Sound Quality Vs. Measurements

[dvv]

Originally Posted by jcx

care to illustrate that with circuit examples?

We did, JCX, go back a few pages and it's all there.

Since Thorsten is on one track, and I on another, and since both projects are ongoing, I expect there will be updates, but cannot say when.

Even our topologies are broadly similar, the two biggest differences being that Thorsten opted for a Single Ended input stage and a lateral FET driver pair, while I stuck to classic fully complementary and a predriver-driver stage. Not really surprising, Thorsten and I go back a long way.

Speaking strictly for myself, this one I'm doing for myself and for the fun of it, so once it's finished, built and tested in real life and time, it will be compiled as a complete project, including PCB artwork, and posted here and elsewhere. That should happen nearer the end of the year, since I am developing this on my free time, which I don't have too much of, unfortunately, so progress is not exactly lightning fast.

[dvv]

Originally Posted by bcarso

Yes, I'm aware of the HD and IMD higher terms appearing with a little feedback, even local. I'm not sure I've yet seen convincing examples of the pervasive notion that global feedback gives rise to these high-order terms.

What I propose: two amplifiers, same nominal transfer function. As similar as possible but one with a lot of local and correspondingly less global FB. Suppose that all of the terms, HD and/or IMD, are very low.

Do we still have people who will hear the supposedly deleterious effects of the higher global feedback? I'm presuming that we don't violate with obvious problems like having insufficient slew rate.

Brad, I can't speak for Thorsten, but in my case, it's a simple example of real world order of things.

For many, many years now, I have held Harman/Kardon's products in high regard, not all to the max, but some yes, to the max. I own two of their integrated amps, one from 1993, the other from 1999. The first is a SEPP job, rated at 50/70 WPC with 17 dB of global NFB, and the other uses two pairs of output devices per channel, delivers 85/130 WPC into 8/4 Ohms, uses just 12 dB of global NFB, and is said to be capable of very high impulse currents. That's HK's "HCC", or "high Instantenious Current Capability", which is severly let down by zero explanation for how long, into what loads, etc.

The key constant was the uncommonly low overall NFB factor. These amps always did, and still do, sound more open, more relaxed and somehow "cleaner" than the vast majority of their competitors. Especially when you clean up after HK, which has, ever since I came in touch with them, consistently managed to find as poor as possible trimmers and pots. If you want to know what a junk pot looks like, look inside an HK. I speak from personal experience, because I have changed quite a few of them, and each and every time, the effect ranged from clearly audible to surprisingly large positive effects. And I didn't stick in anything particularly wild, just Alps Blue pots.

So it made me wonder. Looking at their schematics (which I have a collection of, if anyone's interested, let me know), I saw some to my mind very clear and focused thinking, rather consistent across the line and over time, although of lately, they have changed their overall outlay, but I haven't had an opportunity to hear those new samples, so I can't comment or compare.

Anyway, I investigated further with some local work, taking a few low overall NFB samples and comparing them with typical and not so typical commercial fare, yielding the result that I concluded that those with more local and less overall NFB generally sounded better, cleaner, less strained, more open, with better sound stages. The key word here is "in general", some were sort of let downs, others shined, etc.

Now, I don't mind admitting that I am a freak, some of my views would probably be called extreme - for example, I believe there's no such thing as too big a power supply, I think of a say 300 VA toroid as a dinky little thing. However, please note that I refer to personal projects, not commercial offerings, therefore there are no price constraints. The sky (and my wallet) are the only limits.

On the other hand, unlike our own Wayne here, I do not believe one really needs more than say 100W/8 Ohms at home, unless one has unusually inefficient speakers (say, less than 87 dB/2.83V/1m), a view which is probably slanted by the fact that I have a rather small room. But, that's 100 Watts into 8 Ohms, which should become ideally 200 Watts into 4 Ohms and 400 Watts into 2 Ohms, in short, I want real world power, not power on paper. Or as near as doubling down, as Wayne says, as possible.

I also believe hard in fully protected electronics and speakers. The logic that protection circuits inhibit good sound may be true only of poorly designed and/or adjusted protection circuits, and I must admit to having had dealings with Japanese products with unusually aggresive protection circuits; amps made to work well into pure 8 Ohm loads, and if it should drop to say 6 Ohms, they are in trouble because they are highly strung up, pushed to the limit.

And there you have the design brief. It should have real world power, not just as specs, but in real life, and it should use no more than 20 dB of overall NFB, while still having its overall THD and IM at below 0.1% worst case, i.e. 2 Ohm loads.

And, of course, it has to sound good, where "good" primatily means my own definition of good, but also that MOST people agree it sounds good.

Any objections, suggestions, ideas, views, Brad?

[dvv]

Originally Posted by tvrgeek

dvv,

Hmmm. This could be key to what I am chasing. Getting less high order distortion while not crippling bandwidth like the Rotel does. In the mean time I am working on notch filters for the crossovers to hopefully make them less susceptible to the possible effects. Unfortunately, no answer yet from e-mu on why their drivers are crashing. I have gotten conformation I am not the only one who has had this problem.

On how, after increasing the size of degeneration resistors, not sure where to go from there. So much to learn. Parallel the MC cap with a resistor?

Commercial product designers are, unfortunately, bossed by their sales people and are pushed to produce great specs rather than great sound. One of the things they are forced to do is to use a lot of overall NFB.

This yields better classic THD figures, however, since it also implies greater open loop gain, this also means a more narrow open loop power bandwidth. This in turn tends to produce more HF distortion open loop, and hot ding, you CAN hear it.

For around 30 years, we have accepted the axiom that an amplifier should have an open loop bandwidth of no less than 20 kHz, yet you will find in many commercial offerings an open loop bandwidth of 4 kHz and such like. But whack it with a lot of global NFB, and you get classic THD figures like 0.01% or less.

Mind you, even so, you can still end up with a good sounding amp, but I contend that if given a free hand, those same designers on their own would do much better in real world sound, but with worse classic THD and IM figures, which would be harder to sell to the spec reading, little understaning public.

On this matter in general, I think John (Curl) is much better versed, he's been at it, in both lower cost and cost no object arenas, for decades now.

[wahab]

Originally Posted by dvv

For around 30 years, we have accepted the axiom that an amplifier should have an open loop bandwidth of no less than 20 kHz, yet you will find in many commercial offerings an open loop bandwidth of 4 kHz and such like. But whack it with a lot of global NFB, and you get classic THD figures like 0.01% or less.

Mind you, even so, you can still end up with a good sounding amp, but I contend that if given a free hand, those same designers on their own would do much better in real world sound, but with worse classic THD and IM figures, which would be harder to sell to the spec reading, little understaning public.

This is not true , i simulated a lot of amps including Thorsten s one
and it does not make exception to the rule that low feedback amps
have very high level of IMD , even if loads of active devices are used....

As an exemple , a comparison with a very simple design that has about
66 dB open loop gain and 1 khz OL bandwith.

[nigel pearson]

I once asked my friend Hubert . " What's wrong with Michael " . Reply " He is working to unify Quantum Theory and Einstein ( put crudely ) and is getting somewhere . It wont let him sleep " . My friend's view was it would kill Michael . Well time went by and I would bump into him now and then . I think I would have been able to call him a freind if I had made more effort . You would have to be in the moment to know the magic of what I will say next . One day I say to Michael " what do you think about negative feedback " he answers " well I never have , but if I had I might have said this . Some amplifiers need a lot and some almost none at all " there was then a pause and a very searching look at me . Bursting into laughter he then said triumphantly " you can be sure if one thing , nearly all have the wrong amount ." He then elaborated , however I had got the point . It changes someone if they listen to Michael . Lets be clear Michael only believed in the mathematical answer . However he doubted the maths if he didn't like the sound , which was often from what I understand . Someone called him Oxford's unknown Stephan Hawking . Another quote is it might be 300 years before we fully understand his work .

[dvv]

Originally Posted by wahab

This is not true , i simulated a lot of amps including Thorsten s one
and it does not make exception to the rule that low feedback amps
have very high level of IMD , even if loads of active devices are used....

As an exemple , a comparison with a very simple design that has about
66 dB open loop gain and 1 khz OL bandwith.

Wahab, do not make judgements on basis of one or two samples only - one swallow doth not a spring make.

My point is, there are many possible topologies, and they will not all behave the same way. That's the trick of it, to pick one and really make it work, which implies that your are literally intimate with the adopted topology, that you know your tansistors inside out, etc, etc, etc. Slapping it together to simply work electrically is easy, but making it work as best as it can, well, that's THE trick and that takes time and effort.

You investigated Thorsten's first sample only. I don't know how far he's got, but rest assured he has worked on it and the latest version, whichever it may be, will work better than the initial one.

It's even possible that he has gone as far as to change the topology.

Just one off hand example. Let's say you decide to use BF 471/472 transistors. They are very hard to work with because they are finnicky and very picky, but are good to use because once you get to know them, their sound can be most rewarding. But, they have their very own set of requirements, especially related to bias. Run them at 3 mA, 4 mA and 5 mA and you will get three sets of results where the 3 mA performance looks like from another planet in comparison with the 5 mA results. Totally another matter altogether, no way anyone would even guess that it's the same transistor.

Then again, much depends on their operating load. Give them 10k, that's one thing, give them 30k and that's something else altogether. Not just THD, but overall behavior.

Their specifications notwithstanding, run them at 3 mA and they will do wonders; run them at 5 mA (constant current) and they may well wreak havoc on you. As I said, very picky, very finnicky. But worth the extra effort.

[DF96]

Originally Posted by dvv

For around 30 years, we have accepted the axiom that an amplifier should have an open loop bandwidth of no less than 20 kHz, yet you will find in many commercial offerings an open loop bandwidth of 4 kHz and such like.

Who is "we"? People who never ever listen to music which has travelled through any op-amp anywhere in the record-replay chain? People who don't understand servo systems? It is not an "axiom" (that would imply some science or maths was coming); more like a popular and persistent myth.

[dvv]

Originally Posted by DF96

Who is "we"? People who never ever listen to music which has travelled through any op-amp anywhere in the record-replay chain? People who don't understand servo systems? It is not an "axiom" (that would imply some science or maths was coming); more like a popular and persistent myth.

Where have you been since 1978? Not on this planet, it seems.

Do read Otala's texts published by IEEE in 1978 and 1979.

[nigel pearson]

F_n = F_{n-1} + F_{n-2} ........................?

[abraxalito]

Originally Posted by dvv

Do read Otala's texts published by IEEE in 1978 and 1979.

Have you been off the planet to miss the explanation from Bob Cordell as to why and how Otala was mistaken?