Contemplating. Let us not argue if amps sound different. Been there, done that. The differences between good ones are very slight and money best spent on better speakers, so not to argue this point. The question is what the source of the difference is. I have no doubt that on some source material, I can accurately identify differences. Other source material, I can not. Now, back to science.

First conjecture: Any distortion or artifact below about -90dB is totally inaudible by a human.
Second: Any artifact over about 22K is totally inaudible even for young females.
Third: Minor differences in equalization less than 1 dB may be identifiable in A/B comparisons, but in sustained listening your brain just " balances the room" so they are irrelevant.
Final conjecture: The above requirements are simple and any decent amp can meet or exceed them. Most really cheap ones can too.

I am going on the concept that if we can't measure a difference, but we can hear a difference, we are not measuring the correct thing.

Hafler showed it is easy to build an amp where the output is identical to the input except for gain in a steady state test. I have several. Very good, but not great. Personal observation are that impact sounds are more effected than steady state. Clapping more than a violin. Eric Clapton Unplugged applause segments are a great example. Clearly different between amps, yet the vocals and guitars sound indistinguishable to me. Yes, differences in speakers make a much larger difference. A poor speaker plays these segments in a truly horrid and unnatural way.

So, what is different between the impact based sound and the string based sound? I believe this is a clue.

What is going on? Suspects in a transient state, not a steady state.
First suspect: Output impedance and how it interacts with the load.
Second suspect: Amp being constant voltage, constant current, or somewhere in between. Yes, this is related to the first suspect.
Third suspect: Phase. Considering what happens in recording , a dark horse.
Fourth suspect: Something else.

Doing tweeter and midrange testing, I noticed a variance between my bench amp (30W Rotel) and a DH160. For grins, I tested a $9 chip amp as well. What I noticed was how the microphone recorded the output of a 3 cycle gated pulse. This is a very transient measurement. I noticed for the same driver, a difference in ratio of initial peak, second and third to the reference drive waveform. Differences in ringing could be noted as well. Now, the mass of the driver, suspension etc. is a huge factor, but I am talking the same driver, no crossover, baffle mounted.

Going to the suspects, I am dismissing phase for the reason stated. I don't have a candidate for the fourth. Back in the 70's, it was TIM, but that has been resolved by wide bandwidth amps.

My conjecture is that the impedance of the amp may have a lot to do with the difference, not just the bass extension as normally thought. (Rx effect on Qts) Could this be the reasoning for the "softer" sound of a transformer coupled tube amp where direct coupled , to me at least, has a more "detailed" presentation?

Anyone go down this route before? I am thinking if there was a quality factor we could develop and built or adjust to it I believe there is no right answer, as there seems to be a preference between tubes and transistors that can not be attributed to long duration transfer functions. Carver showed that with the silver 7 and matching transistor amp experiment. (kind of backfired, but he did sell a lot of tube amps in the process)

The differences between good ones are very slight and money best spent on better speakers, so not to argue this point.I disagree. :p

You may, but it is not contributing to the discovery process.

I am going on the concept that if we can't measure a difference, but we can hear a difference, we are not measuring the correct thing.


True, but until one can prove it (= in a DBT levelmatched), there is no problem. Have there ever been proven?

Are your listening tests controlled?

The ability of the power suppy to provide instantaneous current at higher frequencies is one factor causing lesser amps to have trouble with more transient sound patterns. Also amp slew rate, bandwidth, damping factor, etc factor in there too, but I suspect that the biggest differences among amps which sound good as compared to thse that don't during transients will usually be in the power supplies impedance as a function of frequency.

Nin, Please refer to my original set of suppositions on the theory thread.

Ugly, Power supply regulation( or lack of) would be a very good candidate if the levels were much higher and transients more demanding. Lack of "life" when playing really loud nasty things like Night on Bald Mountain would be a much easier sell. Easy to see possible linear distortion by rails dipping. Actually, in my B&K I doubled the main caps and added 1u poly bypass caps right at the outputs. I "think" it helped mid-range clarity and impact. Bypassing the input and feedback cap made a big LF difference. As these were totally subjective and done years ago, I was not covering them here. Being as objective as I can, I can't really say if I can heare a difference between the B&K and Hafler.

Beyond conventional regulation specs, instantaneous power supply impedance in the nano-second range, does have an effect on transistor switching speed and power dissipation. So does die temperature. Just can't see this would cause either the pulse shape difference as captured, or CD track as I heard, as the frequencies are just not that high. We are talking Khz, not Mhz and in the linear range, not switching.

So, I have a measured observation. A few clues. Back to my other thousands of distractions.

PS: If someone would like to gift me with a really fine amp ( Krell, Niam, Aragon etc) I would be glad to see if in the next 20 or so years while I listen to my entire music collection I could hear a difference between it and my B&K or Haflers. :) Oh yea, as I bi-amp, I need two!

Well, there are challanges where one can get money if one can hear the difference in a DBT, so why not do that? :)

OK, I appreciate you putting things in the terms of actual gear. It sounds like your gear probably really does need to be pushed before the distortions due to power supply noise gets obviously bad. If power supply noise can be ruled out as being the obvious worst case offender then probably next down the list for quality gear is still amp nonlinearities but this time due to the device parameters such as transistor voltage vs current relationships and circuit toplogy such as strength of negative feedback employed, resulting gain*bandwidth etc.

Perhaps part of the problem may be that even the best gear can't hit the types of numbers, per your 90dB quote, even while operating at the gears own listening level/minimum distortion sweet spot that even under ideal conditions all frequencies of distortions are inaudible. Consider some numbers: your 90dB down for lowest level of audibility quote, ie 1/inverselog(90/20)= 0.000032%. I'm not sure what is typical of good gear at optimal operation levels but that number strikes me as being exceedingly low and would likely be a tough thing to achieve for the entire audible band.

Presumably the distortion capabilities of real amps is not a flat line with respect to frequency. Could it be that even with the low levels of distortions achieved by good amps there are still some frequencies that poke above the audibility threshold? I would also presume that an amps distortion profile as a function of frequency would vary among amps. Especially model to model and manufacturer to manufacturer. Perhaps what distinguishes amp sounds is the way that each amp presents what audible distortion remains even if it is too tiny for humans to call out as being distortion related differences.

I'm just throwing out some ideas since I don't really know what modern amps are even capable of when run at the sweet spot for ideal performance. Clearly (at least IMO), distinguishable audible differences among different gear implies different colorations of the signals. The question is: what is the nature of these colorations? Maybe I've made a good case for seeing where at least some of the problem lies.

All good observations. Yea, 90 is strict, anything below about .01% I do not believe is audible. That should be even easier! I did hear once of a Japanese gentleman who could reliably identify a pure tone with .001% harmonic distortion. Only case I ever heard of.

Non-linear distortion is definitely not flat with respect to either power or frequency. It usually rises smoothly for both. A good example is to look at any OP-Amp data sheet. I have never seen evidence of spikes or dips like you can get with mechanical systems. A couple of OEM's, Benidini ? for one, if I remember was pushing micro-phonics and cut open their transistors to pot them. The sample I heard I did not care for. Circa 1988.

FET vs bi-polar is a potential issue as FET's tend to produce even order, while bi-polar tend to odd. But, all of the amps in question have enough feedback that they should be sufficiently low in steady state. I am not considering the 5 to 10% second order of a vintage tube that is like pouring melted butter over the music :) As I do not hear differences in things like strings or voices, but the biggest differences are impact related, I an going back to some sort of dynamic issue.

I need to reduce the ambient noise in my office and make some new and more precise wav files with pulses of different frequencies to see if I can exaggerate what I observed first. Pictures of waveforms with subjective correlation is what I am after.

Thinking back, pure subjective, when I bought my first B&K 140 in 88. I brought home a seemingly endless series of amps. All about 100W. I had some Ditton 44's with Seas tweeters. In those days, I at least believed I could hear big differences. Not scientific, but I was searching for the best I could afford with no special preferences.

Sanyo Plus 55 The amp I was replacing. Was not a bad unit actually.
CM Labs. My other amp at the time. Not bad in the brief time I could keep it from blowing all the outputs. I converted it into a sub amp.
Adcom: "Ballsey", but no definition
Brown: Dark
Bedinni: Left me totally cold.
Aragon. Loved it, but it was $1200. I guess I was too cheap.
PS. Loved it too. Don't remember why I did not buy it.
Acrus. Came out too late, but I thought it was as good as the Aragon.
Krell No complaints, could not afford it. $5K I think.
Trans-Nova. Used. Had 2. Liked their sound but both had power supply issues
MV-50. Classic tube. Weak bass, no highs, but gad what mids.
Rotel. As expected. Competent but not special.
Quad, used. Did not seem to like the Celestions. Not big enough.
Onkyo. At that time, no longer top notch. a "dip" in their history I guess.
Niko used. Vintage and sounded it.
B&K 140, what I bought after hearing one broken in. ( harsh new) I heard one A/B/C against a Krell and a Lynn and could not hear a difference. ( in store, Theil speakers), so I bought the cheap one, B&K. Hence my argument all really god amps sound about the same.

Years later I started buying used Hafler DH220's and a 160 off e-bay. Can't tell them from the B&K. They are my main system now.

The only reproduction experience I ever had that made me believe I was in the room live was a half-track master played on a Revox through Mark Levinson (26 and 27 I think) electronics and original B&W 801's. It was a solo upright bass. AKG tube direct to tape.

See if this picture links.
OK, triple pulse at 600 Hz, near-field pickup. Rations of waves, 83.1%, 100%, 100.5%
The question is with all other parameters exactly the same, will different amps behave differently?
http://i36.tinypic.com/288xiwx.jpg

Forgot to describe. Top trace is the voltage ( reference) output from the amp. Lower trace the mic pickup. As you can see, the amp does a fine job of constant voltage. This is a Rotel RA 840 driving the Zaph midrange.

Is that triple pulse bandwidth limited? I never saw a short sine-wave test signal just switched on and off like that.

I would expect the test signal to look more like Tracks 3 - 5 Multi-burst signals:

http://www.linkwitzlab.com/burst-cd.htm

I just crafted this pulse with a recorded output from a software wave generator, then went in with an editor and sniped it. That is why it is not quite as perfect as it could be. I need to built it into a loop file as having MS player run it is erratic when the analyzer has priority. Things we can do with bits, we could not do with tubes. I picked 600 Hz ( actually dis a bunch) as it is in a range where I expect a lot of power in an audience applause. I have a feeling what I am hearing in in the mid-range. Notice it is not just the amplitude difference, but the first cycle is asymmetric.

I wish I had a scope program that would allow me to slide traces around and overlap them like I cold in the "good-ol-days" The SE scope is very stable, but very flexible. I think it is Zelscope who I need to send a few bucks to as a more useful scope program.

Sigfried's pulses (yes I bought his disk) have a bit more shaping that would make this observation less apparent. It was with his tracks I first noticed the pattern. Although he knows far more about this than I, sometimes discoveries come from those not well enough informed to know they are on the wrong track!

More distractions. Just got my hands on an HK A300 amp. Clean, working, original outputs. Gotta rebuild it. ( caps, pots etc) Should be interesting to add to the amps to test.

I think the best way to measure the differences people might hear with different amp+speaker combos is to measure the actual output in the room with a microphone. What should be measured, I couldn't say, exactly. When I hear differences due to something (what? who knows?) upstream of the speakers it is usually in the timbre of certain instruments in recordings that I am very familiar with. Cymbals are a good example, but I can hear it elsewhere, too. So, if there was a way to measure very accurately in a controlled manner, the characteristic signature (spectrum?) of, for example, a cymbal's reproduction, and then show a difference in that signature with another amp connected to the same speakers, then you would be on your way.

Now, I think that there are a lot of important variables, here. I do not know that measuring in an inert acoustic environment would necessarily be the best place to perform such measurements. It may be that the room's contribution is important to revealing the difference that might be there. I also think that you would probably find that the differences that could be revealed would be very specific to different amp+speaker combos.

But I don't think measuring electrical differences at the amps' outputs is the place to be measuring. I think that (obviously) the speakers are an integral part of the differences that can be heard.

