Spice simulation - Page 6 - diyAudio
Go Back   Home > Forums > Design & Build > Software Tools

Software Tools SPICE, PCB CAD, speaker design and measurement software, calculators

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 15th May 2007, 05:48 PM   #51
ingrast is offline ingrast  Uruguay
diyAudio Member
 
Join Date: Sep 2004
Location: Montevideo


Rodolfo
  Reply With Quote
Old 15th May 2007, 07:22 PM   #52
Nordic is offline Nordic  South Africa
diyAudio Member
 
Nordic's Avatar
 
Join Date: Sep 2005
MAN!!! I hate it when I cant find something... I saw this article about a computer/or software that does thousands of evolutions on a base concept, compareing it etc... it was pretty new...

The jist of it basicaly was, if you have enough of those babies, humans only have to ask... and supply some parameters...
  Reply With Quote
Old 15th May 2007, 07:55 PM   #53
diyAudio Member
 
Join Date: Sep 2006
Quote:
Originally posted by peufeu
You know, in EE school we had a class called "Numerical analysis" which covers computer simulation of pretty much anything.

The first thing that they teach you is that you should know HOW your simulator works, so that you know WHAT you are simulating, and how much you can trust the results.

Simulations are just a way to solve equations which represent an arbitrarily chosen part of reality. And reality is always right.

Drawing a circuit in Spice and then pressing a button to get a nice curve, means nothing.

If the effects you are interested in are well represented by the models your simulator uses, and are not swamped in reality by stuff that the simulator doesn't model, then you'll get something usable. When you know what you're doing, that tends to be the case.

However, when the things you try to study are not in your models, "knowing what you're doing" can mean many things, like not running the simulation at all since you know it will be useless, or hacking subcircuits from measurements to get a better model, or running it anyway, and comparing with real measurements, and studying the interesting differences.

But don't spit on simulations. Without them, you wouldn't be sitting in front of your computer, because it could never have been designed.

Stuff that SPICE lacks :

- thermal effects (self-heating, thermal inertia, thermal runaway, influence of temperature on transistor hFE isn't well modelled)
- Gummel-Poon is full of holes, like Cbc variation with Vce, etc
- Some transistor models are really suspicious
- Doesn't model layout, ground loops, etc (Protel does signal integrity and crosstalk modelling though)
- MOSFETs and JFETs aren't very realistic
- all transistors are perfectly matched (in real life, you'll get offsets)
- etc

I agree completely. SPICE is tremendously useful as long as you respect its limitations.

Often, the insight that it provides is not seriously compromized by its limitations, and I have to admit that I have uncovered many of my own mistakes and misconceptions in SPICE simulations. The ability to probe node voltages and currents at will (without "disturbing" the circuit) allows for a lot of useful (and quick) poking around. It also doesn't take much time to get a simulation up and running to the point where it begins to deliver useful results, and is certainly not something that is limited to big companies.

Although it doesn't do a very good job of modeling board layout, I have often found it useful and insightful to add some parasitic inductances here and there to see their effect.

But there is no substitute for building the thing and measuring it diligently. I tend to reject papers where the claimed results are based solely on SPICE simulations. Candidly, I must also admit that there have been times when I have designed and built a circuit without SPICE, and when it did not do what I expected, I resorted to SPICEing it. In those cases I probably would have saved myself hours if I had done it the other way around.

Bob
  Reply With Quote
Old 15th May 2007, 08:20 PM   #54
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Recently, on another topic :

- hey check this circuit
- it will go into thermal runaway
- but no, it works well in spice
- it will burn all the same

Does someone know of some simulation software with the new transistor models like VBIC / Mextram, and that doesn't cost $20K ? I'd really like to try those.
  Reply With Quote
Old 15th May 2007, 09:37 PM   #55
diyAudio Member
 
Join Date: Sep 2006
Quote:
Originally posted by peufeu
Recently, on another topic :

- hey check this circuit
- it will go into thermal runaway
- but no, it works well in spice
- it will burn all the same

Does someone know of some simulation software with the new transistor models like VBIC / Mextram, and that doesn't cost $20K ? I'd really like to try those.

I know what you mean. Like putting a finite element analysis thermal package right in with SPICE. It would be nice in a really great free SPICE simulator like LTSPICE, but I don't know of any.

Bob
  Reply With Quote
Old 15th May 2007, 09:52 PM   #56
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Well, finite element is overkill, but philips has open-sourced their Mextram model, so some GNU long beard folk should have done something with it, right ?
  Reply With Quote
Old 15th May 2007, 09:58 PM   #57
PB2 is offline PB2  United States
diyAudio Member
 
PB2's Avatar
 
Join Date: Sep 2004
Location: North East
Blog Entries: 1
Quote:
Originally posted by peufeu
Recently, on another topic :

- hey check this circuit
- it will go into thermal runaway
- but no, it works well in spice
- it will burn all the same

Does someone know of some simulation software with the new transistor models like VBIC / Mextram, and that doesn't cost $20K ? I'd really like to try those.

Your example above really falls into the category of don't expect it to do what it is not coded or set up to do automatically. I believe that such a thermal analysis could be done, given that there are electrical analogies to thermal "circuits", however I don't plan on doing this any time soon. Obviously, you could have a voltage that represents temperature in the real physical design. It's all just more work and more CPU cycles.

Pete B.
  Reply With Quote
Old 15th May 2007, 10:13 PM   #58
peufeu is offline peufeu  France
diyAudio Member
 
Join Date: Mar 2001
Location: Lyon, France
Yes, that's what my above example was meant to show : blind trust in tools is, well, blind.

