"Objectifying the Golden Ear" in latest Electronics Design ...
in light of this discussion, imagine my surprise when i got the latest electronic design trade journal today. those that don't get this journal can read the article here:
http://electronicdesign.com/Articles/ArticleID/16804/16804.html
bruce hofer comments, too.
fyi
mlloyd1
in light of this discussion, imagine my surprise when i got the latest electronic design trade journal today. those that don't get this journal can read the article here:
http://electronicdesign.com/Articles/ArticleID/16804/16804.html
bruce hofer comments, too.
fyi
mlloyd1
Bob Cordell said:
Bob,
Did you say it explicitly? No. But it is implicit in every argument that you (and others) have made. Not once have you discussed their power supply design, choice of active devices, choice of passive devices, topology (excluding feedback loops), etc. It's always, always, always: Halcro got good reviews, Halcro has ultra low distortion, therefore...
And even if you don't follow the ellipsis with spoken or written words, does the conclusion not jump out at you?
Indeed, it's a very old trick used by lawyers attempting to sway juries. And they get away with it simply because it is so difficult to pin what someone didn't say on them.
Your reasoning--that lowering conventional distortions might also reduce unconventional distortions--is, in fact, a double-edged sword. To what unconventional distortions do you refer, sir? Would you kindly enumerate them for me?
Well, that's kinda the problem, isn't it?
By definition, we don't know what we don't know.
And yet, I am reminded of your continual castigation of John Curl for attempting to lower other distortions, simply because his distortions differ from yours. In light of what you've said above, you might be a little more forgiving in the future.
And his attempts also appear to have lead to good reviews.
Now, at this point I would like to note several things:
1) I don't read Stereophile. I have read an occasional single issue passed to me by a friend. All told, I've probably read three or four issues over the years. Those issues have not impressed me to the point that I have taken a subscription to the magazine. I will, however, defend the usefulness of listening as an evaluation process. If that means I must come to Stereophile's defense, then I will do so, even if I do not necessarily agree with any individual review or reviewer.
2) I do read Absolute Sound. Unfortunately, their usefulness to me has declined considerably since they abandoned their multiple reviews of the same product strategy. Worse yet, Patrick Donleycott (whose last name I have probably misspelled) is no longer with the magazine. Between PHD and HP and one or two others of the old crew, I could usually get a pretty good idea whether it was worth my trouble to listen to something. I still read it to keep up with what's new, but I find less and less in each issue to interest me.
3) Harry Pearson (of Absolute Sound) used to like one of the Halcro amps, but dropped it rather quickly. What reason for his disenchantment? I don't know. However, any experienced listener has surely fallen in love with something that sounded different, only to find that, given time, what was different was not necessarily better. Was that what happened with Pearson? I am only suggesting a possibility, not necessarily a probability, and certainly not a certainty.
4) I have not heard any of Halcro's products. I am scrupulous about not voicing an opinion on something that I have not heard. I have heard many high-feedback designs over the years (I used to work in the retail end of the market) and never found one that I wanted to live with long term. They ranged from Yuk! to initially exciting, but annoying over the long haul. From that I conclude that the odds are against a Halcro design being my "ultimate" amplifier, but it's a long ways from 'odds against' to 'impossible.' Unfortunately, high end retail is nearly nonexistent in the Southeastern US because it's one of the poorest portions of the nation and there's little money to be made even when two channel audio is in fashion. As such, it's unlikely I'll hear a Halcro anytime soon.
5) My experience in listening to live classical and jazz has tended to lead me towards lower, not higher feedback in the circuits I play with. Does that make it impossible that a high feedback design might somehow quell distortions other than THD and IM? Absolutely not. But my experience leads me to the conclusion that THD and IM are, to put it mildly, flawed. Enough models have been put forth that indicate that feedback actually creates new problems that it's hard to claim that there are no proposed mechanisms for new distortions. Are they, in fact, relevant? I don't know. All I can say is that I have found by trial and error that wide bandwidth and low negative feedback sound more like real music. Does that mean that I endorse TIM, SIM, HIM, FLIM, FLAM, and BAM? No. (Actually, I probably shouldn't have put FLIM FLAM in there because some half-wit will take it upon themselves to interpret that as meaning that I am ridiculing other proposed distortion mechanisms. Nothing could be further from the truth.) I feel that we desperately need new ideas as to how to measure distortion. It's just that I have not studied all the proposed distortions in depth and do not profess to have an opinion as to their validity or lack of it.
Grey
John, have you ever considered designing tube gear (again?)? Judging from these forums, it's a much friendlier society.
I think the fundamental problem is there isn't a standard method for measuring audio equipment using a non-periodic random signal. Using a simple sine wave (or square wave or sawtooth wave) is a somewhat crude at this stage of the game, don't you think?
