Graham Maynard said:
OK. Practically any amplifier with basic global negative feedback will require a load-isolating output inductor if a reduction in phase margin is to be avoided when driving a capacitive load.
If we wish to do away with the inductor, then we need to ensure that the phase margin when driving a resistive load is in excess of optimal - and that almost necessarily means a lowering of negative feedback factor at HF. This will increase THD and depending on toploogy, slew rate as well (for instance, we could just make Cdom bigger).
The Leech feedback scheme mentioned earlier, which is used in his famous DIY amplifier, is one scheme that removes the need for an output inductor, but in doing so actually worsens the output impedance and increases the THD.
There are some designs out there with very little or no negative feedback that have absoutely no problem driving even the most capacitive of loads. however, by virtue of their lack of feedback, these amplifiers are plagued by high output impedance and low damping factors - much more badly than a well designed amplifier with high global negative feedback an a small output inductor.
Cheers,
Glen
john curl said:
I know that the output impedance of your JC-1 is lower. I don't see why that should invalidate anything I have said. My last paragraph isn't appliciable to your JC-1 anyway, as by your own admission, is uses generous global negative feedback.
Your JC-1's THD is also hundreds of times higher than the Halcro, and its bandwidth into loads above 2 ohms isn't as good.
How about that?
Almost all the distortion in the JC-1 is second harmonic. I could remove it with a cascoded vas stage, but I chose not to. Big deal! I am NOT trying to compete with Halcro in low distortion. I think that is a waste of time. I am only concerned with the ORDER of the harmonic distortion. Also, I have 8 times more peak output current, lower output impedance, higher input impedance, and a very high slew rate, and at 1/5 the price. That's enough for this model.
Hi Rodolpho,
The delay variation in time you quote certainly looks like that sort of range, I did measure and list them at the time. Yet look at these values compared with the individual choke group delay variation for the same frequency with the perfect (virtual) amplifier in post#211.
And these are all steady sines, not even music with all of its variations and asymmetricalities. The loudspeaker system induced variation of output voltage in time due to amplifier characteristics will also change as input waveform varies wrt to prior loudspeaker system energisation, as well as with output (power) current demands.
Re your 'PS', what would you like me to add ?
Hi Glen,
The point I have been trying to get through all along is that THD is not the most important aspect. If you are writing about THD at 20kHz then 0.01% is way beyond what anyone could hear anyway, so I suggest that you might indeed try 'lightening up' as you suggest; the THD might go up and yet the amp actually sound better !
Long ago I found it easy to manipulate internal feedback responses in order to minimise manually adjusted fundamental nulled distortion at say 10/20kHz. But I saw that this was due to me reactively tuning some internal part of the circuitry. The reproduction - though seemingly more pure - sounded different - wrong - as if it had been (borrowing Carlos' word) sterilised.
And every time I made an internal feedback adjustment I was annoyed at having to readjust the null phase, even though nothing else had been changed !!!
The work was taking ages and for no improvement in reproduction, ie. the notable improvement in steady sine measured THD figures into a dummy load was meaningless for music reproduction via loudspeakers !!!
Since then I've seen all sorts of similarly tuned internal stages, feedforward etc., all claiming to reduce distortion ..... so for decades I have been saying to myself 'oh yeah?' knowing that they have not prioritised upon the audible waveform distortion which is the most important to our senses ?
Hi John,
EW did not do the audio community any favours in relation to your (never seen) written-up investigations !
The same happened with some (never seen) of Bengt Ollson's work in the late '90s.
Your post is too interesting for me to cover at this moment. Back shortly.
Cheers ......... Graham.
The delay variation in time you quote certainly looks like that sort of range, I did measure and list them at the time. Yet look at these values compared with the individual choke group delay variation for the same frequency with the perfect (virtual) amplifier in post#211.
And these are all steady sines, not even music with all of its variations and asymmetricalities. The loudspeaker system induced variation of output voltage in time due to amplifier characteristics will also change as input waveform varies wrt to prior loudspeaker system energisation, as well as with output (power) current demands.
Re your 'PS', what would you like me to add ?
Hi Glen,
The point I have been trying to get through all along is that THD is not the most important aspect. If you are writing about THD at 20kHz then 0.01% is way beyond what anyone could hear anyway, so I suggest that you might indeed try 'lightening up' as you suggest; the THD might go up and yet the amp actually sound better !
Long ago I found it easy to manipulate internal feedback responses in order to minimise manually adjusted fundamental nulled distortion at say 10/20kHz. But I saw that this was due to me reactively tuning some internal part of the circuitry. The reproduction - though seemingly more pure - sounded different - wrong - as if it had been (borrowing Carlos' word) sterilised.
And every time I made an internal feedback adjustment I was annoyed at having to readjust the null phase, even though nothing else had been changed !!!
The work was taking ages and for no improvement in reproduction, ie. the notable improvement in steady sine measured THD figures into a dummy load was meaningless for music reproduction via loudspeakers !!!
Since then I've seen all sorts of similarly tuned internal stages, feedforward etc., all claiming to reduce distortion ..... so for decades I have been saying to myself 'oh yeah?' knowing that they have not prioritised upon the audible waveform distortion which is the most important to our senses ?
Hi John,
EW did not do the audio community any favours in relation to your (never seen) written-up investigations !
The same happened with some (never seen) of Bengt Ollson's work in the late '90s.
Your post is too interesting for me to cover at this moment. Back shortly.
Cheers ......... Graham.
Graham,
I've seen your post somewhere about this issue
From what I see : you are trying to say that the amplifier output distortion spectrum, at high frequency when driving inductive load is not as little as predicted by simple THD measurement, especially when in class B / AB transition region ?
and become even worse with output inductor connected ?
or maybe you are trying to point another issue ?
I've seen your post somewhere about this issue
From what I see : you are trying to say that the amplifier output distortion spectrum, at high frequency when driving inductive load is not as little as predicted by simple THD measurement, especially when in class B / AB transition region ?
and become even worse with output inductor connected ?
or maybe you are trying to point another issue ?
stoolpigeon said:
Unfair comparison, you should go listen to a set of JC-1s before making a cheap shot.
The JC-1 has enough current gain to blast the 90 amps out, and the powersupply of the JC-1 is nothing like that of a featherweight Carver amp.
(Mr Carver also manufactured a few amplifiers that were rather good, but expensive too, suppose you have not audited them either)
Hey Mr 911 Curlera, high time for a JC-2 thread on this wacko club !
Yes, I sort of did sound like Bob Carver on that response. Do you know Bob?
What we say here should not be that sort of comparison contest, but I made my design tradeoffs and Halcro made theirs. It IS ironic that a medium feedback amp would have a lower output impedance with frequency than a high feedback amp, but that is because I apparently have a very low open loop impedance, lower than most designs, and I don't use an output coil. My very high damping factor is not really that important, and neither is the very low distortion that Halcro does so well.
What we say here should not be that sort of comparison contest, but I made my design tradeoffs and Halcro made theirs. It IS ironic that a medium feedback amp would have a lower output impedance with frequency than a high feedback amp, but that is because I apparently have a very low open loop impedance, lower than most designs, and I don't use an output coil. My very high damping factor is not really that important, and neither is the very low distortion that Halcro does so well.
Graham Maynard said:
It is not clear to me to which transfer corresponds the group delay you plotted.
By the way, any multiple driver speaker system unless driven by separate amplifiers for each and also crossed over electronically by some digital linear phase gizmo, will have significant group delay excursions well within the audio band, probably the worst case being 2-way boxes where the crossover frequency falls smack in the most sensitive region for perception.
This is irrespective of the amplifier flatness, output coil or whatever, meaning as I pointed out earlier that to judge relevance of this or that technique, one must not lose sight of the whole picture.
Rodolfo
john curl said:
Your design choices seem to work, especially when one looks at the performance/price ratio.
As an aside, I've been a fan of differential inputs since working as a live sound guy decades ago: unbalanced signals often got corrupted, whereas balanced lines often sounded a lot cleaner even if there wasn't an obvious interference problem.
Hi Hartono,
I have not actually mentioned distortion spectrums. To date I have been trying to point out that there is a need for further study/investigation into a fundamental reason for individually unique amplifier-loudspeaker interface induced reproduction distortions, which steady sine based investigations cannot reveal because these are viewed in time isolation.
I am saying that amplitude distortions wrt to input waveform are not as little as THD predicts, but I am wary of using the word spectrum because this too tends to be a time isolated (wrt a valid music source) sine based concept.
Hi John Ellis,
I find it hard to reply because I just sit nodding my head at your post.
Increasing a C.dom has the potential to slow VAS collector slew such that the output devices can be load driven through a fraction of their bias potential (ie. the load can independently generate an output terminal voltage) before the global loop can control correction, and this as you say with possible input stage overload and new momentary error voltage developed across the emitter resistors. This as with the red and green traces shown in my post#215 link.
I too would prefer direct coupling circuits, but we can never be sure of sources unless transformer coupled, so I still use coupling capacitors, but of a high enough value to minimise LF level shifting, and not just wrt say a -1dB frequency response specification.
I completely agree with you in relation to needing accurate virtual loudspeaker loads, also
"The question ought to be at what point we are not able to hear the differences...",
Hi Rodolfo,
Of course LS have their own problems, but the cheapest and even fullrange loudspeakers still allow us to hear the 'inner activities' of an amplifier enough to prove they are not wires with gain. So we cannot fail to study amplifiers just because loudspeakers and sources are not ideal !
Cheers ........ Graham.
I have not actually mentioned distortion spectrums. To date I have been trying to point out that there is a need for further study/investigation into a fundamental reason for individually unique amplifier-loudspeaker interface induced reproduction distortions, which steady sine based investigations cannot reveal because these are viewed in time isolation.
I am saying that amplitude distortions wrt to input waveform are not as little as THD predicts, but I am wary of using the word spectrum because this too tends to be a time isolated (wrt a valid music source) sine based concept.
Hi John Ellis,
I find it hard to reply because I just sit nodding my head at your post.
Increasing a C.dom has the potential to slow VAS collector slew such that the output devices can be load driven through a fraction of their bias potential (ie. the load can independently generate an output terminal voltage) before the global loop can control correction, and this as you say with possible input stage overload and new momentary error voltage developed across the emitter resistors. This as with the red and green traces shown in my post#215 link.
I too would prefer direct coupling circuits, but we can never be sure of sources unless transformer coupled, so I still use coupling capacitors, but of a high enough value to minimise LF level shifting, and not just wrt say a -1dB frequency response specification.
I completely agree with you in relation to needing accurate virtual loudspeaker loads, also
"The question ought to be at what point we are not able to hear the differences...",
Hi Rodolfo,
Of course LS have their own problems, but the cheapest and even fullrange loudspeakers still allow us to hear the 'inner activities' of an amplifier enough to prove they are not wires with gain. So we cannot fail to study amplifiers just because loudspeakers and sources are not ideal !
Cheers ........ Graham.
Attachments
Hi Forr,
It took me three decades to construct an amplifier better than my own 4xKT88 UL-AB1.
Tube Dude is quite correct.
Think of an amplifier driving a first audio cycle into a LS.
The amplifier voltage always starts out leading current, but by the end of that cycle the current can be leading, thus the amplifier's NFB response can become modified and there can be waveform distortion within that first cycle.
LS do not generate leading current until after they have been energised, and initially they conduct normally at the resonant frequency until resonance has been energised.
Having internal inductance means that amplifier's themselves can react to this changing load reactance within a first cycle.
Don't worry about what other people think; what about you ?
Cheers ....... Graham.
It took me three decades to construct an amplifier better than my own 4xKT88 UL-AB1.
Tube Dude is quite correct.
Think of an amplifier driving a first audio cycle into a LS.
The amplifier voltage always starts out leading current, but by the end of that cycle the current can be leading, thus the amplifier's NFB response can become modified and there can be waveform distortion within that first cycle.
LS do not generate leading current until after they have been energised, and initially they conduct normally at the resonant frequency until resonance has been energised.
Having internal inductance means that amplifier's themselves can react to this changing load reactance within a first cycle.
Don't worry about what other people think; what about you ?
Cheers ....... Graham.
Graham Maynard said:
Graham,
Let me fierst remind you did not clarify the point with regards to the schematics and group delay plot post. To ease things, as I see you put an ideal unity gain amplfier and split the output on one hand to a resistive load and on the other to a passive crossed over 2-way speaker load, in both cases placing 6 uHy//8.2 ohm in series . I tend to believe what you plotted is the differential group delay after the output coils (V), is this right?
To gain perspective, please check a study I posted here to have an idea what loading a typical box presents. With reference to the group delay issue, I attached below from the same model, plots of group delay vs. frequency both for the low pass and high pass ways, only what is plotted is actually voice coil velocity, since what is modeled is not only the crossover network but driver electromechanical equivalent including box loading, so this is a more accurate representation of what happens with sound pressure which is what we hear.
See the values and draw your own conclusions.
Rodolfo
Attachments
- Status
- Not open for further replies.