I think our ways of looking at this differ 😉
My view is that the amp includes the PSU and and you cannot separate them, both affect sound quality in equal measure. If you really want a killer PSU one option is a stack of SealedLeadAcid (SLA) batteries, running the amp off line. Given the price if Hi-Fi it makes 400V of 7AH batteries seem almost reasonable 😉
All things are possible of course, but my view is that GNFB is outdated and not required at all. Smaller stages of local and 2 stage feedback are all that's needed.
I think you could improve the amp more by cutting the GNFB loop in two of more bits, putting a loop around the whole amp is giving the the worst delay possible between the error and the possible correction.
This is a good article to read: HERE which is really the key point of the reason I steer away from GNFB.
Yes I'm sure there is an awful lot of that going on! Since taking the OPT out of the feedback loop in my tube amp it's been sounding rather more natural. Now I have a feedback loop instead between the output tune anode (it's an SE) and the driver cathode - two stages - to lower the impedance driving the primary of the transformer, getting the signal across better to the secondary in the first place 🙂
In fact Schade feedback would probably be ideal if note for the triodes I'm using.
Thanks for pointing me to that paper. I had a quick look at it, and I disagree. For example, he claims that power sent to the speaker is largely reflected back to the amp and not dissipated as heat. In a strict transmission line sense this is true, but it happens within a few ns and the power is reflected straight back from the amp too as this is an even worse match to the speaker cables than the speaker is. The net result is that almost all the power from the amp ends up in the speaker either as sound or heat, which he denies. This spurious 'time domain' or 'transmission line' argument is often put forward by people who know a little about RF but not enough to really understand the situation.
I agree with you that sometimes there are advantages in using NFB over fewer stages than the entire amp. Deciding when to do this or not is one element of good design. Generally you get a bit more distortion, but a bit more stability, so there is a tradeoff. The best thing is to design each stage to be the best it can be, within reason, then improve the whole thing, where necessary, with some NFB. Some people can't even design a single stage properly!
I agree with you that sometimes there are advantages in using NFB over fewer stages than the entire amp. Deciding when to do this or not is one element of good design. Generally you get a bit more distortion, but a bit more stability, so there is a tradeoff. The best thing is to design each stage to be the best it can be, within reason, then improve the whole thing, where necessary, with some NFB. Some people can't even design a single stage properly!
The speed of a feedback amp directly concerns the type of compensation used, as well as speed of the devices. The interesting thing I find with most SS GNFB circuits is the instability into capacitive loads; this comes automatically with the type of compensation used. The stability compensation only responds to changes in output voltage, hence if the output voltage doesn't change, as is the case with a capacitive load, the amp will be only as stable as it was without compensation at all. I wonder if better compensation would improve the sound... I've tried altering schemes, and in simulation OL bandwidth is greatly improved.
- keantoken
- keantoken
Capacitive loads can be a problem even with just local feedback. A cathode/emitter follower with a capacitive load can oscillate all on its own, which is why wise people always add a series resistor to the output.
That's my design philosophy, for both solid state and hollow state. Optimize the open loop performance, then add just enough NFB to improve it. There are some problems that you can't fix without NFB. Pentodes have an intrinsically high r(p) that will lead to woofer under damping and sloppy bass. Some finals (e.g. the 807) like to make high order harmonics that sound very dissonant and unpleasant. When I did an 807-based design, local NFB certainly helped to clean that up. Other finals (e.g. the 6V6-oids, 6BQ6) produce mainly h3 and so just need a bit of gNFB to take the edge off. When doing a new design, I like to spend at least a week with it and listening to the open loop performance before finalizing the design.
Also, the improved independence from device characteristics that NFB gives is an additional advantage since you're no longer so dependent on VT types, makes, models, and aging problems.
"Some people can't even design a single stage properly!"
A lot of them are also professional EE's. EE school doesn't teach anything regarding sonic performance. (I've seen texts like Practical Analysis of Amplifier Circuits Through Experimentation by Lorne MacDonald. There is not word one about sonic performance, or the sonic advantages/penalties of one topology v. another. If there is anything to wonder about it's that SS doesn't sound even worse than it already does.) There really is no need for solid state equipment to sound as horrible as a lot of it does. It's not easy designing for good sonics, you have to work at it a bit more, but it can be done since I've done it myself. So far, SS doesn't come up to HS sonic performance, but I've gotten pretty close with SS designs that'll beat anything you could pick up at places like Circuit Village or Mostest Goodest Buy. NFB gets the bad reputation since these folks abuse it to sweep their open loop design mistakes under the carpet, along with the vitality of the music.
It's gotten so bad that sound engineers master CDs to work around the sonic defects of the SS equipment it will be played through. A HS amp will expose every bit of the resulting nastiness. YUCK!
Many EE courses (in the UK, I can't speak for elsewhere) teach only the most elementary circuit design. I have watched students in a lab struggle to build simple op-amp circuits on a breadboard even given the circuit diagram (which in my opinion they should be able to design for themselves). Electronics textbooks still say foolish things about Class B etc. The result is that students (and sometimes their teachers) know less about audio (and RF) than a reasonably well-informed DIYer. It is not surprising that many commercial and published circuits are so bad. There is, however, an opposite problem - people who know a lot on a practical level often have only a vague idea about the real theory so are easily taken in by nonsense from plausible gurus.
Sure! Especially, when they think that an optimal system can be composed of optimal subsystems. 😉
Member
Joined 2009
Paid Member
I don't take this approach, but then again, I've never claimed to be competent 😱
not really, it is possible to generalize the definition of a scalar potential (voltage) when there is a non conservative electric field / changing magnetic field.
you can't separate the voltage from the electric field like this. The electric field is a vector quantity from which your voltage (scalar) is defined. Another way to think about this is that there is no voltage at a point, it is always relative to some other point. Mostly we choose to define the chasis as 'zero', a point of reference to which all other points can have a voltage. So the anode has a voltage relative to some other place. Between them exists an electric field - voltage/field/voltage/field, same beast. Current, on the other hand, is different, it's the movement of charge over time.
An interesting link about negative feedback by Martin Colloms Here referenced by this article.
Both support my opinion that GNFB smears and moves the distortion, not removes it. It does this in a dynamic transient, temporal way, not in a way that is measurable with steady tones.
I.e the less GNFB we need (down to zero if we stick to local feedback) the more musical the amp sounds. I think the very subtle timing information is damaged with GNFB - the timing in the midrange that for instance changes the waveform of a violin or snare drum - subtly shifting cues our ears use - causing the realism to go. In Hi-Fi some people think both that the ear can hear everything but notice nothing, I think we've gone past proving that false now.
In fact I'd go so far as to say I'm not sure I've seen a single schematic that needs a single big loop like a big floppy servo, all can be chopped down into smaller sections.
Both support my opinion that GNFB smears and moves the distortion, not removes it. It does this in a dynamic transient, temporal way, not in a way that is measurable with steady tones.
I.e the less GNFB we need (down to zero if we stick to local feedback) the more musical the amp sounds. I think the very subtle timing information is damaged with GNFB - the timing in the midrange that for instance changes the waveform of a violin or snare drum - subtly shifting cues our ears use - causing the realism to go. In Hi-Fi some people think both that the ear can hear everything but notice nothing, I think we've gone past proving that false now.
In fact I'd go so far as to say I'm not sure I've seen a single schematic that needs a single big loop like a big floppy servo, all can be chopped down into smaller sections.
Just wondering which is the most non-linear part of a tube amp circuit, the tubes or the OPT? Just wondering as I've been experimenting with many different circuits in a 6P3S (6L6) PP amp lately, and could not get it to sound very nice no matter what local or GNFB I used, or whichever circuit, until I finally replaced the OPT's. Just answered my own question I guess? Interesting maybe to note that the Baby Huey type circuit with its type of local feedback did the best of all the circuits before I fitted the new OPT's.
A poor OPT is by far worse than any properly functioning tube I've heard, but circuit differences are perhaps even more profound as long as we're talking full range response.
So you're saying that even a subpar OPT sounded better with a circuit that didn't try to fix it up with a global feedback loop. It may be a case of measures better (with nfb around the OPT) but sounds worse.
Also of note is that local plate-grid feedback as used in the BH will result in a nice low effective plate resistance, which is IMO about the best you can do for a cheap OPT.
The music signal put through a low or zero-feedback amp will almost certainly have come through many high-feedback circuits in its journey from musician to music-lover. Replacing the last such circuit with zero-feedback cannot remove the supposed errors introduced by all the earlier ones in the recording chain. However, what it can do is add some distortion and it is well known, but frequently denied, that a little low-order distortion makes things sound a little better than no distortion. All the people who buy "tube buffers" are evidence for this - these add a little noise, a little (or sometimes a lot) low-order distortion, and restrict bandwidth a little - all are small distortions yet many people claim that for them the sound is improved. You only then have to add the powerful placebo effect and you have the observed results.
Of course, when someone cannot accept the true explanation for a phenomenon they have to accept a false explanation and so the stuff about GNFB, temporal errors, transients etc are trotted out. Much of this denies Fourier theory - some of the statements people make cannot be true for a bandlimited signal.
There are mysteries about amplifier sound, but there is a lot of nonsense too which is widely and sincerely believed. When someone says that a high feedback amp is flat or boring, he may simply be saying that he finds accurate sound flat and boring. Add a tube buffer and 'improve' it, thereby proving that it is the very accuracy of the amp which is the problem!
Of course, when someone cannot accept the true explanation for a phenomenon they have to accept a false explanation and so the stuff about GNFB, temporal errors, transients etc are trotted out. Much of this denies Fourier theory - some of the statements people make cannot be true for a bandlimited signal.
There are mysteries about amplifier sound, but there is a lot of nonsense too which is widely and sincerely believed. When someone says that a high feedback amp is flat or boring, he may simply be saying that he finds accurate sound flat and boring. Add a tube buffer and 'improve' it, thereby proving that it is the very accuracy of the amp which is the problem!
Yes, but the most distorting things in this chain are those that have higher power gain, especially when they deal with mechanical things that have non-linear and frequency dependent impedances. There are 2 such things: mic preamps and power amps. Juts draw a power amplification diagram from microphone to mic preamp, EQ, fader buffer, summing amp, console output buffer, power amp, and see for yourself.
However, too many buffers in consoles using some opamps similar to 741 may add quite audible distortions as well. And some "Harmonizer" can be used to mask such distortions.
Fourier theory has nothing to do with sounds. It is a pure mathematics, sharp and cold, and in some cases it is too heavy to describe some things.
This explanation is common among people who don't understand why chains with distortions they can not measure sound flat and boring. Try to find that flat and boring distortions, and you will understand how to minimize them.
Last edited:
Member
Joined 2009
Paid Member
If there is one thing people on this forum have taught me, is the best way to understand what is nonsense and what is not, is to build and listen and make your own judgments.
You can't generalize about 'high feedback' amps. With high levels of feedback the accuracy of the error amplifier becomes critical and not all designs will be good. As usual, the quality of the implementation is important. Same goes for tube buffers - there are people who find some well engineered tube amps to sound very much like well engineered SS amps.
You may also discover that the speakers are the things that make a big difference to listening experience. Not only the speaker it'self, but the interaction of speaker-amp and speaker-room.
Enjoy the journey 🙂
Wavebourne says it all really, but I'll just add:
I'm not sure of the phenomenon or explanation in question either: please elaborate!
My observations about simple measurements, feedback, TIM etc have been correlated by my ears by listening to various Hi-Fi since the 1970s.
The delay of a signal through 5 stages of amplifier is irrefutable, and nothing to do with Fourier, it's more to do with frequency dependent phase shift of error signals and distortions. Without feedback, it's just coherent delay.
Without a time machine you simply cannot correct these errors with feedback, period!
Many engineers love feedback, and feedback is indeed the perfect device to correct a linear system with errors. Amplifiers and speakers are not linear however so feedback can't and doesn't correct all the errors.
Also if I like the sound of a real violin, but not the sound of a violin played by a high global feedback amp, surely that makes the amplifier inaccurate, even if it measures well?
After all, a real instrument is accurate - or at least real - isn't it? 🙂
The decision of a true explanation is not your sole authority 😉.
I'm not sure of the phenomenon or explanation in question either: please elaborate!
My observations about simple measurements, feedback, TIM etc have been correlated by my ears by listening to various Hi-Fi since the 1970s.
The delay of a signal through 5 stages of amplifier is irrefutable, and nothing to do with Fourier, it's more to do with frequency dependent phase shift of error signals and distortions. Without feedback, it's just coherent delay.
Without a time machine you simply cannot correct these errors with feedback, period!
Many engineers love feedback, and feedback is indeed the perfect device to correct a linear system with errors. Amplifiers and speakers are not linear however so feedback can't and doesn't correct all the errors.
Also if I like the sound of a real violin, but not the sound of a violin played by a high global feedback amp, surely that makes the amplifier inaccurate, even if it measures well?
After all, a real instrument is accurate - or at least real - isn't it? 🙂
My concern is with those that state that high feedback has to be wrong, simply because it is high feedback. Of course there may be problems with storage effects, but this is an example of where feedback is not an appropriate solution which I am happy to accept. I am not arguing for high feedback, as it is often improperly applied. I accept that there may be things we can hear which we have not yet learnt to measure, but the solution lies in better engineering rather than snake oil. My problem is that the explanations which people put forward are often nonsense. As in many other situations in life, I am unconvinced by both sides: the high feedback and the zero feedback people. I don't accept what the "golden ears" say, as almost all of this can be explained by the placebo effect and low-order distortion. I am equally unconvinced by the "almost zero THD" people, as they are believing their instruments before they have demonstrated that their instruments are measuring the right thing.
I can't argue with someone who prefers a real violin to an electronic reproduction, but I can still suggest that his preferred explanation for this may be incomplete. I will happily ignore someone who deliberately distorts (in an easily measurable way) his sound and then says that he prefers it that way and trots out some nonsense to support his preference - yet this is what many "audiophiles" seem to do, and some of them may never have heard a real unamplified violin!
I can't argue with someone who prefers a real violin to an electronic reproduction, but I can still suggest that his preferred explanation for this may be incomplete. I will happily ignore someone who deliberately distorts (in an easily measurable way) his sound and then says that he prefers it that way and trots out some nonsense to support his preference - yet this is what many "audiophiles" seem to do, and some of them may never have heard a real unamplified violin!
What part of switching from global negative feedback to local negative feedback and/or less need for corrective feedback is snake oil?
This surely is better engineering!?
Talking of snake oil though, I feel for those who payed $1000s to buy a system than needs totally transforming, and at a cost twice that of my Sweet Peach amp 😉
Attachments
- Status
- Not open for further replies.