Circuitmaker got a lot better after a while, but the appeal to me was not having to type out netlists. However, my friend Mark conjectured that the program began as someone's senior project, and then they decided to make it a product and get rich LOL.
However, the bugs are hilarious. Many times fourier will fail (one of the favorite error messages: Line too long). However, I discovered, having pasted in a cascade of bootstrapped twin-t notch filters to see residuals anyway, that merely placing this on the schematic allowed fourier to run!!
There are hideous errors in the models, but these can be edited at least. Many of the P channel DMOS have polarity errors in the threshold voltages for example! And I discovered just the other day that Bode and transient analysis for some of these DMOS parts give wildly disparate results. And the plotting routine often chops off the top of the display, although it can be coaxed into a proper display by changing the x axis settings.
However, at least the darlington models have two Vbe's!
However, the bugs are hilarious. Many times fourier will fail (one of the favorite error messages: Line too long). However, I discovered, having pasted in a cascade of bootstrapped twin-t notch filters to see residuals anyway, that merely placing this on the schematic allowed fourier to run!!
There are hideous errors in the models, but these can be edited at least. Many of the P channel DMOS have polarity errors in the threshold voltages for example! And I discovered just the other day that Bode and transient analysis for some of these DMOS parts give wildly disparate results. And the plotting routine often chops off the top of the display, although it can be coaxed into a proper display by changing the x axis settings.
However, at least the darlington models have two Vbe's!
Spice models supplied by the manufacturers are often very inaccurate. I wouldn't expect the ones in the simulator libraries to be any better, ir more consistent. On the upside, Bob Cordell went to a lot of trouble to make decent SPICE models for a number of transistors. I've got a lot more confidence with those than the ones provided with sim software. You can download them free here: CordellAudio.com - SPICE Models It doesn't include everything you ever wanted, but is quite comprehensive, and certainly better than nothing.
His naming convention wasn't that well thought out though; he appended "C" (for Cordell) to the end of each part number to distinguish them from other models. Unfortunately, lots of parts end in "C" anyway e.g. BC560C, so it's a bit confusing. I edited the text file and prefixed all the part numbers with "bob", so they're grouped together in the lists.
I know that feeling, but at least you've learned a lot along the way.
Our sim results often seem to be totally different.
What kind of sim results are you getting?
It's a good thing I only ordered the soon-to-be-obsolete MPSA77 transistors so far.... If the sim is completely off, then I'm not building it.
I'm sure, but now there isn't a free version anymore (that I'm aware of). Bode plots were the bitch that bit me. When I got totally different plots for the voltage across a resistor, and the current through it, my suspicions were confirmed.
The model for the BC560C was not much different from the one in Tina. I made the changes though it didn't make any big difference in the circuit. I don't have time to mess with this right now, sorry. I can say that it performs much better at +/-32VDC than at +/-15VDC or +/-25VDC. The currents were all the same for each power supply, but the distortion went way down with the higher voltage.
In general I seem to be able to get lower distortion easier, but what prompted my comment was the circuit you showed in post 140. IIRC, when I simmed it, distortion at 20KHz was about 50 times worse than at 1KHz whereas in your sim it was only about 10% worse at 20KHz. I was using some other random Darlington as I don't have MPSA27 in my library, so I wasn't expecting distortion to be the same, but did expect the basic behavior to be similar.
In retrospect in wasn't even remotely an apples to apples comparison, partly due to the wildly different models, and partly because my biasing was probably way out - I just copied your resistor values, instead of changing them to set the idling currents correctly (bearing in mind I didn't know what the currents were supposed to be).
P.S. When you quote distortion at 1V peak-to peak, is that input or output?
Last edited:
I'm sure, but now there isn't a free version anymore (that I'm aware of). Bode plots were the bitch that bit me. When I got totally different plots for the voltage across a resistor, and the current through it, my suspicions were confirmed.
CM went unsupported for a while, so one did not feel guilty getting bootleg copies (although I bought mine fair and square, and got the reasonably helpful book along with the CD ROM). Someone told me that Circuitmaker is now back, but said the emphasis seems to be on educational markets.
An empire-builder at Harman was trying to mandate that all of us (even consultants) had to use a very expensive Orcad package. I showed some of them how I could run circles around them and get good results with CM, presuming of course that one knew enough to question things when they seemed bizarre. Of course empire builders will not be disuaded in their agendas. Since that group, the only part of the Consumer Group actually making money, was disbanded and dispersed (I'd been kicked out well before this) by the brilliant new CEO (who awards himself 30 million dollar bonuses annually), some of the folks use CM now in their other worlds.
On models: beyond the table of parameters, simulators still do different things with those data. For example, Nelson Pass helpfully pointed me to one of Cordell's JFET models. But using it wasn't any more accurate, and I discovered that CM switches abruptly from the so-called triode region to a fixed-slope saturation one! So the results for a lot of sims where the drain impedance was important were bogus, and worse still, generally too optimistic. Bad bad Circuitmaker!
I find that more and more I need to conjure models out of math functions, which have their own bugs. But so far I've not been frustrated enough to start to become a programmer, a pursuit that I could see long ago would change my life, and I believe very much for the worse.
That's input voltage.
