I suspect with enough digging on the simulator, lots of little quirks, hidden loops etc can be extracted, that on paper could cause a problem. Whether they do so in real life is another matter.
That said, you always need to take EF3 parasitic and cascode oscillation seriously. That's my experience.
That said, you always need to take EF3 parasitic and cascode oscillation seriously. That's my experience.
Douglas, feel free to use any models of your choice to establish the veracity of my averment that more than two transistor enclosed in the minor loop likely makes said loop unstable, and that, therefore, the minor loop itself requires compensation as a result.
What are your gain and phase margins in the compensation loop? What is its ULGF?
Have you SPICE'd it?
The problem with this circuit is that the current output of the IPS no longer sees a low-impedance node at the input of the VAS (as normal Miller compensation causes it to do). I believe that you yourself, in one of your books, in touting the advantages of good old Miller compensation, stated that the low impedance at the input to the VAS largely prevents a potentially problematic pole from occurring there (that node largely being like a virtual ground due to the shunt feedback of Cdom).
As you should know, simulations being right and reality being right are not mutually exclusive. Each is done based on certain parameter values and model assumptions. There are cases where simulation will be optimistic and there are others where it will be pessimistic.
One should never take comfort in a circuit working in reality when a PROPERLY DONE simulation shows slim stability margins.
Just because you make a thousand pieces that don't oscillate does not mean that you have good stability margins in your local loop. Would you be comfortable as a designer if you had a crystal ball and knew that every one of those amplifiers had a compensation loop with a phase margin of between 15 and 20 degrees? I certainly would not.
As Harry also mentioned this form of compensation, I hope he will chime in.
I will be the first to admit that I may be too conservative in thinking that some additional compensation may be needed in that loop to make it bullet-proof. Mind you, we are talking about stability in a much higher ULGF range than normally associated with a power amplifier global loop. This compensation loop can probably be fime with a ULGF of perhaps 20MHz.
Cheers,
Bob
Since all my TAG-McLaren amplifiers were undoubtedly stable, and remained so in service, I don't accept that there were any 15-degree phase margins lurking in the circuitry. Reality is always right, but simulations often aren't and I can see nothing to be gained by starting a research program to find out why the simulations appear to be wrong in this case. As I said, there seems to be much doubt about the device models.
I have literally lost count of the number of power amplifiers I have put into quantity production, including many in China where I could not have personal oversight of the test procedures. None of them have ever shown stability problems.
In the light of that I would ask you if you really think that your lecturing tone is appropriate.
As I said run the simulations with models of your choice. Alternatively, provide us with models of your choice and I will run the simulation again. I certainly don't think the quality of my models has anything to do with it.
Last edited:
As in any engineering design software there are limits to what can and should be done. The models for the electronic components that are under test can be accurate at the same time they can be lacking in fine details. This is always the problem with this type of design. You can get a basic answer to an engineering problem but until you test an actual unit under all test conditions you can not guarantee that the model reflects real life operations. Yes as time goes by the software and the models usually increase accuracy as more detail is added, but so much of the electronics side is lacking in detail on a simulation side it seems. I think that both D. Self and B. Cordell have valid arguments for their particular design preferences but I bet both would never let a product out without proper testing of a real circuit. I'll take any of these devices and test it at my house, with a 50K watt radio station within about 100 yards and all the microwave and what looks like doppler radar weather stations I bet I will find an RF problem long before any simulation will find a problem.
Hi Doug,
I think the questions I raised were entirely appropriate, and no lecturing tone was intended.
I was hoping that you would give us some insight as to why that circuit has adequate margin against instability in your application and what you may have done to achieve that. Instead, you cast doubt on the use of SPICE simulation to ferret out possible problems. I did say "PROPERLY DONE" SPICE simulations. I certainly agree that SPICE, like any tool, is subject to garbage-in, garbage-out.
Cheers,
Bob
It's much more simple than that: simulation only is cheap, and paper amplifiers always work.
Simulation is a good base for discussions, though. Unfortunately, replicating simulation results is not that simple as it is believed. Adopting new models, cross platform simulators (including re-capturing schematics), schematic debugging, results interpreting and comparing, etc... are rather time consuming.
For example, it's easy to say "use my models" - but this involves creating new devices for the specific simulation platform, and for what? Just to find out that the numbers are the same (which is guaranteed, anyway, algorithms are deterministic and solutions are unique)? How would one know that the models used by X are actually anywhere better than the models used by Y, other than "because I say so"?
Hi Guys
Reality tends to trump virtual every time, especially when it's a 10k+ vs 1 or 2 of the other.
Amazon lists APAD6 for advance order at a nice price, with a pic of the new cover but sadly a word for word description of ed.5. Maybe the publisher needs a kick to get the right description to the distributors? Certainly would help the drool factor.... not virtual.
One sale here for sure.
Have fun
Kevin O'Connor
Reality tends to trump virtual every time, especially when it's a 10k+ vs 1 or 2 of the other.
Amazon lists APAD6 for advance order at a nice price, with a pic of the new cover but sadly a word for word description of ed.5. Maybe the publisher needs a kick to get the right description to the distributors? Certainly would help the drool factor.... not virtual.
