Maker Pro
Maker Pro

use HSPICE to simulate circuits if being a hardware engineer?.....

M

mikelinyoho

Jan 1, 1970
0
regards:

Does it make sense that to use HSPICE to simulate circuits if being a
hardware
engineer?.....
 
T

Tim Shoppa

Jan 1, 1970
0
mikelinyoho said:
Does it make sense that to use HSPICE to simulate circuits
if being a hardware engineer?.....

It works great for me because I'm only a simulation of
a simulated engineer.

How about you?

Tim.
 
T

Tim Wescott

Jan 1, 1970
0
mikelinyoho said:
regards:

Does it make sense that to use HSPICE to simulate circuits if being a
hardware
engineer?.....
Two of the best analog guys who I know personally use SPICE for all
their circuits. Of course, there's a lot of expertise in simulating a
circuit _right_ and knowing what stimulus to give it and what types of
analysis to subject it to.
 
W

Winfield Hill

Jan 1, 1970
0
Tim Wescott wrote...
Two of the best analog guys who I know personally use SPICE for all
their circuits. Of course, there's a lot of expertise in simulating
a circuit _right_ and knowing what stimulus to give it and what types
of analysis to subject it to.

In my experience the most important often-overlooked aspect of Spice
analysis is bench-verifying the component models for accuracy. It's
amazing just how poor most manufacturer's spice models are, almost as
amazing as the blind trust placed in them by typical spice enthusiasts.
 
J

Jim Thompson

Jan 1, 1970
0
Tim Wescott wrote...

In my experience the most important often-overlooked aspect of Spice
analysis is bench-verifying the component models for accuracy. It's
amazing just how poor most manufacturer's spice models are, almost as
amazing as the blind trust placed in them by typical spice enthusiasts.

(1) The models I receive from semiconductor foundries are VERY
accurate. As another poster pointed out, some foundries will accept
your Spice simulation as proof their processing was flawed.

(2) But otherwise I roll my own (except BJT's, they're pretty
accurate, AND easy to model ;-)

...Jim Thompson
 
W

Winfield Hill

Jan 1, 1970
0
Jim Thompson wrote...
(1) The models I receive from semiconductor foundries
are VERY accurate.

Excepting, of course, for the MOSFET subthreshold region?
 
J

Jim Thompson

Jan 1, 1970
0
Jim Thompson wrote...

Excepting, of course, for the MOSFET subthreshold region?

I've already demonstrated that FOUNDRY MICROCHIP models DO include the
subthreshold region.

The discrete houses (power MOSFET's) have not bothered to model that
region.

When I get a spare moment (when hell freezes over :) I'll fit a
Level=7 model to your data to prove the point.

What you need to do is convince the discrete houses that modeling that
region is useful to more people than just you ;-)

...Jim Thompson
 
M

Mike

Jan 1, 1970
0
Jim Thompson said:
On 21 Sep 2005 07:31:01 -0700, Winfield Hill
(1) The models I receive from semiconductor foundries are VERY
accurate. As another poster pointed out, some foundries will accept
your Spice simulation as proof their processing was flawed.

How does that work? "My simulations with your models indicate that the
circuit doesn't work, but we built it anyway, and it works fine. Please fix
your process." The opposite would appear to be just as problematic: "Here
are my simulation results. They show that the circuit works fine. Please
review the results and fix your process."

-- Mike --
 
R

Roberto

Jan 1, 1970
0
sure it does!

Let the computer make the calculations, so you can concentrate on the
real "engieering"

Roberto
 
J

Jim Thompson

Jan 1, 1970
0
How does that work? "My simulations with your models indicate that the
circuit doesn't work, but we built it anyway, and it works fine. Please fix
your process." The opposite would appear to be just as problematic: "Here
are my simulation results. They show that the circuit works fine. Please
review the results and fix your process."

-- Mike --

It's even more simplistic. If process test devices (generally five
points on the wafer) fall outside the foundry-provided worst-case
device models, the wafer can be rejected at foundry expense.

...Jim Thompson
 
K

keith

Jan 1, 1970
0
I've already demonstrated that FOUNDRY MICROCHIP models DO include the
subthreshold region.

I was in a presentation by a colleague that showed that even our
super-threshold models worked remarkably well in the sub-threshold region.
The modeling folks musta made a mistake. ;-)
 
K

keith

Jan 1, 1970
0
sure it does!

Let the computer make the calculations, so you can concentrate on the
real "engieering"

That's fine, as far as it goes. I've seen people (folks here have too)
arguing that the models are correct, even though they produce unrealizable
results. At least in this universe, 'C' isn't negotiable.
 
J

Jim Thompson

Jan 1, 1970
0
I was in a presentation by a colleague that showed that even our
super-threshold models worked remarkably well in the sub-threshold region.
The modeling folks musta made a mistake. ;-)

They're probably fit by a computer program, so that, even with
non-sub-threshold data, the curvature pretty well lines things up.

...Jim Thompson
 
W

Winfield Hill

Jan 1, 1970
0
Jim Thompson wrote...
They're probably fit by a computer program, so that, even with
non-sub-threshold data, the curvature pretty well lines things up.

Over what current range? I've posted real data over six orders of
magnitude, showing failures over all but the top two decades, but I
have yet to see any real *bench* data validating the putative "work
remarkably well" models. Remember, it's easy to make empty claims.

Moreover, keep in mind I'm not simply a nay-sayer: I've also posted
modified models that do work, plus links to literature for the same.

As for real-world uses for this kind of data and model performance,
try LDO regulators at low currents, if you aren't willing to accept
audio amplifiers at crossover. OK, sure, that's not important.

What's important is that the poor folks in Houston can't get out.
 
J

Jim Thompson

Jan 1, 1970
0
Jim Thompson wrote...

Over what current range? I've posted real data over six orders of
magnitude, showing failures over all but the top two decades, but I
have yet to see any real *bench* data validating the putative "work
remarkably well" models. Remember, it's easy to make empty claims.

Real men don't use Level=1 or Level=3 models ;-) That's the missing
link to your understanding.
Moreover, keep in mind I'm not simply a nay-sayer: I've also posted
modified models that do work, plus links to literature for the same.

As for real-world uses for this kind of data and model performance,
try LDO regulators at low currents, if you aren't willing to accept
audio amplifiers at crossover. OK, sure, that's not important.

Real men also don't use power MOS devices for audio outputs... you've
read the discussions here... they use tooobz ;-)
What's important is that the poor folks in Houston can't get out.

Looks like most everyone can't get out. How do you move a few million
people on short notice?

...Jim Thompson
 
W

Winfield Hill

Jan 1, 1970
0
Jim Thompson wrote...
Looks like most everyone can't get out. How do you move a few
million people on short notice?

Exactly. I suppose now you're a bit more understanding about
the difficulties New Orleans faced in fully evacuating before
the storm? Not that either city should be excused from having
workable plans, of course. But, not to worry, George will be
there tomorrow to clear out stalled traffic. Such a relief.
 
J

Jim Thompson

Jan 1, 1970
0
Jim Thompson wrote...

Exactly. I suppose now you're a bit more understanding about
the difficulties New Orleans faced in fully evacuating before
the storm? Not that either city should be excused from having
workable plans, of course.

Parking buses is a real plan? The population difference is also
almost an order of magnitude different... #4 city versus #32... a
whole lot less people to move from N'Orleans.
But, not to worry, George will be
there tomorrow to clear out stalled traffic. Such a relief.

Always one for the SNIDE remark aren't we ?:)

...Jim Thompson
 
Top