John
Man, this thread moves right along! John, remembering that I'm strictly a hobbyist with limited time and resources, here's what I do. First, you need a differential amp. There are lots of schematics around, but you have to have trims for AC and DC CMRR. I actually use an old Tek 1A7A that I've carefully tuned up, but no doubt one could build something far better. The ranges I use most are probably between 1mV/cm and 10uV/cm. You need to be able to put matched probes on different parts of the chassis, and different parts of the same trace or wire, and clearly see what's going on. Next, you need some attenuators with very fine adjust-ability to match signals. Typically, I'll start by feeding both channels of a power amp with a square wave and use the diff amp to compare the output from each channel. The attenuator is placed on whichever channel has the most gain, and adjusted for minimum observed signal. There's always some easily seen difference, and if it can quickly be tied to something obvious, like cap tolerances between channels, that gets fixed. Now, in goes the (mono) music signal. I use a decent CD player for convenience- minimal artifacts and the ability to repeat sections ad nauseum, but IMO it's not sufficient. I just don't have the time to pursue this with LPs, and suspect an arb gen would also be extremely useful. If everything is well matched, there should be almost no difference between channels.
Now, I change something in one channel, say a coupling cap, feedback method, or whatever, and compare the two channels. I'm also listening through headphones from the CD player. As I see differences, I change the gain trim and try to determine if the difference is something in the bass, mid, or high end. No doubt an SA of the difference signal would be most helpful, but I haven't gotten that far. With that clue, and a fair knowledge of component and circuit defects, I try to form a theory of why the change did what it did. I might make another change and see if I can make it worse- if a theory is good, it should allow predictions.
The same technique can be used to compare completely different circuits, but the result doesn't come out with some nice neat name. It's more like, "Gee, this amp blunts the beginning of the bass drum waveform more than that one." IMO, trying to put names on things like TIM isn't very enlightening either. The point is, once you can isolate the difference to some type of transient within some frequency range, the hunt for the cause is easier. Listening and watching the error signal at the same time is fairly enlightening.
This is an awfully superficial description of what I try to do. I'm not at all as obsessed as some, and there's way more one could do with SA, with input-to-output comparisons (tough, because all amps are quite flawed at this level), and especially with the arb gen (could also be the PC). Separate power supplies with different regulators and bypassing might be interesting too. Though I do THD measurements and all the rest, I like the differential techniques because I can use music or test signals, and I don't have to put a name on the distortions. It's tough to put any blame on the test, as long as the two channels of the diff amp match, and that's easy to verify by measuring the same point. It won't reveal the one truth, nothing will, but it will reveal differences to just about any degree you want.
Now, I change something in one channel, say a coupling cap, feedback method, or whatever, and compare the two channels. I'm also listening through headphones from the CD player. As I see differences, I change the gain trim and try to determine if the difference is something in the bass, mid, or high end. No doubt an SA of the difference signal would be most helpful, but I haven't gotten that far. With that clue, and a fair knowledge of component and circuit defects, I try to form a theory of why the change did what it did. I might make another change and see if I can make it worse- if a theory is good, it should allow predictions.
The same technique can be used to compare completely different circuits, but the result doesn't come out with some nice neat name. It's more like, "Gee, this amp blunts the beginning of the bass drum waveform more than that one." IMO, trying to put names on things like TIM isn't very enlightening either. The point is, once you can isolate the difference to some type of transient within some frequency range, the hunt for the cause is easier. Listening and watching the error signal at the same time is fairly enlightening.
This is an awfully superficial description of what I try to do. I'm not at all as obsessed as some, and there's way more one could do with SA, with input-to-output comparisons (tough, because all amps are quite flawed at this level), and especially with the arb gen (could also be the PC). Separate power supplies with different regulators and bypassing might be interesting too. Though I do THD measurements and all the rest, I like the differential techniques because I can use music or test signals, and I don't have to put a name on the distortions. It's tough to put any blame on the test, as long as the two channels of the diff amp match, and that's easy to verify by measuring the same point. It won't reveal the one truth, nothing will, but it will reveal differences to just about any degree you want.
Sure, why not? A signal is a signal. Even a symphony has a Fourier transform. The question will be interpreting what the differences mean, something much more acute in speaker design.
rdf, AFAIK, the MLS signal is sampled during the excitation. That's why it's advantageous from a S/N perspective compared with impulse. The impulse is derived from the MLS via autocorrelation, but in the actual measurement, the DUT is seeing continuous signal.
Conrad Hoffman said:
Differential comparisons have been on the table for at least
32 years and refined to about -80 dB or so (the figure I was
getting in 1975). Unfortunately, the results have yet to generally
reveal to me what makes one amp sound better than another.
The exceptions are the cases where one amplifier could almost
be considered "broken".
$300, the cost Brasfield cites for his setup, is the reason for the acute fanny irritation of some of the adherents to this thread. I would be upset as well if I was trying to sell a kilobuck amplifier.
a long time back I used to make "unsharp masks" with Plus-X or TMax 100 or copy film -- produced incredible B&W prints from my 4x5 negatives -- i am ashamed to admit that with a little conscientious PhotoShopping I can do the same thing in a fraction of the time, still using a $150 Epson scanner and an $800 Epson printer.
