Chris, Two quick comments. 1) I don't think I was bad mouthing my previous employer, at least I didn't intend to. What I wrote were facts, and as a result of our SI correlation work, improvements were also made to the SPICE models. 2) The rest of your writing reminds me to another saying: "You can lead to horse to the water, but you can't force it to drink". I am not saying by any means that (legacy) IBIS models are better than SPICE models in all respects, and I agree those arguments were wasted time. We can't compare apples with oranges, and we should always try to use a model that was made for the purpose that we want to use it for. This even goes within the SPICE world, you would most likely not want to use a level 3 MOSFET model for nm transistors with all kinds of leakage currents and other fancy effects which were negligible in um transistor sizes... The interesting thing is that the *-AMS extensions were available in the same year you sited for people having SSO modeling/simulation problems, 2004, yet no one jumped on these powerful languages to write better behavioral models. If you could do it 10 years ago, they certainly could have done it 4 years ago... Now, writing better models (and EDA tools) takes more effort and time =3D money. I was never able to convince my colleagues to write SPICE netlists without using .global statements, because that would take them some extra effort through the maze of the subcircuits of a buffer. As so many others commented along these lines, as long as there are no catastrophic failures which can be traced back to something specific in the model, people are not going to change it. This is governed by the basic laws of human laziness and greed. The problem is that most of the time things are not black and white, don't fail due one obvious reason, etc... So, how do we get people to write better models? Arpad --------------------------------------------------------------------- -----Original Message----- From: Chris Cheng [mailto:Chris.Cheng@xxxxxxxx]=20 Sent: Tuesday, April 01, 2008 2:05 PM To: Lance Wang; Muranyi, Arpad; si-list@xxxxxxxxxxxxx Subject: RE: [SI-LIST] Re: Do you ever measure any of the things you = model? How appropriated that this thread comes out at the 10th year anniversary = of me shipping both HSPICE and behavioral models to customers that has = SSO, packaging and a whole lot of bells and whistles effects modeled. = Not only they correlated to each other, they even correlate to lab = measurements too, under SSO condition. I personally worked with at least = two of the largest SI tool houses at that time to make sure they work. = And amazingly this is from the same company Arpad was bad mouthing. It = reminds me of this old Chinese saying,"one kind of rice can raise = hundred kinds of different men". So goes employees within the same = company. It really matters what corner of the building your office is = located. Or geographic location for that matter. And how ironic that Lee raise that EE time article about SSO noise = problems in packages in 2004 in the PDS discussion. Let me say again = 2004, four years ago. If you dig up that article, people were whining = about SSO problems. And the vendor was saying........surprise, = surprise......"we give out IBIS models to our customers for = verification"...... Ok, where is the SSO model then ? Where is the SSO = model now ? I am beyond arguing about HSPICE or IBIS is better and why. We've done = too many times and wasted too many opportunities to get it right = already. I am just asking who is ready to give something, ANYTHING, that = will even address the basic problem we all face everyday. I see this IBIS vs. SPICE has a huge collateral damage....bad designs = that did not get checked and have poor SI performance. This unintended = consequence came from the fact that while we are hot headedly screaming = why SPICE is more accurate or IBIS is faster and just as good, we let = the supplier of the models get away with the lowest possible = denominator...IBIS 3.2 or below simple model that is useless to model a = large segment of the highspeed buses, memory & FSB to name a few. Ten years ago I have already demonstrated to my customer I can be = neutral to behavioral or SPICE based models. If you want SPICE, we can = give you models that correlates with lab measurements. If you want = behavioral, we can give you models that correlates with lab measurements = too. I DON'T CARE, I can make them both work. The fact that 90% of my = customers want SPICE is another topic to discuss. So, once again, we have the tool, we have the model specs.=20 WHERE IS THE MODEL ? -----Original Message----- From: Lance Wang [mailto:lwang@xxxxxxxxxx] Sent: Tuesday, April 01, 2008 10:47 AM To: Chris Cheng; 'Muranyi, Arpad'; si-list@xxxxxxxxxxxxx Subject: RE: [SI-LIST] Re: Do you ever measure any of the things you model? Hi, Chris, I believe Arpad wants to express that IBIS is working on them. However, = the sad thing is no model out there yet as you point out.=20 There is no simple modeling technique for advanced models (such as SSO, etc.). I think the key, getting a good modeling standard to fly and = solve the real design issues, is to have the tools. Imaging how many Spice transistor-level models will be there if there is no tools could create = them by clicking buttons. This is the same to IBIS as well... We are not = quite there yet. My two cents, Lance Wang IO Methodology Inc. 978-764-2298 FREE!!!=20 SignalMeth(TM) - A Powerful Waveform Post-Processing Tool=20 Goto www.iometh.com for details -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] = On Behalf Of Chris Cheng Sent: Monday, March 31, 2008 10:30 PM To: Muranyi, Arpad; si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Do you ever measure any of the things you model? I know the fancy BIRD and new specs.