The purpose is not to simulate thermal runaway (for this, a subcircuit is OK), but to simulate self-heating effects in a complete amplifier, which influence both Vbe and hFE, and hence, open loop gain, phase margins, output bias, input stage balance, offset, etc, etc.

I've done the temperature-voltage thing with capacitors to represent thermal capacitance of transistor die. But the SPICE models don't have a temperature node, so you have to play with dependent sources, and modelling hFE variations is a pain.

Also, Mextram is better for other stuff like non-constant capacitances in BJT...
  Reply With Quote
Old 15th May 2007, 10:18 PM   #59
PB2 is offline PB2  United States
diyAudio Member
 
PB2's Avatar
 
Join Date: Sep 2004
Location: North East
Blog Entries: 1
Quote:
Originally posted by GRollins
Jeez, I should think it should be obvious--it's right there in the spec sheets. Devices behave differently when they get hot. But you can't simulate that because you can't know the ambient temperature in the room...
hence the efficiency of the heat transfer from the heatsink to the room...
hence the actual temperature of the heatsink...
hence the actual temperature of the output device...
hence the actual temperature of the semiconductor chip itself...
hence the actual, literal, real world behavior of the MOSFET's bias and transconductance in the practice of delivering music to the listener's ear.
For want of a nail, the kingdom fell.

Grey

You should understand that your statements are based on how you think real world engineering is done.

The fact is that many people can hack together a design on a bench and make it work, but most professionals know that a production design has to work over process, voltage, and temperature (PVT). Why is this so important in industry? Money and reliability ... If you do a design that makes it to production, ships, and a large percentage of the units are returned due to marginal failures, then that product might cause the company losses rather than profits. Big problem.

Small designs, and reprogrammable designs such as FPGAs can be built and tested in the lab, but this is not practical for large designs or chips being custom or semi-custom fabricated.

Simulation of complete systems, or even often subsystems was not practical years ago due to time constraints, however today in the last 10 to 15 years it is.

I worked in the semiconductor industry where we guaranteed chips (complex often more than 100,000 transistors) to behave the same as in simulation for both functional and (analog) timing behavior over process, voltage, and temperature. We used a variety of simulators, probably many that you never heard of, some for functional verification, and others for timing verification. Cells in libraries are characterized with SPICE, but SPICE is just part of the solution.

Designs were not accepted from a customer that had not been simulated since the simulation was part of the contract to fab (expensive) the part. If the part performed as simulated but did not work in the system then it was a customer's error, if the part did not behave as simulated then it was our issue. Test vectors were captured from a simulation run, and run against the real part on a tester.

We made mistakes from time to time, and it was usually an error in the model, or a bug in the simulator but most often the better tools did work well. Obvioulsy, one had to choose a quality tool set since bugs could mean failure/delays for a project. I did designs also as a customer and saw poor quality tools that nearly sunk several projects.

My point here is that people who claim that SPICE and simulation are useless as a blanket statement would be proven wrong by major segments of the engineering industry.

I have used SPICE for probably more than 20 years, and obviously if you do not have a validated set of models for your semis you cannot blame the simulator. You should also become familiar with the limitations of the models.

You are right, you don't know the temperature in the room, but the fact is that you should be testing over the full intended operating conditions to verify that your design is robust.

Pete B.
  Reply With Quote
Old 15th May 2007, 10:43 PM   #60
PB2 is offline PB2  United States
diyAudio Member
 
PB2's Avatar
 
Join Date: Sep 2004
Location: North East
Blog Entries: 1
Quote:
Originally posted by peufeu
You know, in EE school we had a class called "Numerical analysis" which covers computer simulation of pretty much anything.

The first thing that they teach you is that you should know HOW your simulator works, so that you know WHAT you are simulating, and how much you can trust the results.

Simulations are just a way to solve equations which represent an arbitrarily chosen part of reality. And reality is always right.

Drawing a circuit in Spice and then pressing a button to get a nice curve, means nothing.

If the effects you are interested in are well represented by the models your simulator uses, and are not swamped in reality by stuff that the simulator doesn't model, then you'll get something usable. When you know what you're doing, that tends to be the case.

However, when the things you try to study are not in your models, "knowing what you're doing" can mean many things, like not running the simulation at all since you know it will be useless, or hacking subcircuits from measurements to get a better model, or running it anyway, and comparing with real measurements, and studying the interesting differences.

But don't spit on simulations. Without them, you wouldn't be sitting in front of your computer, because it could never have been designed.

Stuff that SPICE lacks :

- thermal effects (self-heating, thermal inertia, thermal runaway, influence of temperature on transistor hFE isn't well modelled)
- Gummel-Poon is full of holes, like Cbc variation with Vce, etc
- Some transistor models are really suspicious
- Doesn't model layout, ground loops, etc (Protel does signal integrity and crosstalk modelling though)
- MOSFETs and JFETs aren't very realistic
- all transistors are perfectly matched (in real life, you'll get offsets)
- etc

You mention layout, we had a group dedicated to simulating package, board, and backplane effects at one large company where I worked. There are advanced simulation tools for modelling interconnects.

I did crude interconnect/layout simulations in SPICE, years ago, where to get reasonably accurate results showing ringing and ground bounce I had to include bonding wire, lead frame inductance and a transmission line model for the board trace. The model is only as good as what you give the simulator.

It is a lot of work and the advanced tools are expensive.

Pete B.
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Need help with Spice simulation overmind Everything Else 4 23rd December 2002 05:58 PM


New To Site? Need Help?

All times are GMT. The time now is 09:55 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2