Are you saying the Fairchild model is the right one?
This is the Fairchild model : http://www.fairchildsemi.com/models/PSPICE/Discrete/Bipolar_Transistor.html
This is the OnSemi model : http://www.onsemi.com/PowerSolutions/supportDoc.do?type=models
Look for BD139-BD140 models.
This is the Fairchild model : http://www.fairchildsemi.com/models/PSPICE/Discrete/Bipolar_Transistor.html
This is the OnSemi model : http://www.onsemi.com/PowerSolutions/supportDoc.do?type=models
Look for BD139-BD140 models.
lumanauw said:
If you have set an output impedance to the signal source, it is completely correct that the IN becomes distorted by the unlinear load from the bjt.
If you do not have set an output impedance to the signal source, uninstall your simulation software and look for another program...
Mike
Hi, MikeB,
You're right about the output impedance of the signal source.
I don't set anything, so the signal source (Vsin) should be an ideal voltage source (Zout=0).
What makes me confuse, with the Fairchild model, there is no distortion in the input trace. Nothing in the CCT is replaced, just replacing BD139 and BD140.
You're right about the output impedance of the signal source.
I don't set anything, so the signal source (Vsin) should be an ideal voltage source (Zout=0).
What makes me confuse, with the Fairchild model, there is no distortion in the input trace. Nothing in the CCT is replaced, just replacing BD139 and BD140.
AndrewT said:
I don't completely agree, but I see your point. BJT models generally seem safer than some of the MOSFET models. It is also very important when comparing performance from two manufacturers' models to make sure the operating points come out right and equivalent. Some manufacturers don't get the Ic-Vbe characteristic right (I really can't understand why), but that alone can throw things off, yet can be corrected to give some useful results with some judicious tweaking.
I would say, experienced among us design, simulate, build, test, listen, iterate.
Cheers,
Bob
lumanauw said:
I'm glad to hear that you're "taking the plunge" into simulation, David. I think that you will will find it to be quite useful, as well as a viable tool, and a great toy, for learning and experimenting.
So... which simulation software will you be using, until you're persuaded to try LTspice? ;-)
Yes, the Fairchild models are the better ones, for BD139 & BD140. I found that out the hard way, wasting perhaps a whole day, before finding Post #8 in the following thread:
http://www.diyaudio.com/forums/showthread.php?s=&postid=1009747&highlight=#post1009747
Regarding elevated THD of ideal voltage source used as input: One way I can have an elevated THD (above .00001%) for such a source is by forgetting to turn off the data-compression feature, in LTspice, in one of the rare cases where I don't have a spice directive on my schematic, to automatically turn it off.
From my experience with LTspice, it seems that THD calculations can also generally be significantly-affected by the "MAX TIMESTEP" setting for spice's internal "solver". I usually set an initial Max Timestep to something like {1/(1000*freq)}, and then increase the "1000" factor until THD no longer changes significantly.
The Max Timestep also affects ALL of the calculations and modeling that are done, similarly. So it should be set to something less than some fraction of the highest frequency you expect the simulator to be able to resolve, in order to get reliable results.
Regarding your simulator "jamming" when using the Zobel network: If you don't have at least a small value set for the ESRs of the capacitors in your circuit, it can make it more difficult for spice to converge on a solution. The same thing goes for the DCR of inductors. I don't see how the cap ESR could be a problem, with a Zobel, since there's already a series resistance, there. But perhaps there's something else in your circuit that is causing a similar type of problem. I suppose it could even be a Max Timestep that is not small-enough. Along the same lines, if you use a pulse input, make the rise and fall times at least slightly more than zero. Does it do the same thing with the Fairchild models for the BD139 & BD140?
After a short learning curve, I hope that you find using the simulator very rewarding, David.
Hi, Gootee,
Yes, I'm still at the very early page of learning. There are many tricks that I don't know about. I'm using DXP2004 simulator.
With arbitrary manufacturer's model, I can see that SIM could be usefull in knowing the "delta" (change) of the result when a parameter/ or design is changed (Like RDF said). The "absolute" result depends on the accuracy of the model, but the "delta", where the result trend goes to , when something is changed, can be seen immediately using SIM.
Yes, I'm still at the very early page of learning. There are many tricks that I don't know about. I'm using DXP2004 simulator.
With arbitrary manufacturer's model, I can see that SIM could be usefull in knowing the "delta" (change) of the result when a parameter/ or design is changed (Like RDF said). The "absolute" result depends on the accuracy of the model, but the "delta", where the result trend goes to , when something is changed, can be seen immediately using SIM.
Hi Tom,
This actually sounds like an interesting series of articles for EDN to run, and they are hitting their target audience.
-Chris
Contacting EDN is an excellent idea. Since great expense was expended making the simulators available to the world, some effort in producing models seems reasonable.
This actually sounds like an interesting series of articles for EDN to run, and they are hitting their target audience.
-Chris
lumanauw said:
I comment on both the OnSemi and Fairchild BD139/140 here:
http://www.diyaudio.com/forums/showthread.php?postid=1290867#post1290867
At least one of the OnSemi models seems to be completely defective. The Fairchilds are much better, I use them often, but they are far from perfect. The more models I look at, the more problems that I find.
Pete B.
teodorom said:
LOL, yeah I guess I was trying to be generous, LOL!
I'm having trouble finding any good models!
As bad as they are, many are usable, with caution, say in the linear region at least for rough work anyway. Some are completely defective.
I agree, it is a serious problem, they need an independent verify phase before they release models. Very basic QC.
Pete B.