It's Official... PSpice Schematics v10.3!

[Ken Smith]

I guess my point here is that while I'd readily admit that the average
design out there probably could obtain a higher yield with no significant
change in cost, there are also times when it's entirely reasonable to accept
a lower yield just so that you can ship the @#$%@# product and get on with
life.

Where I work, if they plan on shipping 10, they buy enough for 9. Anything
less than 100% yeld is considered a problem to solve.

I've heard that the IC yields on high-end 3D graphics chips are
abyssmal -- around 10% -- yet clearly there's a demand for them and it'd be
absurd to suggest that they simply shouldn't be manufactured unless the
yield could be increased.

They have the advange of having millions in the pipeline and a robot to do
the testing. People who use their chips don't have that advantage so they
want 100% of the chips shipped to be tested as good. The IC making
business is quite a diffent world from the IC using businesses.

BTW, I suspect that if you simulate any of those chips with the absolute
worst case tolerances on all the components the yield drops to 0%.

This could well be true today. I bet the makers are trying hard to raise
the yeld by making the parameter spread smaller.

 

[Jim Thompson]

They have the advange of having millions in the pipeline and a robot to do
the testing. People who use their chips don't have that advantage so they
want 100% of the chips shipped to be tested as good. The IC making
business is quite a diffent world from the IC using businesses.

Yep. The real game is to test BEFORE packaging, so the throwaway cost
is minimal... just like your Cheerios, the biggest cost is in the
packaging ;-)

This could well be true today. I bet the makers are trying hard to raise
the yield by making the parameter spread smaller.

In the IC business we have an advantage called _ratioing_, resistors
may be 30% low, but so what, they're ALL low.

Performance in a GOOD analog circuit design depends primarily on
ratios.

...Jim Thompson

 

[Chaos Master]

SuperSpice has a brilliant GUI.

I am trying SuperSpice... my only complaint is that the interface is a
bit clumsy if you're not using a high resolution.

Since I don't like high resolutions, this is a problem.

(Otherwise, SS is a very good tool. Congratulations!)

One interface I liked was the one from SIMetrix.
Very fast and clean -- just has a very bad choice of keyboard shortcuts
(F1-F12 keys for almost everything -- I'm used to CTRL+R or R everywhere
for rotate), and those shortcuts can't be changed.

[]s
--
Chaos Master®, posting from Brazil.
"It's not what it seems, not what you think. No, I must be dreaming."

http://marreka.no-ip.com | http://tinyurl.com/46vru
http://renan182.no-ip.org | http://marreka.blogspot.com (in Portuguese)

 

[Chaos Master]

I heard rumors that Orcad for DOS was now freeware.

AFAIK, not yet. Please correct me -- I'd like to know as I have an
useless DOS + Windows 3.1 machine here which would like Orcad for DOS.

Maybe you've confused with AutoTrax (now Protel?) for DOS... this one is
free.

[]s
--
Chaos Master®, posting from Brazil.
"It's not what it seems, not what you think. No, I must be dreaming."

http://marreka.no-ip.com | http://tinyurl.com/46vru
http://renan182.no-ip.org | http://marreka.blogspot.com (in Portuguese)

 

[Kevin Aylward]

I am trying SuperSpice... my only complaint is that the interface is a
bit clumsy if you're not using a high resolution.

I agree that you definitely want a big monitor set to high resolution.
Up to a couple of weeks ago I was using a 17" at 1280 by 1024. I found
this very acceptable. However, I now have a 19" set to 1600 by 1200, but
he change was simply because my old monitor failed. It was only £100 so,
today, monitor size should not be an issue.

Since I don't like high resolutions, this is a problem.

(Otherwise, SS is a very good tool. Congratulations!)

Thanks. Have you got the latest I finally got around to adding some
16/8/4 Bit AD and 16/8/4 Bit DA converters. The models are pure spice3
so they will also run in LTSpice. I use LTSpice myself as a check on
convergence so I try to make sure I have a set of analog versions of
digital models. If you do try to run the SS generated net list in
LTSpice, you have to press the pink "I" button to generate a default
include file. For reasons uknown LTSpice will simply halt if it dosnt
find an include file, when it should really just issue a warning.
LTSpice also does this with .options it dosnt understand. I have one
option that simply tells the engine to output floats instead of doubles
to reduce fie size. You only need doubles for the calculations not the
fnal output, usually.

My philosophy is that the engine should do its best to run, and only
fail if it actually has to. So if Mikes reading this, how come not
warnings instead of a fatal error?

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Joel Kolstad]

My philosophy is that the engine should do its best to run, and only fail
if it actually has to.

This is an 'expert friendly' argument, Kevin. In general, if an include
file can't be found, with most users it's FAR more likely that that file is
needed for the simulation and in all likelihood continuing will lead either
to erroneous results or no results at all. Far too many beginners are
likely to believe erroneous output if you don't bludgeon them over the head
with the fact that something about the simulation seems amiss.

But you could readily convince me that there should be a checkbox somewhere
for 'treat [various] failures as warnings'; this is not uncommon with, e.g.,
C compilers.

 

[Kevin Aylward]

This is an 'expert friendly' argument, Kevin.

Not quite. I use the escape clause. "...if it actually has to..." This
means that if a novice is using it, it needs to know that's and take
appropriate action

In general, if an
include file can't be found, with most users it's FAR more likely
that that file is needed for the simulation and in all likelihood
continuing will lead either to erroneous results or no results at
all. Far too many beginners are likely to believe erroneous output
if you don't bludgeon them over the head with the fact that something
about the simulation seems amiss.
But you could readily convince me that there should be a checkbox
somewhere for 'treat [various] failures as warnings'; this is not
uncommon with, e.g., C compilers.

Yes.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Joerg]

Hi Jim,

What do you think of LTSpice? I know why you need the Cadence package.
But LTSpice comes with a schematic entry that offers a hierarchy
structure which you mentioned before as very important (and I agree with
that).

Regards, Joerg

 

[Jim Thompson]

Hi Jim,

What do you think of LTSpice? I know why you need the Cadence package.
But LTSpice comes with a schematic entry that offers a hierarchy
structure which you mentioned before as very important (and I agree with
that).

Regards, Joerg

I think LTSpice is fine, although I have misgivings about Mikey's
twiddling of gmin, and other tricks that enhance speed but don't yield
waveform-match to what I see in PSpice Probe.

I think Mark ("qrk") recently ran a chunk of their sonar chip on
LTSpice and saw some bizarre results.

...Jim Thompson

 

[Joerg]

Hi Jim,

I think LTSpice is fine, although I have misgivings about Mikey's
twiddling of gmin, and other tricks that enhance speed but don't yield
waveform-match to what I see in PSpice Probe.

I think Mark ("qrk") recently ran a chunk of their sonar chip on
LTSpice and saw some bizarre results.

Thanks, Jim. I just downloaded LTSpice, having only had exposure to
PSpice before. It does seem to allow the usual step setting so maybe
these kinds of trade-offs can be avoided. What I was surprised about is
the fairly large number of netlists it supports if you want to use the
schematic part as the point of entry on projects.

Regards, Joerg

 

[Chaos Master]

Kevin Aylward is missing:

Thanks. Have you got the latest I finally got around to adding some
16/8/4 Bit AD and 16/8/4 Bit DA converters. The models are pure spice3
so they will also run in LTSpice. I use LTSpice myself as a check on
convergence so I try to make sure I have a set of analog versions of
digital models. If you do try to run the SS generated net list in
LTSpice, you have to press the pink "I" button to generate a default
include file. For reasons uknown LTSpice will simply halt if it dosnt
find an include file, when it should really just issue a warning.
LTSpice also does this with .options it dosnt understand. I have one
option that simply tells the engine to output floats instead of doubles
to reduce fie size. You only need doubles for the calculations not the
fnal output, usually.

Very good... I haven't tried the AD/DA converters, but they look good.
Even more so since they're SPICE3 compatible.

 

[Kevin Aylward]

Kevin Aylward is missing:



Very good... I haven't tried the AD/DA converters, but they look good.

The example circuit is 16Bit-AD-DA.sss.

Even more so since they're SPICE3 compatible.

As an exercise, see if you can decipher the method of the AD. Hint:
instantaneous successive approximation. I might post a detailed circuit
of the method, if I can around to it.

The reason I decide to make them was investigating modulating the power
supply of RF transmitters. One wants to make the supply track the
envelope of the modulating signal to get the efficiency up. So one goes
A/Ds to D/A directly, which seems quite daft initially! The DA switches
little floating PS (say Transformer with diode cap) in a series/bypassed
fashion. The final output is the filtered quantised input used as a PS
to the TX, with low loss. The idea being that one is only limited by the
raw switching speed of the dac'ed PS. A PWM is going to have trouble at
say, 50Mhz modulation frequency.


Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.

 

[Ken Smith]

This could well be true today. I bet the makers are trying hard to raise
the yield by making the parameter spread smaller.

In the IC business we have an advantage called _ratioing_, resistors
may be 30% low, but so what, they're ALL low.[/QUOTE]

In discrete land, getting matched sets does this too but it costs like the
dickens. I've had R-packs made to get the tight ratio match needed it
turned out to be the less expensive option.

Performance in a GOOD analog circuit design depends primarily on
ratios.

Yes, and of them it depends on the ratios that are an easier match.

The trouble starts when the parameter you are trying to control is
frequency. It is hard to get frequency to depend on a parts ratio. This
is a big part of why almost everything is fed into a DSP these days. The
DSP cost less than the discretes that would be needed.

 

[Fred Bloggs]

I heard rumors that Orcad for DOS was now freeware.

Paul C

Did ya' hear that, retard boy? Geez- thnx for the tip- you're quite the
ass-ette to the forum....