The MPSA27 is a small TO-92 darlington with a beta of 10,000. The currents in the VAS and the output stage are now all about 12.4mA.
I think you will be vastly better off losing the darlingtons altogether. I did some sims (yes I know I said I was going to go away for a while, but it was of interest to get some confirmation of my hunches about turnoff time issues with darlingtons) and the difference between the diamond buffer and the darlington output stage at high frequencies is quite striking. But again, you had a defective darlington model which would not show this, and as well as Godfrey and others have suggested, you may do quite well enough with a simple complementary emitter follower, given what you are driving.
As I mentioned in another thread, I've been catching up only about 35 years late reading back issues of The Audio Amateur. Early this AM I got to reading Pass's class A 40W amp project article, which is interesting as one can read his early directions toward what has become the "Zen" phenomenon.
But the amplifier uses darlington devices in the output stage. Lacking a symbol for these distinct from single devices I gaped at the schematic initially, but upon reading learned of their composite structure. But even more interesting: they were Lambda parts, developed for their power supplies, and featured 25 ohm resistors internally across the output device base-emitter junctions, lower than any other integrated power darlingtons I'd heard of. Why so low? So that the first device would run at a healthy current, and so that the output device would turn off quickly.
But the amplifier uses darlington devices in the output stage. Lacking a symbol for these distinct from single devices I gaped at the schematic initially, but upon reading learned of their composite structure. But even more interesting: they were Lambda parts, developed for their power supplies, and featured 25 ohm resistors internally across the output device base-emitter junctions, lower than any other integrated power darlingtons I'd heard of. Why so low? So that the first device would run at a healthy current, and so that the output device would turn off quickly.
I just now tried replacing the transistors in the VAS and buffer with BC327 and BC337, then rebalanced the LTP, and the distortion was almost twice as bad. I increased the current in the VAS and the distortion got a little better. I replaced the buffer transistors with the MPSA77 and MPSA27 and the distortion was still bad. I replaced the transistors in the VAS with the darlingtons, rebalanced the LTP, and the distortion came down to 2.6x10-5%. I don't know what to do. I don't have equipment that can actually measure this kind of distortion. If I can't believe the simulation, then I might as well quit and stick with op amps and gain clones.
I saw the discussion in another thread about Harmon going down the tubes. Sad.
There was an interesting experiment last year in one of Bob Cordell's threads where various people simmed the same power amp with different software and compared results. Interestingly, almost identical results were obtained with LTspice, MicroCap, Simetrix and Cadence, when the same models were used. That gave me more confidence about using Simetrix when almost everyone else on the forum swears by LTspice. Final results are here. (post 3953)
But you're still using it?
I saw the discussion in another thread about Harmon going down the tubes. Sad.
There was an interesting experiment last year in one of Bob Cordell's threads where various people simmed the same power amp with different software and compared results. Interestingly, almost identical results were obtained with LTspice, MicroCap, Simetrix and Cadence, when the same models were used. That gave me more confidence about using Simetrix when almost everyone else on the forum swears by LTspice. Final results are here. (post 3953)
But you're still using it?
I may have to adopt LTSpice as it is rapidly becoming a de facto standard. But CM is the devil I know. Were it not for many years of designing with pencil and paper and calculators, I might not find it adequate. Again, this is a later version than the widely-circulated freeware, which was very limited and even more buggy.
darlington output simulations
Here is a schematic, and sim results, for a 10kHz 50ns rise/fall square wave input to the darlington output stage alone, with a 1k source resistance. Note the curious and long-lasting initial ramps on the tops and bottoms of the waveforms. This is due to devices not turning off quickly, and worse, corresponds to current flow that will make the power rails sag as well.
Not shown here, but when the input is a 100kHz sinusoid of similar p-p amplitude, the distortion is 1.36% and the stage self-biases to about twice the quiescent current after a few cycles! Not at all good. Of course, one could argue plausibly that most audio signals won't exercise things this much, but the effects on distortion show up quite a bit already at lower frequencies and levels.
Here is a schematic, and sim results, for a 10kHz 50ns rise/fall square wave input to the darlington output stage alone, with a 1k source resistance. Note the curious and long-lasting initial ramps on the tops and bottoms of the waveforms. This is due to devices not turning off quickly, and worse, corresponds to current flow that will make the power rails sag as well.
Not shown here, but when the input is a 100kHz sinusoid of similar p-p amplitude, the distortion is 1.36% and the stage self-biases to about twice the quiescent current after a few cycles! Not at all good. Of course, one could argue plausibly that most audio signals won't exercise things this much, but the effects on distortion show up quite a bit already at lower frequencies and levels.
You have your current sources connected to ground whereas mine are connected to the + and - rails. I don't know if that makes any difference.
thanks again for your help. I don't know enough about these devices to understand a lot of what you're saying.
Last edited:
Believability isn't a yes or no thing, it comes in degrees. I think simulators are very useful and can give fairly accurate results if good models are used. But even if the sim can't tell you exactly what the distortion of your amp is, it can still tell you if this circuit has much lower distortion than that circuit.
The simulator can also be very useful when looking at other aspects besides distortion, e.g. PSRR, slew rate, stability etc.
No difference. Of course in "real life" a current source/sink would be deriving its electrons from one of the power rails.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.