[SI-LIST] Re: IBIS Model Quality (or lack thereof)

My favorite analogy is the data sheet. What if the typical data sheet
was as accurate as the typical IBIS file?


On Tue, 2002-03-26 at 11:22, Dunbar, Tony wrote:
> 
> All,
> 
> As with other threads on the SI List, we are teetering on the brink of
> getting away from the original question or topic. I apologize to Todd
> up-front for pitching in and contributing further to the distractions with
> my own comments. I also applaud Adam for his contribution and those he has
> made in the past because, as far as I'm aware, he's the only Si vendor IBIS
> model-maker that I ever see on the List. I'm ready to be flamed for that,
> but I suspect I won't feel much heat because, somewhat indicative of the
> problem, there are so few other IBIS model-makers on the List and/or have
> any affinity for the end-user situation, as Todd indicated in his opening
> post.
> 
> As regards the matter of IBIS model "cleansing service" and using that term
> as a eupemism for "sanitization", I'm afraid the latter is often the more
> appropriate one to use. The fact that this should even be necessary for a
> simulation model is quite appalling to me. I, along with my customers, have
> fought this for over 7 years. I know there are many on the list who have
> fought it for longer. Please note that I am not saying that a model should
> not be checked when it is received and prior to being used; that's only good
> engineering practice.
> 
> But let me ask the following question, which, though stated in hypothetical
> terms, it is not too wide of the mark:
> 
> If your ASIC or FPGA or any other Si vendor supplied digital simulation
> models (e.g. VHDL or Verilog) to the same quality that many IBIS models are
> supplied, would the industry tolerate it to the same extent poor quality
> IBIS models are tolerated?
> 
> Sorry, but poor quality IBIS models is a pet subject of mine, since it
> consumes so much of my time and that of my customers.
> 
> Regards,
> Tony
> 
> -----Original Message-----
> From: Joe Socha [mailto:Joe.Socha@xxxxxxxxxx]
> Sent: Tuesday, March 26, 2002 9:45 AM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: IBIS Model Quality (or lack thereof)- Cleansing
> service
> 
> 
> 
> Hello All,
> 
> It is matter of Knowledge and Time. And, as we all know time is money.
> Yes the Vendors need to be IBIS savvy, and they have to want to put in the
> time to check their models. I believe Todd is asking the semiconductor
> community to let us know which of them takes the time to check / test/
> verify/ correlate their models before releasing them to the public. ( I just
> saw the response from Adam at Fairchild.)
> 
> Plexus provides engineering and SI services, and as such we cleanse all the
> model we use. (I prefer to describe it as cleansing rather than Sanitize
> because sanitize sounds too drastic.) This includes reviewing all the data
> testing, modifying the model as needed, working with the vendor as
> necessary, testing the model on some standard loads and other activities
> such as that. Unfortunately or fortunately depending on your point of view,
> the cleansed model is owned by the customer paying for the work we do. This
> becomes a competitive advantage for our customer, but unfortunately does not
> benefit the overall user community
> 
> I like to use the free " Visual IBIS editor" from Innoveda ( formerly
> Hyperlynx) There is a new version ( V 3.0) on the web site, and it has some
> significant improvements over V2.0. As with all tools it is not perfect, but
> it is a time saver for me.
> http://www.innoveda.com/products/datasheets_HTML/ibis.asp   When dealing the
> semiconductor supplier I always ask them to download it so we can talk"
> apples to apples" for comparisons and understandings. We have also produced
> some models for semiconductor companies and we run the models in several of
> the simulators and overlay the results to show that under some typical
> topologies the models/simulators all produce similar results.  
> 
> So it comes down to time. Somebody has to put in the time. The semiconductor
> company, the user, or outsourcing. 
> 
> 
> Thanks,
> Joe Socha
> ----------------------------------------------------------------------------
> -------
> Joe Socha:   Technical Lead, Signal Integrity Analysis 
> Plexus Technology Group
> Snail Mail: 400 Amherst Street               Nashua, NH 03063
> Email: joe.socha@xxxxxxxxxx
> Direct: 603.864.1334, Fax: 603.864.1301, Main: 603.864.1300
