So have this spice model been compared to the real thing to see how much reality is in the model? That is the question on my mind.
PB2, I recently loaded this amp into LTSpice but lost it a month ago when laptop died! This is great. For what it's worth, my simulations ran well but I never finished tweaking.
Thanks again.
Thanks again.
My spice models do what I tell them to do and sometimes they do what they want to do. Never can compare subjective sound to computer screen. There is no way. Simulators are amazing tools. Even more fun ensues if you like the results when you actually build and tweak some more.
Pete the DC bias of the input stage is not balanced, so this effects the output stage DC bias as well. Is the 2SC diffamp model good? Seems the gains might be too low?
Is there a tutorial on modeling transistors? or maybe using a simple static model ie hybrid pi? s/b good enough for frequency sweeps Bode plots.
Is there a tutorial on modeling transistors? or maybe using a simple static model ie hybrid pi? s/b good enough for frequency sweeps Bode plots.
Last edited:
Value of R3 on the schermatic is surely wrong, as it s written 33.2K..
I suppose it is 33.2R , and i did sim using this latter value..
At first try, it did oscillate..
To suppress this behaviour, i had to remove the lead compensation
capacitor C4 and increase the lag compensation capacitor C2 to 47pF
to compensate for the overshoot, although it wasn t enough to
remove it completely with these modded values..
Anther solution is to reduce C4 to 4.7pF, as the value of 39pF displayed
on the schematic seems way too high for this compensation scheme..
In this latter case, C2 can be increased to 33pF to suppress any overshoot.
This solution seems more satisfactory.
I suppose it is 33.2R , and i did sim using this latter value..
At first try, it did oscillate..
To suppress this behaviour, i had to remove the lead compensation
capacitor C4 and increase the lag compensation capacitor C2 to 47pF
to compensate for the overshoot, although it wasn t enough to
remove it completely with these modded values..
Anther solution is to reduce C4 to 4.7pF, as the value of 39pF displayed
on the schematic seems way too high for this compensation scheme..
In this latter case, C2 can be increased to 33pF to suppress any overshoot.
This solution seems more satisfactory.
Last edited:
Wahab you edited your post after I responded to your question about R3 being in the Kilo-ohms, adding everything else after.
Not according to Pete's model, if it did, you must be using your own model/s then>
Why did it oscillate, since the real ones don't? Seems better to find out and fix your model, before simply changing the compensation. It would also behoove one to validate such finding of the spice sims with real tests to the hardware before recommending such changes.
.
Yes I got that and was not suggesting the model could be "virtually" listened to. I just wondered after tweaking if the model and the amp act at all the same. There is a lot of model work on these forums and having used models for circuits and transducers I found the models left a lot to be desired. Several of the circuits used here I could not even get to work in the model but that may have been that software or my lack of circuit model skills. So you seem like one who realizes the model needs "adjusting" to match reality and wondered if after calibration of the model if it follows the amp behavior well? That is it.
Thanks- sum
Hi infinia,
Often the real compensation components will differ from what you get in a simulator. They might work out to be the same if you also modeled all the stray effects - exactly. Obviously too much work for any normal human!
Hi PedroDaGr8,
That is very kind of you, thank you for the offer. I'll send you mail.
-Chris
Often the real compensation components will differ from what you get in a simulator. They might work out to be the same if you also modeled all the stray effects - exactly. Obviously too much work for any normal human!
Hi PedroDaGr8,
That is very kind of you, thank you for the offer. I'll send you mail.
-Chris
I think you should NOT let a designer loose on a repair job as much as a repair tech should be not in charge of a design job.
Remember folks we have here an already well designed amp that has been well documented and built in commercial quantities. If this is being currently simulated as an oscillating design we need to re-check the models and re-sim. BTW I don't think Pete's model has a sim that is oscillating tho.
Chris regarding audio amps, the stability compensation design values really don't differ too much from accurate models. If they do what's the whole point of the modeling anyway?
