Make a folder for new models inside wich you put generaly
a text containing the parameters , then :
Simetrix --> File ---- Model library ---> Add/remove library.
Symetrix will open a browsing window ,go to your new folder and double click,
simetrix then display the new files , click on the folder name and then on "add".
Here, most models are not very good, use these from Cordell:
CordellAudio.com - SPICE Models
You can contact me as I may have better models of popular transistors that aren't in that list.
If it helps, I have a SPICE model of the 2N5551 on my web site that I developed while I was writing my book "Designing Audio Power Amplifiers". See CordellAudio.com - Home.
Cheers,
Bob
Much thanks Bob. I am working my way through your book to learn a bit about amps. Amps just plain got me intrigued. They are a lot more subtle than they look. The problem is that books don't answer questions.
I would guess you have the 5401 as well as you used that in the book. I am playing with an old Hafler DH-120 as a learning experience. The way it is built, it is hard to probe, so learning Spice seemed reasonable.
BTW, if you are not using LTspice, I urge you to try it. It is free and in my opinion maybe the best. I use nothing else. Everything in my book was done with LTspice, and my two chapters on SPICE simulation are LTspice-centric and can help one learn LTspice. There is also info on my web site about downloading and getting started with LTspice, and also ready-to-run LTspice amplifier simulations. You can literally be LTspice'ing in ten minutes.
Best,
Bob
Quick update. I had to make a couple small edits on Bob's models to run in Simetrix. No "mfg" tag for example. I also dug out the Exicon models. I am pretty much getting a hang of it. The function that I can't get use of is the pole zero function. It spits out a list of complex numbers and sort of says, have a ball. It would be very nice if one of these tools actually provided Bode plots. I may try to move up to LTSpice eventually now I have a little clue what I am looking for. I have a few more functions to try.
At the beginning, the quality of your simulations will depend more on your tenacity and attention to detail, than the simulator used. You should verify your models and see in what areas they aren't realistic. Otherwise, you will not know what you're missing. Even Cordell's models don't model quasi-saturation, so transistors will behave overly nice at low Vce.
I don't know about P-Z analysis, I don't even know what it is, but in LTSpice I just use the AC analysis to see bode plots. LTSpice also has Cartesian and Nyquist mode (I understand Bode and Cartesian, but Nyquist is still quite mysterious to me). But surely Simetrix has an AC analysis or bode plotting function.
Possibly "pole-zero analysis" - i.e., extracting an analytical Fourier or LaPlace (s-domain) expression from amplitude and phase data.
Many incarnations ago I could locate a dominant pole, and maybe 1 or 2 others, by looking at a Bode plot. It wouldn't surprise me if this kind of graphical analysis is still discussed in AC networks texts, but I don't know of one offhand. I seem to recall that the topic surfaces from time to time in the LTSpice Yahoo group but the responses are not encouraging. See, e.g., < http://tech.groups.yahoo.com/group/LTspice/message/53761 >.
