[sdiy] spice flamewars, Bob Pease vs Barrie Gilbert

[karl dalen]

Splendid! 
On wich email list did this argumenting take place?

Reg
KD

 --- Harry Bissell Jr <harrybissell at prodigy.net> skrev: 
> sorry this is so long... I cut out a lot of the
> headers to make it more readable and to protect the
> innocent from exposing their e-mail addys.
> Thought a lot of folks would get a kick out of it.
> I wrote to say that Spice has its place, once you KNOW
> when it is or is not lying.  thread follows H^) harry
> 
>  From: Ward Silver
>        Subject: Re: ALLEGED solutions with SPICE...
> 
>     The issue of SPICE lying or not lying reminds me
> of Steven Hawking's quote that, "God may not play dice
> with the Universe, but sometimes he throws them where
> you can't see them."
>      
>     Having worked with a fair number of students and
> young engineers over the years, I can tell you that
> any simulation at all is dangerously addictive to
> them.  They have absolutely no clue what the pitfalls
> are, what the symptoms of breakdown are, or even how
> to open the hood and really look at the data.  It's
> treated like a "magic box" in many cases. 
> Furthermore, since very few of them are experienced
> builders, they will make things in ways that SPICE
> assumes they won't and shouldn't.  And that more
> experienced engineers don't.
>      
>     It's like giving a brand-new driver the keys to a
> stock car.  EVENTUALLY, they may figure it out, but
> only after a lot of grief and after someone older and
> wiser takes them in hand.  It shouldn't have to be
> that way at all, but...
>      
>     1) SPICE (standing for simulation in general) is
> introduced too early and with too few cautions in the
> engineering curriculum, often in substitution for lab
> work.  This is BAD JUJU and leads to violations of
> Will Roger's First Law - "It's not what you don't
> know, it's what you know that ain't so."  Simulation
> should be done only in parallel with hands-on lab work
> and not in substitution for it except in rare
> circumstances.
>      
>     2) There is effectively zero emphasis on the
> simulate-build-compare cycle that shows them how the
> tool's version of reality differs from what the real
> electrons are doing.  Without that feedback, they
> don't learn when the simulation is getting into deep
> water.  Frankly, I think an entire quarter-long course
> should be devoted solely to use and misuse of
> simulation tools.
>      
>     3) High-speed stuff - anything over a few MHz - is
> really dangerous in the hands of inexperienced folks. 
> Until they get to senior-level courses, if at all, the
> subjects of parasitic reactances, transmission line
> effects, crosstalk, EMC, and other important are not
> covered.  Yet, they use models that cheerfully support
> frequencies into the GHz.  It's hard to explain to
> them why all that stuff in the VHF spectrum is really
> fiction.
>      
>     It's nuts that we give this powerful, but
> sensitive, tool to these young engineers with as
> little guidance and caution as we do.  Both industry
> and academia need to work harder on dealing with this
> problem.
>      
>     Regards,
>      
>     Ward Silver
>      
>      
>             From: Paul Rako
>         
>         Subject: Re: ALLEGED solutions with SPICE...
> 
> 
>         Hmmm-- looks like we have a good fight on our
> hands here.  I used to try and get between combatants
> and say "calm down", but today I am bored and so I say
> to both these esteemed gentleman--- are you going to
> take that shit from HIM???!!?!?!?!?!
> 
>         Ha ha ha--
> 
>         OK, on a more level keel-- I can offer a few
> observations.  First it is not surprising to me, an
> application engineer, that Mr Gilbert thinks that
> SPICE is just dandy.  Most IC guys do.  If you live in
> the transistor design world and have 10 to 50 million
> a year to feed both a Process and Modeling and CAD
> (sorry for the redundant "and" Barrie) yes, SPICE can
> give you a first-order indication of how an IC will
> work.  However, for us board-level guys-- SPICE is a
> lot less useful and a lot more dangerous.  At high
> speeds the board is a pretty important component and
> unless you drop the 30 grand for hyperlynx you can't
> factor the board's effect into the design on a
> computer.  
> 
>         I am also amazed that Analog Devices does not
> have any trouble with their models-- not even on new
> processes?  I have seen even our huge-dollar IC spice
> tell lies on a new process.  Sure it gets worked out
> but that can often take a couple of spins and that is