=3D20 I am asking who is stepping up to the plate and deliver them to =3D customers.=3D20 The last time we talked about IBIS with SSO, it was many many many years = =3D ago. The spec and BIRD were out then too, yet I have heard of no one =3D delivering model like that. In the mean time, DDR1 is gone and DDR2 is = =3D on its way to become DDR3, am I the only one seeing SSO problems in them = =3D ? Or everyone just look the other way and claim perfect correlation with = =3D only one driver switching and call it a day ? I know 100Gb/s SERDES are the heat right now. But there are those of us = =3D need to ship systems with memory on board. I am just asking for = help.=3D20 Now let me have a show of hands, who's willing to step to the plate ? .......sounds of cricket........ =3D20 -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Muranyi, Arpad Sent: Monday, March 31, 2008 6:58 PM To: si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Do you ever measure any of the things you model? Chris, SSO is a fairly difficult thing to model behaviorally, but we have at least two BIRD-s which intend to provide means to the legacy IBIS style behavioral modeling to help that situation. BIRD 95.6: http://www.vhdl.org/pub/ibis/birds/bird95.6.txt and BIRD 98.3: http://www.vhdl.org/pub/ibis/birds/bird98.3.txt Since we are getting very close to the release of the next major revision of the IBIS specification, these BIRD's will most likely become an official part of the IBIS 5.0 spec rather soon. And the reason I used the words "legacy IBIS style" above is because if someone is not shy about using any of the *-AMS extensions of IBIS, they can write out their heart's content using those modeling languages to include any SSO effect they wish. These extensions have been available since IBIS 4.1 was ratified in January of 2004. I agree, it may be debatable whether it is easier to simulate SSO with transistor models instead of figuring out how to model those effects behaviorally. However, I would like to caution the non-suspecting bystander that more transistor models than you would like to see make use of that handy little shorthand feature called .global, making SSO simulations useless or impossible. Add to that some encryption, and you are out of luck. Arpad =3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D= 3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D=3D 3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D= 3D =3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D= 3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D=3D 3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D= 3D =3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D=3D3D3D -----Original Message----- From: Chris Cheng [mailto:Chris.Cheng@xxxxxxxx]=3D3D20 Sent: Monday, March 31, 2008 6:32 PM To: Muranyi, Arpad; si-list@xxxxxxxxxxxxx Subject: RE: [SI-LIST] Re: Do you ever measure any of the things you = =3D3D model? There is always this little problem called memory. FBDIMM is not taking = =3D =3D3D off. DDR2/3 is around everywhere. I guarantee you I've seen SSO in DDR channels, a lot. How does one =3D3D correlate DDR models without SSO ? -----Original Message----- From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Muranyi, Arpad Sent: Monday, March 31, 2008 6:25 PM To: si-list@xxxxxxxxxxxxx Subject: [SI-LIST] Re: Do you ever measure any of the things you model? Chris, I think the dance floor is called SERDES these days. I haven't seen anyone using transistor level models in that field to get BER estimates down to double digit exponents. Whether it is done in C, Matlab, *-AMS, doesn't matter, its still behavioral modeling, isn't it? Of course, this is a slightly different topic from matching I-V and V-t curves, but no one can escape the correlation steps in this world either, as they shouldn't in the SPICE world either... And I think this thread started mainly around the topic of correlation. Arpad ---------------------------------------------------------------- -----Original Message----- From: Chris Cheng [mailto:Chris.Cheng@xxxxxxxx]=3D3D3D20 Sent: Monday, March 31, 2008 5:31 PM To: Muranyi, Arpad; si-list@xxxxxxxxxxxxx Subject: RE: [SI-LIST] Re: Do you ever measure any of the things you =3D =3D3D3D model? Arpad, I don't think we need to go into a flame war again. I just want to know who is actually shipping IBIS 4.0 or 4.1 models to = =3D =3D3D =3D3D3D customer that can do SSO. We are all dress up, I need to know where's the party ? The Gala Ball of SPICE models for customers with ten million dollar PO = =3D =3D3D =3D3D3D is out of my league, I am settling for the IBIS for the mass punk rave. = =3D =3D3D =3D3D3D Please tell me where is the dance floor ? ------------------------------------------------------------------ ------------------------------------------------------------------ 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 technical documents are available at: http://www.si-list.net List archives are viewable at: =3D20 //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 =3D20 ------------------------------------------------------------------ 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 technical documents are available at: http://www.si-list.net List archives are viewable at: =20 //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 =20 ------------------------------------------------------------------ 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 technical documents are available at: http://www.si-list.net 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