second Note also says: "One option to address this is that the user not | utilize the Tx AMI_GetWave functionality in this case, and set the Tx | GetWave_Exists to False for time domain simulations. " Could someone please explain to me how a user can do this? Some tools may not allow the user to edit the Reserved Info parameters for the EDA tool. Then what? Edit the .ami file? What if the tool performs a check in the DLL and detects the GetWAve function and the discrepancy between the DLL and the .ami file? What should the tool do then? Stop with an error, or continue based on the DLL, or continue based on the .ami file? We must address this in the specification, or not make such suggestions. (Keep in mind the tool will also need to make different selections for its convolutions based on what goes on in this area). 6) The next sentence in the second note says: "Another option is to | use deconvolution to be able to properly combine the Rx filtering from the | Rx AMI_Init function with the output of the Tx AMI_GetWave function and the | channel itself." Who is this sentence talking about? The model maker, the model user, or the EDA tool vendor? (I just happen to know the answer, but is this obvious to a novice reader)? This is another example for writing a sentence that sounds really smart, but doesn't say anything practically useful to the reader. Thanks, Arpad ==================================================================== -----Original Message----- From:ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Ken Willis Sent: Wednesday, August 11, 2010 10:22 AM To:bob@xxxxxxxxxxxxx Cc:wkatz@xxxxxxxxxx; 'IBIS-ATM' Subject: [ibis-macro] Re: AMI Flow BIRD - Editorial Comments. Hi Bob, Thanks for the feedback. I have taken care of these items in the attached update. Thanks, Ken Willis Sigrity, Inc. 860-871-7070 kwillis@xxxxxxxxxxx -----Original Message----- From:ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Bob Ross Sent: Tuesday, August 10, 2010 5:51 PM To:kwillis@xxxxxxxxxxx Cc:wkatz@xxxxxxxxxx; 'IBIS-ATM' Subject: [ibis-macro] Re: AMI Flow BIRD - Editorial Comments. Ken: Throughout, use Rx and Tx for consistency (not RX and TX) Section 2.2, Step 10 and Section 3.2, Steps 6, and 6a GetWave --> AMI_GetWave All "i.e." should be followed by a comma. Bob Ken Willis wrote:Hi everyone, I have done the spell check, verified that the function and parameter names are consistent and correct, and put in the | characters wheretheywere missing. Attached is this latest version for everyone?s review. Thanks, Ken Willis Sigrity, Inc. 860-871-7070 kwillis@xxxxxxxxxxx <mailto:kwillis@xxxxxxxxxxx>------------------------------------------------------------------------From:ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] On Behalf Of Walter Katz Sent: Tuesday, August 10, 2010 4:12 PM To: IBIS-ATM Subject: [ibis-macro] AMI Flow BIRD All, AMI Flow BIRD ready for final review. Walter Walter Katz 303.449-2308 Mobile 720.333-1107 wkatz@xxxxxxxxxx www.sisoft.com-- Bob Ross Teraspeed Consulting Group LLC Teraspeed Labs 121 North River Drive 13610 SW Harness Lane Narragansett, RI 02882 Beaverton, OR 97008 401-284-1827 503-430-1065 http://www.teraspeed.com 503-246-8048 Direct bob@xxxxxxxxxxxxx Teraspeed is a registered service mark of Teraspeed Consulting Group LLC --------------------------------------------------------------------- IBIS Macro website :http://www.eda.org/pub/ibis/macromodel_wip/ IBIS Macro reflector://www.freelists.org/list/ibis-macro To unsubscribe send an email: To:ibis-macro-request@xxxxxxxxxxxxx Subject: unsubscribe --------------------------------------------------------------------- IBIS Macro website :http://www.eda.org/pub/ibis/macromodel_wip/ IBIS Macro reflector://www.freelists.org/list/ibis-macro To unsubscribe send an email: To:ibis-macro-request@xxxxxxxxxxxxx Subject: unsubscribe