One sale here for sure.
Have fun
Kevin O'Connor
Possible, but that's my perception as well. I would call it "pompous" rather than "lecturing", though.
I have run the simulation with several alternative models and the results are still the same: uncompensated unity minor loop gain phase shift exceeds 200degrees. Perhaps you can provide us with models you trust?
Hi Guys
Both amazon.ca and amazon.com have the same advance price discount, and the same erroneous description.
Bob is normally pleasant and quite thoughtful in his posts, but I thought he seemed preachy in the post in question. Considering the rather circular arguments that have gone on here, all based on simulation, it would be easy for someone even as smart as Bob to get pulled into the vortex.
I don't believe this has to do with the position of the Atlantic Ocean. On the other hand, there has been a lot of solar activity...
Have fun
Kevin O'Connor
Both amazon.ca and amazon.com have the same advance price discount, and the same erroneous description.
Bob is normally pleasant and quite thoughtful in his posts, but I thought he seemed preachy in the post in question. Considering the rather circular arguments that have gone on here, all based on simulation, it would be easy for someone even as smart as Bob to get pulled into the vortex.
I don't believe this has to do with the position of the Atlantic Ocean. On the other hand, there has been a lot of solar activity...
Have fun
Kevin O'Connor
I have run the simulation with several alternative models and the results are still the same: uncompensated unity minor loop gain phase shift exceeds 200degrees. Perhaps you can provide us with models you trust?
Michael, if you are referring to the schematic you posted here, no simulations are required to explain why you got instabilities in the minor loop. You did not degenerate the VAS (or TIS if you prefer) - which would be highly unusual for any practical implementation.
I have run the simulation with several alternative models and the results are still the same: uncompensated unity minor loop gain phase shift exceeds 200degrees. Perhaps you can provide us with models you trust?
I think I have made it quite clear that I have no interest in finding out what is wrong with the simulations. Reality is the test.
Sims vs 'real life' vs How to muck up someone else's design
If I may repeat what I've said before ad nauseum, 'the only reason we do dis sims is cos we believe they have some resemblance to real life'. I have great respect for Bob & Doug, not cos their pontificating sounds authoritative .. but cos they back up their pet theories & prejudices with 'real life' examples.
'Real life' is the true measure of the worth of dis sims. If someone designs a production amp which my sims show is wonky .. then provided I trust him to do the proper tests .. and I put both Bob & Doug in this august group ... it is up to ME to find out why his 'real life' stuff works and why my sims are wonky.
This is the Chinese statement of 'Scientific Method'. The theory (sims etc) has to fit 'real life' .. not the other way round.
It's perfectly valid to sim one of Doug's 'real life' amps provided you use models as close to what he uses as possible. If you then replicate his 'real life' test results for THD & stability, you can be fairly confident your sim is within striking distance of 'real life'
You can then quite validly claim that eg 'using 2n3055 in the output stage results in instability & the release of Holy Smoke'. Your sim of small changes to his circuit will then have contributed to the store of human knowledge.
But if you take one of his designs and
Michael, I have not downloaded ALL your examples ... but those that I have certainly ... exhibit the above features. In addition, they appear to have wonky stability (understandably from the above list) ... and though quite complex, have performance which would be considered blah pre 1980, let alone in da 21st century.
Bob has chapters in his book on validation of models. It would help the credibility of your claims if you say a bit more than 'I've used them for years to troubleshoot real circuit problems succesfully.' Were these switching circuits? A useful answer please.
It would help your credibility even more, if you post a complete circuit of YOURS that demonstrates what you consider good practice & performance. This is preferably a 'real life' device but even a complete SPICE model would be an improvement over what you have shown so far.
This would demonstrate you have some idea of how to do a good amp. As it is, you have only demonstrated you know how to muck up someone else's design.
If even YOU won't vouch for the performance of your circuits, why be surprised that other people won't waste time working on them.
If I may repeat what I've said before ad nauseum, 'the only reason we do dis sims is cos we believe they have some resemblance to real life'. I have great respect for Bob & Doug, not cos their pontificating sounds authoritative .. but cos they back up their pet theories & prejudices with 'real life' examples.
'Real life' is the true measure of the worth of dis sims. If someone designs a production amp which my sims show is wonky .. then provided I trust him to do the proper tests .. and I put both Bob & Doug in this august group ... it is up to ME to find out why his 'real life' stuff works and why my sims are wonky.
This is the Chinese statement of 'Scientific Method'. The theory (sims etc) has to fit 'real life' .. not the other way round.
It's perfectly valid to sim one of Doug's 'real life' amps provided you use models as close to what he uses as possible. If you then replicate his 'real life' test results for THD & stability, you can be fairly confident your sim is within striking distance of 'real life'
You can then quite validly claim that eg 'using 2n3055 in the output stage results in instability & the release of Holy Smoke'. Your sim of small changes to his circuit will then have contributed to the store of human knowledge.
