I do not; I use spice to estimate, then start with the listening tests using a real speaker.
If you use that approach you will over compensate and the amp will sound leaden and slow.
However, I'm just one man with one opinion, so go for it if you disagree!!
Hugh
If you use that approach you will over compensate and the amp will sound leaden and slow.
However, I'm just one man with one opinion, so go for it if you disagree!!
Hugh
What is the "side product" (looks a little like IMD graph) next to 2khz 2nd harmonic rising above noisefloor BB at 8 watt ?
I am sorry, Irribeo, I cannot answer that, we will need to wait X arising from his bed.......
HD
HD
Thank you very much Hugh.
You are the fastest gunman
By the way, this week I improved my KEF Q100 coaxial speakers after adding mass to the cases -> clearer sound, drier bass and more 3D. It is a pity that the real sensitivity is between 84 and 85 dB 🙁
You are the fastest gunman
By the way, this week I improved my KEF Q100 coaxial speakers after adding mass to the cases -> clearer sound, drier bass and more 3D. It is a pity that the real sensitivity is between 84 and 85 dB 🙁
I also wonder how much the longer wires/leads of the lower value mica used for testing add inductance that might be a bigger change than the capacitance.
That might be the +/- beat note (imd) between the 2nd harmonic and or 120Hz mains frequencies.
Hi Hugh,
if it were possible do you think it would be desirable to have the harmonic profile of 1w remain at higher powers?
if it were possible do you think it would be desirable to have the harmonic profile of 1w remain at higher powers?
ED,
Yes It would be desirable..... but as you load up any amp you inevitably incur undesirable distortions. I do not know to do this at this stage but I have only worked on this effectively for ten years!
It may be possible with pre-distortion sometimes called forward error correction.
HD
Yes It would be desirable..... but as you load up any amp you inevitably incur undesirable distortions. I do not know to do this at this stage but I have only worked on this effectively for ten years!
It may be possible with pre-distortion sometimes called forward error correction.
HD
First sound with 94dB sensitive XKi MTM speakers (dual Beta 8cx(no CD fitted) and aluminum dome tweeter. This speaker is a big in size (31in tall and 55lbs) and sound. First sound test driven by ALPHA BB. I can't give it more than maybe 1.5vrms and the SPL from one speaker is too loud to be comfortable. Will be a good match for ALPHA 20 4R version.
XKi is my own semi-6th order bandpass reflex with Karlson couplers. XO is passive time-aligned.
XKi is my own semi-6th order bandpass reflex with Karlson couplers. XO is passive time-aligned.
Thanks Hugh,
I asked that because I got a configuration that maintains that profile with more power and in AB class. The harmonic profile very much resembles that of a SET but with much more power, hence the reason of the question. I must say that such a result is still in simulation mode but by experience I know it is not far from what it will be in a real circuit.
Hi Hugh,
After 4R, gain and damping factor, I played a little with the compensation cap 🙂
From the simulations the Alpha seems to be very stable!
I tested with a square wave and adding a variable capacitive load to the output (1nF 5nF 20nF 100nF)
Even when I reduce the compensation cap to 10pf, all the square waves still look great, even with the 100nF load.
AC analysis also looks good, just a little more bandwidth, almost up to 1Mhz.
This is with the 4R version.
So far I don't see any oscillations with 10pf.
Is 10pf not high enough ?
Regards,
Danny
After 4R, gain and damping factor, I played a little with the compensation cap 🙂
From the simulations the Alpha seems to be very stable!
I tested with a square wave and adding a variable capacitive load to the output (1nF 5nF 20nF 100nF)
Even when I reduce the compensation cap to 10pf, all the square waves still look great, even with the 100nF load.
AC analysis also looks good, just a little more bandwidth, almost up to 1Mhz.
This is with the 4R version.
So far I don't see any oscillations with 10pf.
Is 10pf not high enough ?
Regards,
Danny
Attachments
Last edited:
Member
Joined 2009
Paid Member
I'm fully on board with Hugh on this one (he taught me about this back in 2010 or so). My experience has shown that the ear is the best judge for fine tuning the compensation cap. It's not just spatial imaging, it affects the sound in other subjective ways. When Cdom is too high the sound is subjectively compressed and dull, when it is too low it is subjectively expanded and tizzy but when you find the sweet spot you will find the amp takes on a 'magical' sound relative to the other positions. You can 'lock' in this value even listening in mono.
Square wave testing is not as simple as most realize - with very fast edges you really have to pay attention to cables, terminations and how the signal gets from the signal generator to your amp and then to the O'scope. In most cases it will get a little ugly all by itself. And once set up I find the square wave test really does not tell you all you want. I have found, like Hugh, that it leads to over-compensation, even gross-over compensation if you blindly plug away at it (it was particularly noticeable on my TGM2 amp many years back which had a CFP LTP, inspired by Hugh of course).