> ----------------------------------------------------------------------------
> -------
> 
>  -----Original Message-----
> From:         pwelling@xxxxxxxxxxxxxx [mailto:pwelling@xxxxxxxxxxxxxx] 
> Sent: Monday, March 25, 2002 6:56 PM
> To:   twester@xxxxxxxxxxx; si-list@xxxxxxxxxxxxx
> Subject:      [SI-LIST] Re: IBIS Model Quality (or lack thereof)
> 
> 
> Todd,
> 
> Many IBIS Models will indicate in the comments section that they are a =
> SPICE
> to IBIS conversion or a Benched model. Even though they say they are =
> bench
> models, sometimes they don't correlate very well because test =
> conditions
> versus real world loading are two different things sometimes. Sometimes =
> the
> power supply / ground isn't as clean as desired, load impedance varies =
> with
> trace topology, etc.
> 
> I have had reasonable correlation on some models and some not.=20
> 
> The first thing (in my oppinion) that needs to done is to at least =
> parse the
> models - before they are published. So many times I have opened models =
> just
> to find stupid little syntax problems and things like 1/0 Buffer =
> (instead of
> I/O Buffer) for the I/O type. They could not have parsed it before
> publishing it, so how did it run when they tested the model - and on =
> what
> simulation tool? Do vendors test the model on more than 1 tool, it's
> supposed to be universal, but I think they would be surprised how the =
> model
> might behave.
> 
> I opened one today (a new device) that had several errors in it - =
> stupid and
> complex.
> 
> Then there is the problem of what revision IBIS models the Simulator =
> will
> support. On older simulators, I have had to strip out 3.x model =
> information
> to get them compile or to even run without hanging up the system.
> 
> IBIS can work, as long as you know the models, where they came from, =
> have
> "sanitized" them, and your simulator will support them. Generally, you =
> can't
> just use them off of the web without some work. Some model houses will
> create them for you, but be sure you understand what you are getting. =
> Some
> just SPICE to IBIS, some create minimal point models, dV/dT's are =
> typical
> (whatever that is), etc.
> 
> Yes, SPICE will give more accurate results, take longer to run, and are
> generally not applicable for routed board level trace simulation within
> budget and schedule constraints.
> 
> I agree with you that we need help with IBIS models and it has to =
> originate
> with both the device vendor and simulator vendor.
> 
> 
> Philip Ross Wellington
> Mgr. Signal Integrity & EMI
> L-3 Communications CSW
> 
> 
> -----Original Message-----
> From: Todd Westerhoff [mailto:twester@xxxxxxxxxxx]
> Sent: Monday, March 25, 2002 4:28 PM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: IBIS Model Quality (or lack thereof)
> 
> 
> 
> Oh no, here we go ...
> 
> For the record, we've had *VERY* good luck getting HSpice to correlate =
> with
> IBIS simulations - the problem is, we've had to create most of the IBIS
> models ourselves ir order to get things to match.
> 
> Please, let's save the "ultimate HSpice accuracy" arguments for another
> thread (preferably one that gets redirected to /dev/null).  I'm well =
> aware
> of IBIS' limitations, and we've done lots of real-world comparisons for =
> our
> current applications.  IBIS will work just fine if we have decent =
> models ...
> AND, I'll be able to do all the post-route analysis I want.
> 
> The question remains, folks ... what semiconductor vendors are willing =
> to
> "step up the plate" and tell us how they test their IBIS models to =
> ensure
> correlation to a known source (measurement or HSpice)?
> 
> Inquiring minds want to know!
> 
> Todd ;-)
> 
> Todd Westerhoff
> Signal Integrity Engineer
> Hammerhead Networks
> 5 Federal Street - Billerica, MA - 01821
> email:twester@xxxxxxxxxxx - ph: 978-671-5084
> 
> "starting With the man in the mirror
>  asking him to change his ways
>  And no message could have been any clearer
>  If you wanna make the world a better place
>  Take a look at yourself, and then make a change"
> 
> 
> - "Man in the Mirror", Michael Jackson
> 
> 
> 
> 
> 
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx
> [mailto:si-list-bounce@xxxxxxxxxxxxx]On Behalf Of Chris Cheng
> Sent: Monday, March 25, 2002 5:18 PM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: IBIS Model Quality (or lack thereof)
> 
> 
> 
> Why use IBIS at all ?
> If you are convinced that IBIS is not good enough (which I am), don't
> simulate and just follow whatever the vendors design rules say. After =
> 
> all,
> if you deviate from the rules based on your incorrect IBIS models, do =
> 
> you
> think the vendor will really take responsibility if it doesn't work ? =
> 
> If you
> follow the design rules, what value does the simulation add to you
> design ?
> If you are serious enough to simulate, you must be working on something =
> 
> that
> is unique about your own system that needs to be custom optimized
> beyond
> what the component guideline can tell you. Get the SPICE model and do a =
> 
> good
> job of modeling your own system. If a small little start up like mine =
> 
> can
> get SPICE model from vendors, anyone can. Its just a matter of a few =
> 
> NDA and
> extra effort.
> 
> -----Original Message-----
> From: Todd Westerhoff [mailto:twester@xxxxxxxxxxx]
> Sent: Monday, March 25, 2002 12:59 PM
> To: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] IBIS Model Quality (or lack thereof)
> 
> 
> 
> Hi all,
> 
> We have been doing a lot of comparisons between IBIS and HSpice models
> lately, with a lot of less-than-satisfactory results.
> 
> I have an open question for the semiconductor suppliers among us:
> 
> What testing is performed after an IBIS model is created, to ensure
> that the
> completed model behaves as expected?  I'm interested in responses from
> vendors that actually SIMULATE the IBIS model before distributing it =
> 
> for
> customer use.  It looks to me as though many of the models I've
> received
> were never used in a simulation before being released ...  and, as a
> potential customer, I find that quite disheartening.
> 
> So the question is - what processes are being used to test IBIS models =
> 
> and
> ensure they function properly before their release to customers?
> 
> As always, replies both on and off the list are greatly appreciated.
> 
> Todd.
> 
> Todd Westerhoff
> Signal Integrity Engineer
> Hammerhead Networks
> 5 Federal Street - Billerica, MA - 01821
> email:twester@xxxxxxxxxxx - ph: 978-671-5084
> 
> 
> "starting With the man in the mirror
>  asking him to change his ways
>  And no message could have been any clearer
>  If you wanna make the world a better place
>  Take a look at yourself, and then make a change"
> 
> 
> - "Man in the Mirror", Michael Jackson
> 
> ------------------------------------------------------------------
> 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:
> http://www.freelists.org/webpage/si-list
> 
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
> 
> List archives are viewable at:   
>               http://www.freelists.org/archives/si-list
> or at our remote archives:
>               http://groups.yahoo.com/group/si-list/messages20
> 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:
> http://www.freelists.org/webpage/si-list
> 
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
> 
> List archives are viewable at:
>               http://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:
> http://www.freelists.org/webpage/si-list
> 
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
> 
> List archives are viewable at:    =20
>               http://www.freelists.org/archives/si-list
> or at our remote archives:
>               http://groups.yahoo.com/group/si-list/messages=20
> 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:
> http://www.freelists.org/webpage/si-list
> 
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
> 
> List archives are viewable at:     
>               http://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:
> http://www.freelists.org/webpage/si-list
> 
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
> 
> List archives are viewable at:     
>               http://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:
> http://www.freelists.org/webpage/si-list
> 
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
> 
> List archives are viewable at:     
>               http://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
>   
> 
-- 
=============================================================================
Mike Mayer
mwmayer@xxxxxxx


------------------------------------------------------------------
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:
http://www.freelists.org/webpage/si-list

For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field

List archives are viewable at:     
                http://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

Other related posts: