Well, I seen the mantra over and over again, endlessly in forums and posts, by those pushing the objectivist agenda -- hence the use of " ' " characters ...
However, if one accepts that choice of passive components does affect, alter the audible quality significantly - which is what a lot of designers cheerfully acknowledge they do to 'voice' their designs - then what is the 'truth', which is the combinations of components used for a particular design that is, at least, 'correct'? Either the parts used make a difference, or they don't - you can't have both things at once ...
However, if one accepts that choice of passive components does affect, alter the audible quality significantly - which is what a lot of designers cheerfully acknowledge they do to 'voice' their designs - then what is the 'truth', which is the combinations of components used for a particular design that is, at least, 'correct'? Either the parts used make a difference, or they don't - you can't have both things at once ...
Man amp or girlie amp .......
Wayne, it's inappropriate for the author to judge his own work, it's very bad taste. How well I did should be judged by other people, certainly not me. Every father loves his child and is hardly objective.
Some of my calc indicate that in impulses, its output will be limited by the supply line voltage above all, even into 2 Ohms. I don't need that much, but I went for it as a matter of principle - once completed, I will make it all available publically, and there's no telling what someone else may run into.
Frank, notwithstanding the text above, it IS a fact that say 'scopes can give you some valuable information and even a pointer (not guaranteed, but very likely) to how it may sound.
Just depends what you are looking for. For example, harmonic decay order and behavior - if that is as it should be, it is LIKELY that the amp will sound at least very good.
Or what is your settling time like - if it's reasonably brisk, the amp is LIKELY to sound clean and focused.
And while it is certainly no ideal tell-tale, it can shorten your work very considerably. It could save you weeks of work.
The thing is, instruments are a sort of a counter guarantee. If your harmonic decay pattern looks like it should, the amp will PROBABLY sound fine, but if the decay pattern is wierd, it is guaranteed not to sound very good.
But, as I said, in the end, it's all down to the designer and his hearing.
In my experience, ANYTHING that has gone from a principle to a mantra has also gone sour along the way. Never heard a mantra which didn't end up with a bitter taste.
How do you 'audibly eliminate' the (alleged) distortion? My guess is that most of the tweaks you perform actually change the frequency response more than any distortion. Certainly swapping passive components will do this, unless you ensure exact matching of both value and parasitics.fas42 said:
Similarly, those people who think they can 'improve' an amp by reducing NFB are merely expressing a preference for peaky bass due to a low DF. Those same people, if they could overcome their abhorrence of tone controls, would probably like to turn the bass up a bit. They can't admit that, as tone controls are out of fashion for audiophiles, so they achieve a similar result by degrading their amp. Fortunately, as a bit more distortion is not that noticeable, they can get away with this.
The type of distortion that 'offends' me are largely due to non-linearities at low levels, inadequate power supplies, and interference - I doubt FR comes much into play here. The nonlinear behaviour typically occurs over poor metal to metal interfaces, wherever mere low level pressure over poorly mating surfaces is expected to perform an adequate job: the classic example is the RCA plug/socket interface.
Power supplies are well known for their generally below par performance - the voltage rails sag and modulate badly when stressed, and are also extremely susceptible to mains noise interference. And interference both via RF and mains is a major quality killer, this is one of the most difficult to fully tame.
I don't worry about resistors, capacitors are far more fertile grounds to improve behaviour, where they are plainly inadequate.
Until you do actual experiments, this is just handwaving and conclusion-jumping.
It would be interesting to repeat the Carver challenge with a class D amp and a 100 watt chip amp .
Many years ago a friend and I thought about fooling the press with a class D tube amp . I knew nothing of the problems and probably wouldn't have gotten very far . The idea was to use 807 tubes as they are accepted as audio devices and were cheap . Genuine RCA at $10 . I just read a pair of EL 36 TV tubes gives 300 watts class C so think it might have been a winner . The plan was to say a tad over 20 watts and see if any of them would have the common sense to test the output . What we expected was for them to say . " One could be forgiven for thinking the amplifier had 40 watts " . Also " whilst obviously of higher distortion than modern amps one soon hears through that and enjoyed the euphonic sound of a good tube amp " . Joy of joys if we had done it . Dead band in tubes , yikes .
Someone has had a stab at it . Before anyone says it wasn't about efficiency . A comment offered was to use Quad ESL direct to the panels and use that as the integrator .
http://www.montagar.com/~patj/tubeamp-pwm.gif
Many years ago a friend and I thought about fooling the press with a class D tube amp . I knew nothing of the problems and probably wouldn't have gotten very far . The idea was to use 807 tubes as they are accepted as audio devices and were cheap . Genuine RCA at $10 . I just read a pair of EL 36 TV tubes gives 300 watts class C so think it might have been a winner . The plan was to say a tad over 20 watts and see if any of them would have the common sense to test the output . What we expected was for them to say . " One could be forgiven for thinking the amplifier had 40 watts " . Also " whilst obviously of higher distortion than modern amps one soon hears through that and enjoyed the euphonic sound of a good tube amp " . Joy of joys if we had done it . Dead band in tubes , yikes .