There are other reasons a good designer usually builds in extra wiggle room on phase margin and gain margin and it should be very rare to get an oscillating design on the repair bench unless you are doing something very silly. This is also the main reason for using dominate pole compensation > keeping you out of danger of wide component tolerances and even to a lesser degree the extra parasitics. Otherwise when going extra wide-band you tend to have an extra lead in your back pocket anyway. I use to do this by hand with 6 cycle log log graph paper using the HF transistor models checking with max/ min gains. You can pretty much account for extra poles due to known transistor parasitics very well. The spice sims are really good at this IF you have verified the transistor models.
Finally testing in the lab for closed loop BW and with small signal step responses or square waves gives the reasonable assurance your models are good. Large signal V steps will also check bandwidth and/or slewing estimates.
I seem to recall Self covering the diff pair imbalance on his web site suggesting that a mirror on the diff pair is the solution. The DC OLG will be much higher with the mirror. I just plugged in the models, did not check them in anyway - I think they are good, you can try other transistors and will probably see similar results.
Several here have been working on modeling. Andy_c did some excellent work here:
Improved SPICE Models for MJL3281A and MJL1302A - Section 1
There are other threads here, you could try a search. I would sure like to see an easier method.
It don't see any oscillations, he changed the sim model then claimed it oscillated to support his claim way back that the comp caps should change. He is just playing games as I think you too have noted.
I would expect the model to be quite good for predicting stability since the poles are set by the stage operating points and the cap values used for compensation. This assumes that the caps are larger than the transistor parasitics, or that the parasitics are reasonably accurate if not.
The faster output devices are an issue but since it is unity gain it may not matter all that much.
For some reason I'm really bothered by that diff-amp imbalance it really affects the output stage biasing. I'm sure you saw it too when you looked at biasing the last stage of your model. It should rear its head again on crossover distortion sims. I'm probably going to play with the first stage a bit more and looking at the other references you gave me as well.
Pete thanks for your work and help on this amp modeling exercise! I'm learning something.
Q> If someone asked you what the models nominal Hfe is at the particular operating bias current, say for example the diff-amp pair. How could we determine the answer and how could we account for manufacturing tolerances of that parameter during a sim? looking for stage gain variation and pole changes.
So you are smart, and i m an idiot playing games?..
Please, spare me your arrogance, since it s not even backed by an
evident knowledge..
So if I change models, the amp will change his beheviour, but
at the same time you pretend that the caps are the dominanting
components for the amp stability...
That is bad faith, no more..
Everybody with the slightest knowledge in compensation
network would realize that the lead compensation cap is
way too high, and that the output stage has no influence
over it, as it s took at the vAS output.
Nelson Pass did use this compensation scheme in his 40/70W
class A amp that was published in the late 70s..
I did check stability using a step , not a sine wave..
Devices used to sim : 2SA970/2SC2240 2SA1360/2SC3423
2SA1837/2SC4793 2SA1943/2SC5200
I already have andy_c models, and they doesn t make any difference.
Last edited:
No, this amp is not well designed, wether this is mr Pass or
anybody else that did the work...
A single dual differential amp should work flawlessly at first try,
as it s a proved design, the most used on earth, but this
particumliar model seems to concentrate many bad habits
of the late 70s...
wahab maybe you can contact the manufacturer and tell them their many thousands amps in the field have a problem, I'm sure they would be interested in your solution.
Me I own a couple and they work just fine! thanks for your help, not.
Me I own a couple and they work just fine! thanks for your help, not.
Last edited:
This amp seems to be device dependant, so it s no surprise
that yours are running well, but at the same time, there s much
people who have trouble with it , as witnessed at DIY audio and
other forums..
That would a another false conclusion by just another opinionated person with a obvious chip on the shoulder. Try this for size> That they are valued by DIY folks who know audio and just what there getting for an honest dollar. Additionally that these amps are valued enough to be kept in service for so long and would rather be repaired when abused than tossed aside like most of the other audio products of that era.