> big time and big money.  And the errors in SPICE are
> most often in the very same corner cases where Mr.
> GIlbert brags that SPICE does so well, over
> temperature and process and time.  
> 
>         If you want to be a software engineer I think
> that is a fine and honorable profession but I, like
> Mr. Pease, continue to be disturbed by this modern
> effete tendency by EEs to think that their job is to
> sit at a workstation all day playing with a computer. 
> Then they toss the design over a wall to another
> clueless geek that does the layout.  Then everybody is
> so surprised and blame the "tools" when the dang thing
> doesn't work.  "It can't be us-- were are so smart we
> never make mistakes"-- right.  It did work on the
> screen after all.
> 
>         I also tend to dispute Mr. Gilbert's dismissal
> of prototyping as to being a valuable contribution in
> the design process.  Very smart people seem to have a
> hard time understanding that they are not being paid
> to learn-- they are being paid to teach-- teach those
> nice people in manufacturing how to make a few million
> of whatever we make so any of us can get paid.  The
> goal is to get into a real physical piece of green
> plastic or shiny silicon that can go into production. 
> Worshiping SPICE keeps it in that computer screen
> phase for the majority of the development and then the
> "genius designers" can blame layout or the tools or
> something and all that happens is another 6 months
> gets added to the design cycle.  After working twenty
> years as a consultant I can assure everyone that the
> businessmen that pay the bills are lot happier to see
> a air-ball prototype that works but has problems
> rather then some computer print! out that shows what a
> rosy world we live in.  At least you can put the
> air-ball into a pretty box and make the show in
> Chicago.  That screen is not reality-- it is a
> dangerous drug and although a little can ease your
> pain, a lot of SPICE and you will just be a junky with
> your eyes glazed over staring into the screen and no
> sense of what you are really trying to do-- like
> getting real hardware going out the door.
> 
>         Now Bob does like to get on his high horse
> about SPICE,  but you have to understand any good
> preacher really can't go soft on the devil and keep
> the faithful from harm.  I have seen him advise young
> engineers on how to mix SPICE with a few real-worlds
> tests-- maybe those air-balls dead-bugged on a board,
> and then from that see where SPICE lies (and it lies
> by definition-- as my Professor once said--
> "everything is based on linear analysis but even a
> resistor goes nonlinear when you put enough juice in
> it to melt it").  Then it is time for breadboards--
> real green plastic boards with real black plastic
> chips and you still have to count on three spins to
> get everything right.  
> 
>         Sure SPICE is great-- when the box is small
> enough.  How appropriate that in college we called the
> vectors and matrix classes "boxes and arrows".  Yup,
> that matrix is only right in a very narrow regime--
> with very good models and proven processes and perfect
> design rules and a manufacturing operation at 6-sigma
> and on and on and on- yup-- SPICE works just fine. 
> But in the real analog world SPICE is doing a lot of
> damage-- not only to budgets but to impressionable
> young minds.   And if your company is making optimum
> use of it's resouces the processes are new and models
> imperfect and the manufacturing is good-enough, not
> perfect.  
> 
>         So I guess I would have to say that we
> shouldn't hold a digital opinion in an analog world--
> yeah, Bob does rant about SPICE but he understands in
> can teach a few things, especially to idiots like me. 
> But the more experienced you get you may now where to
> distrust SPICE, as Mr. Gilbert so well demonstrates. 
> I maintain that an Israeli commando with a slingshot
> will be more effective thane a Boy Scout with an Uzi. 
> I suspect Bob sees indiscriminate SPICE use as
> analogous to giving Boy Scouts Uzis.
> 
>         I just did a PSPICE simulation of our very
> fast LMH6624 used as a trans-impedance amp.  I
> integrate the output with another LMH6624 and feed
> that back to the "+" pin of the main amp to servo out
> the DC.  It simulates just fine.  But all my
> experience tells me "DANGER WILL ROBINSON".  I know
> fast amps just don't like getting hooked tail-to-nose
> like this.  I will build it and report if the real
> circuit works or not.
> 
>         Paul
> 
=== message truncated ===