Browsing this morning, and came across this statement:
"Btw. I wonder why all those simulations never attempt to drive a real - that is, a complex -load? They would be amazed what nasty stuff happens to those thd figures...."
From here: http://www.diyaudio.com/forums/solid-state/161293-recovery-clipping-soft-clipping.html#post2085488
So why isn't this done routinely? There is so much discussion and argument about the meaning of distortion and it's relationship to perceived audio quality (including the point that the normal THD figures don't account for real loads or the full output range). Perhaps this could get help resolve some of that. I've no direct experience with simulating anything but very simple circuits, but I do see it done for almost every solid state design here (in principle, could apply just as much to tube designs, though they seem to be less frequently tested by simulation).
I would think it would be fairly simple to come up with a few complex loads, that would approximate speakers and run the sims with them. It could not be comprehensive, but it should give a pretty good idea where the problems lie, and maybe why some designs perform better than others - or maybe even what kind of loads should be used with particular designs. With some experience, it should be possible to get a rough consensus on some model loads that could become part of a common protocol.
Anyone?
Sheldon
"Btw. I wonder why all those simulations never attempt to drive a real - that is, a complex -load? They would be amazed what nasty stuff happens to those thd figures...."
From here: http://www.diyaudio.com/forums/solid-state/161293-recovery-clipping-soft-clipping.html#post2085488
So why isn't this done routinely? There is so much discussion and argument about the meaning of distortion and it's relationship to perceived audio quality (including the point that the normal THD figures don't account for real loads or the full output range). Perhaps this could get help resolve some of that. I've no direct experience with simulating anything but very simple circuits, but I do see it done for almost every solid state design here (in principle, could apply just as much to tube designs, though they seem to be less frequently tested by simulation).
I would think it would be fairly simple to come up with a few complex loads, that would approximate speakers and run the sims with them. It could not be comprehensive, but it should give a pretty good idea where the problems lie, and maybe why some designs perform better than others - or maybe even what kind of loads should be used with particular designs. With some experience, it should be possible to get a rough consensus on some model loads that could become part of a common protocol.
Anyone?
Sheldon
Last edited:
Asymmetric distortions
Hi Sheldon
To gain reactive impedance and reactive power is essential in testing an amplifier for the eventual difficult loads it will face. Throw masses of asymmetric conditions that amplifiers are usually intolerant to. Good use and knowledge of what can be done with the Argand phasor/vector diagram and use of complex notation to develop loads that, at resonance, will sometimes force a phase shift of between 20 to 30 degrees between current and volts - is a strong part of the development of the PSU and output stage. For anyone used to doing a delta star transform in LCR circuits to simulate a far from static load, causing asymmetric clipping... this is routine.
Hi Sheldon
To gain reactive impedance and reactive power is essential in testing an amplifier for the eventual difficult loads it will face. Throw masses of asymmetric conditions that amplifiers are usually intolerant to. Good use and knowledge of what can be done with the Argand phasor/vector diagram and use of complex notation to develop loads that, at resonance, will sometimes force a phase shift of between 20 to 30 degrees between current and volts - is a strong part of the development of the PSU and output stage. For anyone used to doing a delta star transform in LCR circuits to simulate a far from static load, causing asymmetric clipping... this is routine.
Attachments
Last edited:
I've posted that remark on complex loads...it has a couple of consequences.
Firstly one cannot do without a SOAR protection.
Next consequence is to rethink the compensation.
Even if the amp is intended to drive one speaker directly without passive crossover it is only then simple if the speaker is in a perfectly dampended enclosure and radiates into an anechoic space...
Vented enclosures are a bit more complicated and horns yield a sort of spiralling
impedance curve in the complex plane.
I have noticed that particular topologies of power stages employing voltage feedback react most strangely to real world complex loads that is possibly why those topologies never made it in the real world.
Not everything is as easy as it appears at first glance.
Firstly one cannot do without a SOAR protection.
Next consequence is to rethink the compensation.
Even if the amp is intended to drive one speaker directly without passive crossover it is only then simple if the speaker is in a perfectly dampended enclosure and radiates into an anechoic space...
Vented enclosures are a bit more complicated and horns yield a sort of spiralling
impedance curve in the complex plane.
I have noticed that particular topologies of power stages employing voltage feedback react most strangely to real world complex loads that is possibly why those topologies never made it in the real world.
Not everything is as easy as it appears at first glance.
I routinely simulate with 8R//2uF, and a very tricky one is 8R//100nF. This naturally has implications for phase shift, lag compensation, and THD.
Most SOAR protection affects the sound. You need to be very careful in the design of protection.
BUT, you do find that horrific sims do not always give unstable amps. The technology is very useful, and saves money, but nothing beats a real build and measure.
Hugh
Most SOAR protection affects the sound. You need to be very careful in the design of protection.
BUT, you do find that horrific sims do not always give unstable amps. The technology is very useful, and saves money, but nothing beats a real build and measure.
Hugh
Thanks Guys,
I'm not surprised that amps are simmed for various reactive loads, as part of a stability check. But I was more driving at the distortion contribution that hafran mentioned, and its profile. I know Hugh, that you strive for a particular profile, or at least to avoid a particular profile. Just thinking that this should be a more or less standard test with sims. Seems like the tedious part is building the circuit model under test. Might as well do some more extensive distortion testing with realistic loads, as long as one has gone to all the trouble to model the amp?
This may be something commonly done, but I haven't noticed it here. Other than the usual test at 1kHz, the only addition I've periodically seen, is distortion at other frequencies - generally with a resistive load.
Sheldon
I'm not surprised that amps are simmed for various reactive loads, as part of a stability check. But I was more driving at the distortion contribution that hafran mentioned, and its profile. I know Hugh, that you strive for a particular profile, or at least to avoid a particular profile. Just thinking that this should be a more or less standard test with sims. Seems like the tedious part is building the circuit model under test. Might as well do some more extensive distortion testing with realistic loads, as long as one has gone to all the trouble to model the amp?
This may be something commonly done, but I haven't noticed it here. Other than the usual test at 1kHz, the only addition I've periodically seen, is distortion at other frequencies - generally with a resistive load.
Sheldon
Sheldon,
In my observation, there is blistering debate on the worth or otherwise of distortion measurements, and most audio companies are very loathe to polarise their buyers by jumping into either camp. Witness the wars waged here, and the many blood-speckled monitors around the world.......
For some reason, not much research has been done, at least in the public arena, on the correlation between distortion and the subjective listening experience. That is not to say people don't have strong opinions, of course!! Distortion measurement is objective, performed by engineers, whilst the later is subjective, experienced by golden eared individuals presumably influenced by psychotropic drugs. This is not a sexy area; damned if you do, and damned if you don't. The fact is that human hearing is very complex, and a great deal of psychoacoustic background should first be covered to suitably quantify the 'subjective listening experience' - do you get my drift?
When you realise that most gear is, or at least has been, sold on THD specs, you apprehend that disturbing the status quo may have commercial risks. Thus no audio company is prepared to back such research, and it's left to impecunious universities.
Jean Hiraga published his view thirty years ago that a monotonically decreasing profile was the best, with H2 at -60dB or less and subsequent harmonics 5-10db uniformly lower thereafter. Most in the field agree, but there is still serious debate. One thing is certain; THD is not very effective in predicting the subjective listening experience.
OTOH, and there is always a rider, human senses are so unreliable that many engineers say that only instrumentation can be trusted to measure an amp. They have a good point, but the correlation is still poor.
In the end, and given the subjectivity of the design exercise, it comes down to part engineering, part art. But then, there are many technologies like this - the automobile, for one, so it's not hallowed ground. And knowing your interest in the modern automobile, I'm sure this resonates with you......
Cheers,
Hugh
In my observation, there is blistering debate on the worth or otherwise of distortion measurements, and most audio companies are very loathe to polarise their buyers by jumping into either camp. Witness the wars waged here, and the many blood-speckled monitors around the world.......
For some reason, not much research has been done, at least in the public arena, on the correlation between distortion and the subjective listening experience. That is not to say people don't have strong opinions, of course!! Distortion measurement is objective, performed by engineers, whilst the later is subjective, experienced by golden eared individuals presumably influenced by psychotropic drugs. This is not a sexy area; damned if you do, and damned if you don't. The fact is that human hearing is very complex, and a great deal of psychoacoustic background should first be covered to suitably quantify the 'subjective listening experience' - do you get my drift?
When you realise that most gear is, or at least has been, sold on THD specs, you apprehend that disturbing the status quo may have commercial risks. Thus no audio company is prepared to back such research, and it's left to impecunious universities.
Jean Hiraga published his view thirty years ago that a monotonically decreasing profile was the best, with H2 at -60dB or less and subsequent harmonics 5-10db uniformly lower thereafter. Most in the field agree, but there is still serious debate. One thing is certain; THD is not very effective in predicting the subjective listening experience.
OTOH, and there is always a rider, human senses are so unreliable that many engineers say that only instrumentation can be trusted to measure an amp. They have a good point, but the correlation is still poor.
In the end, and given the subjectivity of the design exercise, it comes down to part engineering, part art. But then, there are many technologies like this - the automobile, for one, so it's not hallowed ground. And knowing your interest in the modern automobile, I'm sure this resonates with you......
Cheers,
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.