Tempus fugit
a long time back I used to make "unsharp masks" with Plus-X or TMax 100 or copy film -- produced incredible B&W prints from my 4x5 negatives -- i am ashamed to admit that with a little conscientious PhotoShopping I can do the same thing in a fraction of the time, still using a $150 Epson scanner and an $800 Epson printer.
Tempus fugit
I guess the comparisons are good enough for my ears and no better- I can hear certain coupling cap changes, and can see artifacts in the upper frequency ranges associated with them. What other tests can do that? I also question my opinions when minimal differences show up, and often I've fooled myself about things that really didn't have an effect. It's a reality check because we're dealing with very subtle sonic differences, and anybody who thinks they never get fooled is just kidding themselves. Remember, I said the method wouldn't reveal the one truth. I don't think any test can sit in judgement of good and bad, only tell you about differences from some reference, be it numeric or another piece of hardware. What if the technically perfect amp were developed, and proven through some yet unknown method to be so, and yet it sounded much worse than "A" grade, whatever that is. I see no fundamental reason to rule out that scenario. It occurs to me (since photography was another hobby) that if I had a control panel for reality in front of me, I could probably adjust the color, saturation, brightness and contrast slightlyl to improve on whatever was out there. Without knowledge of the original scene however, I'd have no way of setting them to "correct".
Nelson, that's about right on the -80dB figure, not near as good as a THD measurement, but if you SA the results, you get way below that. IMO the output will be more useful as it says something about errors in music reproduction, not artifacts from a single tone- where's the amp doing a good job and where's it falling short. In the Geddes thread on distortion below the noise floor, the premise is that things at -100dB or more are important. I think the jury is still out on that one, and since I see obvious problems with music reproduction at -80, and have more confidence that we can hear -80, that seems like a good place to try and improve things.
rdf said:
Well, I don’t recall claiming to be a "top expert" in any field, but I thankyou for holding me up to such a high standard and for the valuable, albeit somewhat less than insightful character evaluation, sir.
My general sense of self-assuredness is hardly as fortuitous as you’d like to believe, despite the stridency with which I have attacked some of the more asinine utterances made here.
rdf said:
Well that's fantastic, but it would hardly get much of an argument from me, despite the fact that I would probably go even further with your last sentence...........
A good task for all of this energy might be to discuss what form a new type of measurement might take and why.
The value/shortcomings of even the cutting edge in measurements does not bring us any closer to a meaningful picture of the differences that ARE clearly hearable by experienced practitioners of this art.
And, although I seriously doubt from experience that a bridge can be built between the artist and the engineer, I do believe that to move forward there must be a new way of measuring what we hear... This coupled with the available technolgy for sampling and manipulating complex data sets opens up the very real possibilty of creating a new measurement approach. Defining the variables and their relationships would be a starting point.
IF we start looking at our amplifiers as a combination of resonant responses and filtering, what do we see? If we get away from steady-state test signals and start comparing input to output, based on full spectrum data with real loads, what is there to interpret?
I have my ideas; anyone else? With what's availible from a software standpoint, creating a test setup is not as out of reach as it seems. The real challenge is defining what to look for.
Regards, Mike.
The value/shortcomings of even the cutting edge in measurements does not bring us any closer to a meaningful picture of the differences that ARE clearly hearable by experienced practitioners of this art.
And, although I seriously doubt from experience that a bridge can be built between the artist and the engineer, I do believe that to move forward there must be a new way of measuring what we hear... This coupled with the available technolgy for sampling and manipulating complex data sets opens up the very real possibilty of creating a new measurement approach. Defining the variables and their relationships would be a starting point.
IF we start looking at our amplifiers as a combination of resonant responses and filtering, what do we see? If we get away from steady-state test signals and start comparing input to output, based on full spectrum data with real loads, what is there to interpret?
I have my ideas; anyone else? With what's availible from a software standpoint, creating a test setup is not as out of reach as it seems. The real challenge is defining what to look for.
Regards, Mike.
Charles Hansen said:
Well thankyou fro the critique of my memory and reasoning skills.
But why, may I ask, did you design an amplifier with such low open loop distortion?
To merely prove that such exemplary objective performance can be attained without the use of negative feedback? I somewhat doubt it. By your own admission it took a considerable R&D effort to get the THD as low as what you did and I think that it would be a bit disingenuous to deny that there was at least a casual relationship between good objective performance and good subjective performance - or that the former was, at least in part, a design aim.
Your latest amplifier does measure a lot better than the previous model, does it not?
Charles Hansen said:
There are people here who design for high open loop linearity just like you do before applying these things that you merely deride as "band-aids". Those who eschew things such as EC, nested feedback loops or global NFB do not have a monopoly on linearity.
Just a thought.
- Status
- Not open for further replies.