Danny - you may be safe at 10pF depending on parasitics. But, sometimes it is only when the amp is tortured that you find out if it's a large enough value to prevent problems. An example would be clipping - driving the amp to clipping can provoke nasties if the compensation is marginal. It might also show up some sensitivity to certain speaker cables etc. I would use your ears - if it sounds just as good at 30pF as it does at 10pF then use 30pF.
Last edited:
Almost. In my simulations 2H reach -40dB...3H reach -50...-60dB with higher harmonics is almost noise.
Member
Joined 2009
Paid Member
After all these years I have found almost no correlation between simulated FFT and sound. I feel that I should be able to, but most of my amps simulate very similar, in fact they look mostly just like above with dominant 2nd and 3rd (2 higher than 3rd). But they sound different. I expect a really 'bad' or 'strongly flavoured' FFT would be quite audible but with nfb SS amps this isn't the case. Most of them are low distortion, like this one, where the harmonics are theoretically below audible threshold. But they sound different nonetheless. Apart from the ASKA amplifier (the one and only, original) most SS amps with gnf seem to my ears to sound better the lower the overall distortion.
Last edited:
ED,
I think that is a tube headphone amp?
H2 at -40dB H3 -60dB H4 -60dB is typically a tube profile. With SS I try to aim for -65dB to -72dB, just on the level of audibility, the Woo product would be very tubey and perhaps to many it would be a bit too much tubey!
Gareth,
Thank you for your valued input and your kind comments. I have to admit I am a little diffident about the profile conferring wonderful sound quality these days; I suspect a lot of it is about feedback ratios.
Square wave testing is very hard on the Boucherot network used on most feedback amps, sometimes called the Zobel. That resistor gets very hot with any frequency of square wave! Testing by the ear - or by cat as JLH would describe it - needs confidence, and most people do not test their ears at all. They say it's subjective, important to ignore it. That's a pity, because music is subjective, and it's the reason we design for! So there are some absurd notions in this game.......
I suspect that the majority of bad sound comes back to the compensation. As you say, get it wrong and it really destroys the sound quality of any beautiful amplifier.
I see the AKSA come up again and again. I never really understand why, when my later amps are superior. Perhaps I should produce it again!! Prasi, are you reading? We need to get moving!
Cheers,
Hugh
I think that is a tube headphone amp?
H2 at -40dB H3 -60dB H4 -60dB is typically a tube profile. With SS I try to aim for -65dB to -72dB, just on the level of audibility, the Woo product would be very tubey and perhaps to many it would be a bit too much tubey!
Gareth,
Thank you for your valued input and your kind comments. I have to admit I am a little diffident about the profile conferring wonderful sound quality these days; I suspect a lot of it is about feedback ratios.
Square wave testing is very hard on the Boucherot network used on most feedback amps, sometimes called the Zobel. That resistor gets very hot with any frequency of square wave! Testing by the ear - or by cat as JLH would describe it - needs confidence, and most people do not test their ears at all. They say it's subjective, important to ignore it. That's a pity, because music is subjective, and it's the reason we design for! So there are some absurd notions in this game.......
I suspect that the majority of bad sound comes back to the compensation. As you say, get it wrong and it really destroys the sound quality of any beautiful amplifier.
I see the AKSA come up again and again. I never really understand why, when my later amps are superior. Perhaps I should produce it again!! Prasi, are you reading? We need to get moving!
Cheers,
Hugh
I don’t hear any tizziness sound yet. I think I will leave it at 36pF silver mica though. I have some 22pF NP0.
Danny,
I did not answer your comments about the stability of the ALPHA.
Yes, I'd noticed it too, but I think it's the result of the first and second stages of the amp, the AKSA/Lender. The VAS is driven by voltage at the base and current at the emitter, with a very tight connection with the LTP. This makes it very stable and extremely fast. The high pole of this amp is high, around twice that of a standard, single base-driven VAS, and therefore compensation is easier because the phase shift is less than most amps, around 4 degrees at 20KHz.
In fact, I'm not using lag compensation, the usual compensation between base and collector of the VAS. I'm taking phase lead, from output to the fb node. This is more benign for the audio signal.
Thank you for your continuing interest,
Hugh
I did not answer your comments about the stability of the ALPHA.
Yes, I'd noticed it too, but I think it's the result of the first and second stages of the amp, the AKSA/Lender. The VAS is driven by voltage at the base and current at the emitter, with a very tight connection with the LTP. This makes it very stable and extremely fast. The high pole of this amp is high, around twice that of a standard, single base-driven VAS, and therefore compensation is easier because the phase shift is less than most amps, around 4 degrees at 20KHz.
In fact, I'm not using lag compensation, the usual compensation between base and collector of the VAS. I'm taking phase lead, from output to the fb node. This is more benign for the audio signal.
Thank you for your continuing interest,
Hugh