Someone has had a stab at it . Before anyone says it wasn't about efficiency . A comment offered was to use Quad ESL direct to the panels and use that as the integrator .
http://www.montagar.com/~patj/tubeamp-pwm.gif
Last edited:
I'm curious what you consider experiments - I change something, it may have an effect, which is either positive or negative; or, it has null impact. Over a period of time, many years of it in fact, I've determined that certain procedures, actions always improve the performance of what I'm dealing with. So, strangely enough, this tends to encourage me to think I'm going in the right direction.
Many of the things I do automatically now, because years ago I effectively did and undid the tweaks numerous times - the poor connection problem was a nightmare until I bit the bullet, and hardwired everything. If you want "proof" that lousy connections are not good for electrical circuits I think one can track one or two studies ...
When people say I guess is it a figure of speech or do they do experiments ? No offense as I am as guilty as any of this . I always say something when I have evidence when I can . Usually my evidence is nothing change however the sound seemed to . Often I find that in open loop plenty changed . Now what does that tell you ?
I could suggest close loop is like boiling water . If it was drinkable before why do it ? Class B and output impedance I guess . I guess what Dvv says is Pasteurize your water / amp ? Guessing , we have to sometimes ?
I think some say science is the proving of a well thought out guess ? Supporting that guess is tough as others take great joy in proving it of no importance . I seem to remember that the last conic section to be explained was the hyperbola . It was solved by a geometric rotational method . It was derided by the purely numerical mathematicians . That was until another fleshed it out with the usual rigor they demanded . I forget which of the great mathematician it was . I do remember he wouldn't accept credit for it and refereed back to the geometric proof . That is the one taught today to students to spark their interest . To be honest to prove conic sections to one like me with no visual concept would be to shut very firmly the door .
I could suggest close loop is like boiling water . If it was drinkable before why do it ? Class B and output impedance I guess . I guess what Dvv says is Pasteurize your water / amp ? Guessing , we have to sometimes ?
I think some say science is the proving of a well thought out guess ? Supporting that guess is tough as others take great joy in proving it of no importance . I seem to remember that the last conic section to be explained was the hyperbola . It was solved by a geometric rotational method . It was derided by the purely numerical mathematicians . That was until another fleshed it out with the usual rigor they demanded . I forget which of the great mathematician it was . I do remember he wouldn't accept credit for it and refereed back to the geometric proof . That is the one taught today to students to spark their interest . To be honest to prove conic sections to one like me with no visual concept would be to shut very firmly the door .
Last edited:
You have defined "confirmation bias" almost perfectly. And nice framing about what *I* consider to be an experiment- it's not me, it's a few hundred years of researchers figuring out how to determine what's real and what's not.
There's plenty of literature out there describing how sensory experiments are set up, performed, and interpreted. Besides my little summary in Linear Audio Volume 2, see, for example, Floyd Toole's excellent book, the outline of data interpretation at tonmeister.ca, numerous papers by Lipshitz and Vanderkooy, and various ramblings at seanolive.blogspot.com. This list represents about 0.01% of the available literature. Although not related to audio, the book "Flim Flam" gives delightfully entertaining and educational accounts about how doing poorly designed experiments can lead to hilariously wrong conclusions due to the way human brains work.
Or you can play away in blithe disregard of over a century's worth of careful research on how perception works.
Both true and untrue.
True if you take model designed for say 40 dB of global NFB and reduce this to say 20 dB at the spur of the moment. You'll be lucky to get away with no oscillations the first time it encounters an evil load.
Not true if the amp has initially been designed for 20 dB of global NFB. To be able to RELIABLY use just 20 dB of global NFB, your amp needs to have some serious performance under open loop conidtions, and, even more importantly, it has to be very stable open loop. Not ultimately so, but VERY stable, let's say it should be stable into 4 Ohm loads with a highly reactive component (e,g, impedance drop to beow 3 Ohms, phase shift -40 degrees), which is a tall order, but quite possible if you work at it long and hard enough.
As for a bit more distortion, I agree completely. I have long recognized that the true purpose of NFB in audio power amplifiers is to ensure superlative stability under very adverse conditions. With 20 dB of global NFB, it is not too hard to achieve a THD factor of <0,07%, 20-20,000 HZ, into 4 Ohms, 28 Vrms output, assuming a triple output stage power trannies. But you do need to know the transistors you are using inside out.
Rising THD and even IM figures under heavy loading (meaning full nominal power into a difficult load) is incomparably less obnoxious than having the amp choke and gasp for voltage, current or both. Your transient peak may not be perfectly executed, but at least it will be executed. It generally will not be, although this assumes a truly evil load, say a nominal 4 Ohms fropping down to 3 Ohms with a -60 degree phase shift. Most amps will choke instantly, or their protection circuits will trigger.
As in the car consumption business, here we have the question of what is taken as the yardstick, what is defined as "an evil load". A 3 Ohm load with a -60 degree phase shift is equivalent to a load impedance of 1,5 Ohms ...
The big problem we have is that the task being performed is to 'fool' the hearing system into perceiving something which is not actually there. And probably no-one can precisely specify what are all the necessary conditions for that to happen -- we can get a pretty good idea by doing all manner of little exercises, which 'test' one aspect or another of that, but it's mightly hard to come up with the 'Maxwell Equations', so to speak, which completely determine adequate quality -- it's always a mind game ...
Fortunately, I stumbled onto the 'disappearing speakers' trick, and that's proved to be an excellent marker - how close, how far am I from that very clearcut behaviour of a system. So, the sound will nearly always change when one does one thing vs. another - but, has it moved me any closer to achieving that goal ...?
Fortunately, I stumbled onto the 'disappearing speakers' trick, and that's proved to be an excellent marker - how close, how far am I from that very clearcut behaviour of a system. So, the sound will nearly always change when one does one thing vs. another - but, has it moved me any closer to achieving that goal ...?
Frank, the problem of low level operation is one NOT inherent to class AB, let alone pure class A.
It is inherent to the bean counting, which rules the industry (nay, the world today). Using their heat sinks, I am not one bit surprised that most Japanese amps use 20-30 mA of quiescent current, while most Europeans (what's left of them) go for 40-60 mA. I am much happier with 110-130 mA per device, and believe right or wrong that going above 130 mA per output devices no longer brings any audible effects, thereafter you might as well hog it and go class A all the way.
But that means that each output device is dissipating just under 7W at idle, and for a three pair amp, that's 42 Watts. This clearly implies much more efficient and robust heat sinks than are found in dinky little commercial amps. This in turn means more internal real estate, more weight and (AAAAARGH!!!) more money. You can't get away with heat sinks rated at say 1.8 or 1.5, you need to get down to 0.8 or 0.7 - I use two 0.6 heat sinks and believe me, they do get nicely hot.
Payback is that the amp works to about 3 Watts/8 Ohms in pure class A before it crosses over to class B. Given that statistically I spend 98% of my listening time with a constant dissipation below 1W, it turns out that this way I spend like 99% of my time in pure class A. Class B serves me only to do occasional hard transients, or when the feeling takes me to be a headbanger every now and then. But that's 15 lbs of heat sinking alone.
If "careful research" has identified what I, Pano and others have experienced at times in audio replay I would be interested to know. Since I personally have not come across any such work, I will in the meantime have to continue in "blithe disregard", since I find what my endeavours can deliver well worthwhile. The usual standard of audio replay is too lacklustre, and unsatisfying in the medium and long term, to accept as the last word, so I shall have to persist on my lonesome way ...
Here's a suggestion. Rather than ennumerate randomly, let's all try to list, in order of importance we think should be, the most important apsects of power amp design according to us.
My list is rather short:
1. Unconditional stability even into very difficult loads;
2. As symmetrical as possible rise and settling times;
3. A specific harmonic decay rate down to 7th harmonic;
4. As good as possible TRANSIENT power delivery, limited only by supply voltage for peaks lasting 100 mS or less;
5. As clean as possible power supply, possibly with distributed filtering;
6. Intimate designer knowledge of the properties and nature of the devices used;
7. Design for low global NFB rates, open loop full power response should be 40 kHz or more, and
8. Full but unintrusive electronic protection from overheating, short circuits, too much DC, overvoltage and overcurrent.
That's it.
Actually, under 5., I prefer to use separate lines for the VAS and the output stage, with the VAS lines electronically stabilized. If for any reason not feasible, then the input stage and the VAS should be separated from the rest using capacitance multipliers.
My list is rather short:
1. Unconditional stability even into very difficult loads;
2. As symmetrical as possible rise and settling times;
3. A specific harmonic decay rate down to 7th harmonic;
4. As good as possible TRANSIENT power delivery, limited only by supply voltage for peaks lasting 100 mS or less;
5. As clean as possible power supply, possibly with distributed filtering;
6. Intimate designer knowledge of the properties and nature of the devices used;
7. Design for low global NFB rates, open loop full power response should be 40 kHz or more, and
8. Full but unintrusive electronic protection from overheating, short circuits, too much DC, overvoltage and overcurrent.
That's it.
Actually, under 5., I prefer to use separate lines for the VAS and the output stage, with the VAS lines electronically stabilized. If for any reason not feasible, then the input stage and the VAS should be separated from the rest using capacitance multipliers.
'Performance' is an objective word, in the context of electronics. I assume you have careful measurements to confirm this?fas42 said:
Could you offer an explanation of how 'lousy connections' can audibly affect domestic audio yet a similar effect seems not to be seen in radio or instrumentation, where low level IM would be much more damaging, unless seriously bad connections are used?
Strange. Unless your amp is unstable open loop (very unusual), NFB is like to degrade stability. The true purpose of NFB is well-described in all the standard textbooks: lower distortion, lower output impedance, wider frequency response, behaviour set by passive components rather than active components etc.dvv said:
- Status
- Not open for further replies.