Mike There are several errors I have seen in SPICE descriptions of I/O buffers. Missing clamps is just one example. More important are the basic issues with correlation between the IV curves that SPICE predicts and what we have actually measured. I have downloaded IBIS models from manufacturer's web pages. I have compared those IV tables with measurements I have made on sample ICs with lot date codes near the published dates of those IBIS models. Sometimes there is good agreement. In many cases I've seen room temperature, nominal voltage IV measured curves that fall outside the limits of the Min/Max of the SPICE extraction. Typically the measured curves were stronger than the max in the SPICE derived model. There are two ways that Teraspeed Labs has made min/typ/max IBIS models from measurement. One is taking a sample in characterizing it over temperature and voltage and stating in the model that the min and max represent voltage and temperature only. Over the years we have seen many die from many different lots of the same manufacturing lots. We have noticed that there is very good consistency over time of the performance of the I/O buffers unless there was an obvious process change. Modern IC process tend to be very consistent. I've also had IC vendors or their fabless customers tell me that there is not much variation between lots. The other way we have made Min/typ/max models is to get process skew ICs. We measure different process corners to get the different IBIS corners. This is easy to get corner lots for a CMOS process if they want to spend the time but difficult for the bipolar processes. Tom Dagostino Teraspeed Consulting Group LLC Device Modeling Division 13610 SW Harness Lane Beaverton, OR 97008 503-430-1065 tom@xxxxxxxxxxxxx <mailto:tom@xxxxxxxxxxxxx> Teraspeed Consulting Group LLC 2926 SE Yamhill St. Portland, OR 97214 http://www.teraspeed.com -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of MikonCons@xxxxxxx Sent: Monday, March 22, 2004 9:34 AM To: si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Tools to Create IBIS Models In a message dated 3/21/2004 10:37:43 AM Pacific Standard Time, scott@xxxxxxxxxxxxx writes: If a Spice model is created with impropper assumptions or by leaving out important elements of the circuits (like ESD structures) it will not be very accurate. IBIS models, on the other hand, can be created directly from measurement and are as accurate as the measurements we take. Teraspeed Labs has measurement based IBIS models that accurately reflect the device operation. With full process corner silicon, we have developed models which are more accurate than the original Spice models (as reported by our customers, the IC vendors themselves.) ************** Scott: You are right on target re: ESD structures. I have found the clamping networks (or lack of inclusion of their affects in the IBIS models) are a very common source of error. It is these very networks that contribute the maximum variation of driver and receiver output/input capacitance as a function of signal voltage, which in turn alter the rising and falling edges of the signals. That is one of the key elements in my paper. Some of your comments raised my eyebrows a bit..."IBIS models, on the other hand, can be created directly from measurement and are as accurate as the measurements we take. Teraspeed Labs has measurement based IBIS models that accurately reflect the device operation....and..." If you want to know if your IBIS model is accurate, do the same, or create it directly from measurements, as we do." I have seen and appreciated your modeling work (which IMHO was reflected in one of the best papers at DesignCon 2004), but I am puzzled as to how one would extrapolate lab measurements to models that correctly reflect/predict the corners of silicon processes. I would think that any reasonable number of measurements on real devices would still fall within the one-sigma range of performance. I have used model parameter extrapolations (based on historically documented process tolerances supplied by cooperative silicon vendors) for very critical modeling applications, but there is always some inherent error risk in those extrapolations. I also found that my clients were initially reluctant to add compensating design margins (with its attendant cost impact) when such extrapolations were not confirmed by room ambient lab tests. (But they ultimately agreed because of the criticality of their application.) Respectfully, Mike Michael L. Conn Owner/Principal Consultant Mikon Consulting *** Serving Your Needs with Technical Excellence *** ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List FAQ wiki page is located at: http://si-list.org/wiki/wiki.pl List technical documents are available at: http://www.si-list.org List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu ------------------------------------------------------------------ To unsubscribe from si-list: si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field or to administer your membership from a web page, go to: //www.freelists.org/webpage/si-list For help: si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field List FAQ wiki page is located at: http://si-list.org/wiki/wiki.pl List technical documents are available at: http://www.si-list.org List archives are viewable at: //www.freelists.org/archives/si-list or at our remote archives: http://groups.yahoo.com/group/si-list/messages Old (prior to June 6, 2001) list archives are viewable at: http://www.qsl.net/wb6tpu