Dale
I've never used P_Z analysis in Spice - not particularly human readable output and the numerical algorithms are often poor
even though I found the reverse, the behavioral source Laplace() function can occasionally be helpful - but numerically problematic in .Tran
Bode Plot is the most often used Classical Frequency Response view - think of it as an "isometric" view of the 3-d magnitude, phase vs frequency plot (with a little processing to separate log magnitude, and phase )
in EE, Control Theory we have lots of textbook, tutorial examples of how to use Bode Plots, Bode asymptote approximations
Nyquist is the view straight down the frequency axis, we mostly care about the stability region near |1|, 180 degree phase point - so the useful scale is linear - but gain/attenuation still ranges orders of magnitude so simple plotting SW is often frustrating about the axis ranging
In my experience Nyquist is nearly ignored in practical day-to-day EE usage
Bode plots of the "simple Middlebrook" loop gain probe is what I use most often in exploring stability of feedback loops in sim - with the understanding that .AC analysis in Spice is very limited
"small signal" linear stability is required but not that hard to achieve at nominal output - the more interesting issues can only be looked at in .Tran analysis - rail sticking, nonlinear oscillation in clipping, recovery
and our Spice models all fall short of the real world circuit's complexity so "illustrative" indications of trends, sensitivities are more what you really get from sim rather than hard prediction of exact compensation component values
even though I found the reverse, the behavioral source Laplace() function can occasionally be helpful - but numerically problematic in .Tran
Bode Plot is the most often used Classical Frequency Response view - think of it as an "isometric" view of the 3-d magnitude, phase vs frequency plot (with a little processing to separate log magnitude, and phase )
in EE, Control Theory we have lots of textbook, tutorial examples of how to use Bode Plots, Bode asymptote approximations
Nyquist is the view straight down the frequency axis, we mostly care about the stability region near |1|, 180 degree phase point - so the useful scale is linear - but gain/attenuation still ranges orders of magnitude so simple plotting SW is often frustrating about the axis ranging
In my experience Nyquist is nearly ignored in practical day-to-day EE usage
Bode plots of the "simple Middlebrook" loop gain probe is what I use most often in exploring stability of feedback loops in sim - with the understanding that .AC analysis in Spice is very limited
"small signal" linear stability is required but not that hard to achieve at nominal output - the more interesting issues can only be looked at in .Tran analysis - rail sticking, nonlinear oscillation in clipping, recovery
and our Spice models all fall short of the real world circuit's complexity so "illustrative" indications of trends, sensitivities are more what you really get from sim rather than hard prediction of exact compensation component values
I have some strange behavior from multisim, a lot of simulationerrors like in this schematic, put the yellow mesurement on the output and it simulate very strange, voltages and amps are totally strange, much higher idle current, and no signal anymore, see pictures, the one without the yellow measurement on the output works oke.
Multisim has also a lot of faulty models, is this suckware maybe. I seek also a model for the semisouth j-j-mosfets.
thanks in advance.
Multisim has also a lot of faulty models, is this suckware maybe. I seek also a model for the semisouth j-j-mosfets.
thanks in advance.
Attachments
It may mean your circuit is on the verge of oscillation at ridiculously high frequencies. This sometimes happens with MOSFETs in LTSpice. Try using Cordell's K1056/J162 models.
CordellAudio.com - SPICE Models
CordellAudio.com - SPICE Models
still multisim has error or better bug,s, If I copy and paste a part of a other design like a mu stage in a excisting schematic , it wil not work anymore, strange voltages and 100 ampere in a tube for example,, if I copy the whole schematic and past it in a nieuw sheet incl the mustage or other, it works again but with a warning.
this I have as long as I use that software, special in windows 7 it makes problems.
regards
this I have as long as I use that software, special in windows 7 it makes problems.
regards
while working at real EE jobs with the ability to buy $$$$ software I've chosen to use Ltspice for sim
its really hard for me to imagine why not use a free, unlimited nodes/devices Spice designed, supported by engineers, for engineers
if you give Mike a fully documented bug report he will often fix it in the next release, every few weeks usually - I've had one fixed over lunch hour!
the UI is intentionally "flat" - not a production schematic entry tool for a dedicated draftsman - rather a working engineer's tool - used internally at Linear for chip design
the convergence, speed are competitive with the $$$$ tools, there is the Yahoo support group for Ltspice specific issues
it is not a dumbed down learning tool - but if you already know Spice you should be running sims in < 1/2 hr - including download time (in Windows - I don't know how long a Wine install takes)
its really hard for me to imagine why not use a free, unlimited nodes/devices Spice designed, supported by engineers, for engineers
if you give Mike a fully documented bug report he will often fix it in the next release, every few weeks usually - I've had one fixed over lunch hour!
the UI is intentionally "flat" - not a production schematic entry tool for a dedicated draftsman - rather a working engineer's tool - used internally at Linear for chip design
the convergence, speed are competitive with the $$$$ tools, there is the Yahoo support group for Ltspice specific issues
it is not a dumbed down learning tool - but if you already know Spice you should be running sims in < 1/2 hr - including download time (in Windows - I don't know how long a Wine install takes)
Last edited: