High NFB is intended to remove the 'character'. That is what it is for.
Opamps are not usually "easy to distinguish from each other". They can be distinguished (with some difficulty) under certain circumstances; most of the claims of 'night and day' differences are just male bragging or the result of instability due to poor circuit design (including random 'opamp rolling'). Remember that opamps are not perfect. They can suffer from common-mode distortion (which feedback cannot remove) or thermal feedback (which can affect LF) or, in some cases, crossover distortion.
Opamps are not usually "easy to distinguish from each other". They can be distinguished (with some difficulty) under certain circumstances; most of the claims of 'night and day' differences are just male bragging or the result of instability due to poor circuit design (including random 'opamp rolling'). Remember that opamps are not perfect. They can suffer from common-mode distortion (which feedback cannot remove) or thermal feedback (which can affect LF) or, in some cases, crossover distortion.
I would agree with your subjective observations, scopes and persons hear differently (unfortunately). This makes good sounding approaches less straitforward.
Would like to share some thinking out loud about your interesting schematic. . Most interesting is output SS buffer, probably it has low output impedance, but “ emulated by NFB “. Still there is some subjective sound (bass and midbass) difference, if the same output impedance achieved without NFB (say 70 mOhms could be with proper emitter follower) and due to NFB. Also, this stage does involve power supplies in the output signal chain, similar to, for instance, Aleph (modulated) Current Source. Although such and similar tricks allow to improve THD, they make subjective sound less natural and less involving, usually more dull. IMHO, best capacitors and bypasses at PS and PCB solve the task only partially.
At the input tube stage, I dislike small idle currents around 1mA. I prefer 10mA or such. Maybe it is my DIY crazyness, but, from previous experience with big tubes at input, I like “brutal” solutions, at present I use EL84-like (6P15P-EV) tubes (pentode connected) at the input long-tail pair of my current push-pull project. Each tube passes through 28mA (23mA plate, 5mA 2-d grid, 5mA 150v zener to 2-d grid), 500V B+, 300V at plate, 150V at second grid. CCS at the “tail” is based on BSP129 mosfet. Even resistive loaded, every tube provides voltage amplification 36, times 2 gives 72. Such input differential stage also observes “no ps current variation under signal”. 80V RMS are easy from such stage. I would appreciate also “no ps current…” at the output current buffer. Excuse me for some offtopic.
Last edited:
Member
Joined 2009
Paid Member
The crucial difference is that the output impedance is not constant across the audio frequency range. NFB tends to flatten the impedance curve. Without NFB some amplifiers don't have a flat impedance curve. Because the speaker does not have a flat impedance curve the interaction can change the sound.
Vladimir, I think we must have some NFB in this case since it's not so heavily biased. Such a short loop doesn't make the sound "dull".
The Aleph uses a clever design that makes the amplifier think the load is only about the half of what is really is. That circuit isn't really an NFB, but has it's own short feedback loop. So there are two NFB loops that works in parallel. Personally I'm a little sceptical to that, and Nelson admits it's a compromise.
I once made an Aleph clone, but couldn't resist modifying it after a decade or so. I inserted a phase splitter and made it complementary.
The thing about perceived bass. Usually it's the treble and midrange that gets modulated by the bass. When that happens our attention is confused and we cannot focus on the bass notes. We can't relax and "sink" in to the bass since our attention gets drawn towards the modulated higher frequencies. This is my personal reflexion.
That's why it's pointless to use a Krell to drive a subwoofer.
I think high THD figures makes the sound a bit more "brittle", even if it usually still is quite vivid and entertaining ( if the NFB is low, that is ). But high THD is tiresome in the long run.
High NFB: Controlled, strained or dull.
Low NFB, low bias: Brittle, vivid.
Low NFB, high bias: Vivid, musical and with all the proper nuances.
The Aleph uses a clever design that makes the amplifier think the load is only about the half of what is really is. That circuit isn't really an NFB, but has it's own short feedback loop. So there are two NFB loops that works in parallel. Personally I'm a little sceptical to that, and Nelson admits it's a compromise.
I once made an Aleph clone, but couldn't resist modifying it after a decade or so. I inserted a phase splitter and made it complementary.
The thing about perceived bass. Usually it's the treble and midrange that gets modulated by the bass. When that happens our attention is confused and we cannot focus on the bass notes. We can't relax and "sink" in to the bass since our attention gets drawn towards the modulated higher frequencies. This is my personal reflexion.
That's why it's pointless to use a Krell to drive a subwoofer.
I think high THD figures makes the sound a bit more "brittle", even if it usually still is quite vivid and entertaining ( if the NFB is low, that is ). But high THD is tiresome in the long run.
High NFB: Controlled, strained or dull.
Low NFB, low bias: Brittle, vivid.
Low NFB, high bias: Vivid, musical and with all the proper nuances.
These two make sense. In the low bias case, HF THD probably isn't purty. In the high bias case, things are much more likely to be constant across the audio band, which is pretty much accepted as being the next best thing if you can't get distortion to negligible levels.
IMHO a full understanding of distortion can only be gained when looking at THD over frequency plus distortion spectrum, and ideally distortion spectrum over frequency as well. Compare 10 kHz and 1 kHz distortion with some of these circuits, and some things should become clear.
Oh, and I'll second Bigun's point re: output impedance. Speaker with passive XOs can commonly use damping factor in excess of 10.
Sounds like that involved the odd bad high NFB amp.
High NFB, high bias seems like the logical route to establish a "blameless" baseline.
Wouldn't you say that full understanding of dist involves understanding of human hearing? Of course if we invented an amp that had 0.00001% THD - that would probably be truly blameless.
Isn't it likely that we humans are disturbed more from some types of dist than others? If we think this seems probable, then the THD chasers who takes all THD as a bunch, aren't they over-simplifying?
My personal belief, that I have mentioned earlier, is that our brain builds up a sonic image from a very advanced real time analysis of base tones, over tones, echoes, and so forth. If we have a complex amplifier that has low THD but where the THD is distributed over a very complex pattern of overtones - isn't it likely that our brain is more "confused" when it encounter complex distortion, than if it is subjected to simple dist, such as from a single ended triode that has almost exclusively 2'nd order harmonics?
Isn't it likely that we humans are disturbed more from some types of dist than others? If we think this seems probable, then the THD chasers who takes all THD as a bunch, aren't they over-simplifying?
My personal belief, that I have mentioned earlier, is that our brain builds up a sonic image from a very advanced real time analysis of base tones, over tones, echoes, and so forth. If we have a complex amplifier that has low THD but where the THD is distributed over a very complex pattern of overtones - isn't it likely that our brain is more "confused" when it encounter complex distortion, than if it is subjected to simple dist, such as from a single ended triode that has almost exclusively 2'nd order harmonics?
Of course. In fact, that's where the distortion spectrum thing comes in.
And unnecessarily complex, because this value would correspond to -140 dB. Last time I checked, the difference between typical peak listening levels and hearing thresholds was a lot less than that.
In a way, of course, and actually that's been known for decades. The first research in this direction dates from the 1930s. Lower-order harmonics tend to be heavily masked.
That being said, subpar ultrasonic distortion tends to make an amplifier sensitive to ultrasonic crap as possibly emitted by DACs or mistracking phono cartridges, due to corresponding intermod (2nd order is actually particularly suited). 2nd order HD of -70...80 dB at 20k or better seems to provide adequate rejection. So you have to be careful, it's not the limitations of human hearing alone that play a role.
Eh... I don't know about that. First off, we don't generally listen to pure sines.
More and more people admit now, that there is something more to be measured than THD and IMD. More specifically, one should use test signal fragments with permanently varying frequency spectrum (like a music) and look for a ways of proper characterization. There is really interesting work along this way, published in russian forum. I do not have enough courage to translate it into english, but who has proper capabilities, will find really interesting results. I enclose that published pdf file. Amps with " 0% " THD show quite big measurable errors with dynamically varying signals.
That's alright, we don't have to know the answer of everything. But it's interesting to ask questions. I'm actually very unsure about a lot of things, to a degree that I sometimes - when I'm slightly depressed - am questioning whether we can differentiate amps and DAC's and other stuff from each other and that the whole thing is a placebo thing. Or at least most of it, blind test seems to suggest that. On the other hand, on blind tests people usually fails miserably. Amplifiers, DAC's, cables.... we can't hear a difference on blind tests. But I think blind tests aren't suited for audio stuff.
When we take home an amplifier to audition it, we usually do so on familiar speakers, and it can take several weeks to find out whether you dig it or not. And you are at home in a relaxing environment.
But let's assume that someone can prove that a blameless amp has so low THD that we humans can't hear it, then the whole diyaudio will be pointless. If so, we would all go to the store and buy ourselves a "blameless" amp.
I mean, why invent the wheel over and over again? Perhaps because it's fun and that we all are silly.
Btw, Sgrossklass, there is a German amplifier brand named SAC, have you heard of that one? They are selling an amp that has 80dB of NFB over the entire audio range. Or at least they claim that, but personally I think it's impossible, they must have fiddled with the numbers - perhaps they have some local loops that they count in.
The magazine 6moons reviewed it. They concluded that it was stunningly detailed, but also that they found the sound "utterly dry".
Perhaps real life is very dry, I don't know.
Sorry if I have offended you, Wahab.
And DF96 - I didn't really understand what you meant.
OK, Wahab. You don't think we shall run after one parameter. Fine. But the way I see it, a blameless amp is designed to minimize THD, and by doing that, IM as well. What other parameters is it designed to minimize? Besides PSRR. What do you mean with DF? I'm not used to all acronyms here on diyaudio.
I see that people gets offended by me calling them, THD chasers, I'm sorry for that. But I'm not sarcastic, what are those other parameters? Please enlighten me.
And DF96 - I didn't really understand what you meant.
OK, Wahab. You don't think we shall run after one parameter. Fine. But the way I see it, a blameless amp is designed to minimize THD, and by doing that, IM as well. What other parameters is it designed to minimize? Besides PSRR. What do you mean with DF? I'm not used to all acronyms here on diyaudio.
I see that people gets offended by me calling them, THD chasers, I'm sorry for that. But I'm not sarcastic, what are those other parameters? Please enlighten me.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.