Re: Circuit simulation software

John Larkin wrote:
On Sat, 24 Jun 2006 16:45:37 GMT, "Kevin Aylward"
<see_website@xxxxxxxxxxxxx> wrote:


The *only* way to be 100% sure of a prediction is actual
measurement after the fact.

Sure. But I can test a production first-article, and production can
test every unit before they ship. If I have a high enough confidence
that a design will work, it's an expensive waste of time to
breadboard or simulate.


Ahmmmm.. Well we are coming totally opposite directions. There is no
f^%$ing chance that typical analogue i.c designs can be produced,
without extensive simulation. The cost of simulation compared to mask
costs and 3 month waits until it comes back from the fab is a risk
*NO* semiconductor company will make. Ever. Indeed, it can be a
sacking if you ship the tape without doing WC/MC.


I'm putting parts on boards, and you're baking silicon, so things are
a bit different. I can sell, say, 95% or more of my original board
layouts, given the option to change values and maybe add a discreet
kluge. And spinning a board takes as little as a week or two, and
costs a couple thousand dollars. Each of us, ideally, applies the
optimum amount of pre-fab verification to maximize payout, and neither
of us can sensibly keep testing and simulating to get to 99.99%
confidence; at some point we have to say "enough" and fab the damned
thing.

But I still think that a lot of people fling out their "designs" too
fast and then simulate too much. I encourage my engineers, and myself,
to get it right by design and spend less, typically zero, time
simulating and breadboarding, and that becomes a habit and a
discipline that pays off in time and quality.

This depends on your definition of "design". For me design *is* running
appropiate simulations. Of course, running simulations blind is not
"design", and I am not suggesting that.

For example, suppose I wanted to set the Vgs voltage on a mosfet
connected diode. Do I know what the formulas are inside out? Sure I do.
Do I bother to look up the spice data for K and calculate the value?
Nope. Its far quicker for me to run a few sims from an initial guess
from experience of knowing the approx Vgs for a given current. I know
what the sensitivities are so I know what things I can wing it with. I
don't even calculate 7*9 by hand.

My pen and paper is the GUI schematic. Its just far quicker to try out
new connections in the virtual world.



It's ideal to go from design directly to a sellable PC
board. You've got to lay out and assemble and test the board
eventually, so you may as well do it now. The skill is in properly
calibrating the risk.

And one never taken today, with ic design.

Do you run at exactly zero risk? 100% of first silicon is always
right?

John

I was referring to the risk of not simulating at all. Of course there is
a risk, and many chips end up with several spins before its gotten
right. IC Companies usually have a *mandatory* simulation sign off
***.

Kevin Aylward B.Sc.
431infoEXTRACT@xxxxxxxxxxxxx
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

"There are none more ignorant and useless,than they that seek answers
on their knees, with their eyes closed"


.

Quantcast