But if you take one of his designs and
- change ALL the devices
- resulting in operating points & currents that are well outside his recommendations
- remove several items that practically guarantee the amp will be unstable with 'real life' loads
- make other major changes to the circuit which you claim are trivial and have no effect on THD or stability .. but other gurus (eg Cherry & others) & pseudo gurus (eg me & others) feel are important
- then claim the circuit is unstable/wonky/evil
Michael, I have not downloaded ALL your examples ... but those that I have certainly ... exhibit the above features. In addition, they appear to have wonky stability (understandably from the above list) ... and though quite complex, have performance which would be considered blah pre 1980, let alone in da 21st century.
Bob has chapters in his book on validation of models. It would help the credibility of your claims if you say a bit more than 'I've used them for years to troubleshoot real circuit problems succesfully.' Were these switching circuits? A useful answer please.
It would help your credibility even more, if you post a complete circuit of YOURS that demonstrates what you consider good practice & performance. This is preferably a 'real life' device but even a complete SPICE model would be an improvement over what you have shown so far.
This would demonstrate you have some idea of how to do a good amp. As it is, you have only demonstrated you know how to muck up someone else's design.
If even YOU won't vouch for the performance of your circuits, why be surprised that other people won't waste time working on them.
Last edited:
If I may repeat what I've said before ad nauseum, 'the only reason we do dis sims is cos we believe they have some resemblance to real life'. I have great respect for Bob & Doug, not cos their pontificating sounds authoritative .. but cos they back up their pet theories & prejudices with 'real life' examples.
'Real life' is the true measure of the worth of dis sims. If someone designs a production amp which my sims show is wonky .. then provided I trust him to do the proper tests .. and I put both Bob & Doug in this august group ... it is up to ME to find out why his 'real life' stuff works and why my sims are wonky.
This is the Chinese statement of 'Scientific Method'. The theory (sims etc) has to fit 'real life' .. not the other way round.
It's perfectly valid to sim one of Doug's 'real life' amps provided you use models as close to what he uses as possible. If you then replicate his 'real life' test results for THD & stability, you can be fairly confident your sim is within striking distance of 'real life'
You can then quite validly claim that eg 'using 2n3055 in the output stage results in instability & the release of Holy Smoke'. Your sim of small changes to his circuit will then have contributed to the store of human knowledge.
But if you take one of his designs and
... you only demonstrate that your choice of devices and perhaps your understanding of some very basic issues is inadequate or inappropriate.
- change ALL the devices
- resulting in operating points & currents that are well outside his recommendations
- remove several items that practically guarantee the amp will be unstable with 'real life' loads
- make other major changes to the circuit which you claim are trivial and have no effect on THD or stability .. but other gurus (eg Cherry & others) & pseudo gurus (eg me & others) feel are important
- then claim the circuit is unstable/wonky/evil
Michael, I have not downloaded ALL your examples ... but those that I have certainly ... exhibit the above features. In addition, they appear to have wonky stability (understandably from the above list) ... and though quite complex, have performance which would be considered blah pre 1980, let alone in da 21st century.
Bob has chapters in his book on validation of models. It would help the credibility of your claims if you say a bit more than 'I've used them for years to troubleshoot real circuit problems succesfully.' Were these switching circuits? A useful answer please.
It would help your credibility even more, if you post a complete circuit of YOURS that demonstrates what you consider good practice & performance. This is preferably a 'real life' device but even a complete SPICE model would be an improvement over what you have shown so far.
This would demonstrate you have some idea of how to do a good amp. As it is, you have only demonstrated you know how to muck up someone else's design.
If even YOU won't vouch for the performance of your circuits, why be surprised that other people won't waste time working on them.
Pow. Ouch! geez. Wow. slam. bam. wham. Ohhh. plop. thud. silence.
-RNM
Sims V real world
Sims "can" be good, in fact Very good, & where would we be without them !
BUT, they are NOT real world built up tested circuits, with real components & PCB's & wiring etc etc.
Supposedly identical components from different manufactures, can & do vary in numbers of various ways. Sometimes it "might" not matter, in other cases it "could" be critical to performance.
ASAIK sims don't take account of PCB parasitic etc matters, & track inductance etc. This "could" make a difference.
Wiring to & from the PCB can be made to look very neat & tidy, but it "might" not be optimum to performance. Sims don't account for this.
Also possible Actual electromagnetic/RFI etc influences on the circuit/components aren't simmed.
So the ONLY way to establish if something performs as per simmed, is to build & test it.
Sims "can" be good, in fact Very good, & where would we be without them !
BUT, they are NOT real world built up tested circuits, with real components & PCB's & wiring etc etc.
Supposedly identical components from different manufactures, can & do vary in numbers of various ways. Sometimes it "might" not matter, in other cases it "could" be critical to performance.
ASAIK sims don't take account of PCB parasitic etc matters, & track inductance etc. This "could" make a difference.
Wiring to & from the PCB can be made to look very neat & tidy, but it "might" not be optimum to performance. Sims don't account for this.
Also possible Actual electromagnetic/RFI etc influences on the circuit/components aren't simmed.
So the ONLY way to establish if something performs as per simmed, is to build & test it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.