Do you really think that a mic is an acurate enough piece of test equipment to even bother with? Could a mic mask over any differences itself, be there any differences? This would then cause the argument that the reason there was no difference measured was due to the mics. Also you are then relying on the speakers to reproduce the sound perfectly themselves. So which mic, which speakers? Hard to say since everyone seems to like and say different speakers are more acurate than others.

Do you really think that a mic is an acurate enough piece of test equipment to even bother with?Mics are extremely sensitive pieces of equipment, actually. Extremely. Even a 50 year old microphone. Hell, probably a 100 year old microphone. Or older. Accuracy isn't what is so important (though they ARE accurate). Precision (reproducibility) and sensitivity are what are important, here.


Could a mic mask over any differences itself, be there any differences?Sure, I suppose. Then all you could conclude would be that you couldn't measure any difference with that technique and methodology (see below). Not that there wasn't a difference. You're limited by the tools at hand.


This would then cause the argument that the reason there was no difference measured was due to the mics.Well, you could make that argument about anything couldn't you? If you can't measure a difference then all you can conclude is that, using this particular methodology and technique, a difference could not be measured.


Also you are then relying on the speakers to reproduce the sound perfectly themselves.No, you're not. I didn't say anything about that. Speaker accuracy at reproduction is another whole discussion. You use the same pair of speakers for the comparison. That variable stays the same. All that is important is that the speakers be precise (they are). But, within reason, not necessarily accurate.


Hard to say since everyone seems to like and say different speakers are more acurate than others.As I said, you use the same speakers for any one comparison. But, yes, as I said in the post, different amp+speaker combos should be tested. But you don't switch both variables. I think it's reasonable to expect to be able to hear an audible difference between two amps with one pair of speakers yet not hear that difference with another pair of speakers.

Granted, the mic has mass, complex impedance, and it's own odd issues, but this experiment keeps that as a fixed parameter. I am looking for not absolute shape, but a difference between amplifiers. The ONLY variable in this experiment is the amp. All other factors are fixed, so I am looking for a difference, not any absolute measurement.

Only if I can find a trend, this group like this, that like something else, would I then go back to the subjective listening to see if the groups can be identified reliably. I will check again, but I do not remember hearing differences that I could detect reliablly with cymbals. Hey have such complex sounds, maybe a good test. Here is another: I have one Guitar recording, Bream Plays Segovia, Some amps, the strings sound metallic. Others nylon. To my ears, 100% reliable audible differences and not limited to a specific speaker. I can take my CD around to stores and evaluate an amp. This effect is not as obvious as the applause on UNPLUGGED. Actually, this was my main test when I bought the B&K. It was the cheapest amp that sounded nylon. (Aragon, PS, Krell, Lynn were all quite a bit more)

Mic is the usual Behringer.

I am having a heck of a time with PC based scopes. Bought Zelscope. Not too bad. The better one seems to be gone, unsupported but on all the WEB's anyway. In trying to make higher frequency pulses on the PC, I run into the low bit rate issue. Not enough resolution for a measurement. Any recommendations for PC based dual sweep, delayed, single shot storage scope would be appreciated. A 7834 for the PC for those old lab rats. ( limited to 96KHz of course)

Another observation is the way sawtooths are distorted. No comparisons yet (distracted by new old toys), but another possible example where the cone is basically unable to follow the input ( Thank you Mr. Newton, I understand) so how it behaves may be telling on the amps ability to control it. I expect quite a difference in the transitions, but not sure why the cone seems ahead of itself on the rise, and behind on the fall. Phase shift from inductance? I wonder if this is exaggerated in the digital domain, as 20Khz sine wave does not look much like a sine until it has been integrated by the mass of the driver.

Here is another: I have one Guitar recording, Bream Plays Segovia, Some amps, the strings sound metallic. Others nylon. To my ears, 100% reliable audible differences and not limited to a specific speaker.Yeah, this is a great example. But, I wouldn't be so certain that the difference here is not more pronounced with some speakers versus others. But that is sort of irrelevant if you are going to try to "measure" that difference with a mic you would use the same pair of speakers.


But what I am getting at is that, for example, certainly we can (somehow) "measure' the difference in the reproduced sound of a trumpet playing an A440 and a violin playing an A440. There has got to be a way to "measure", quantify, and demonstrate that timbre difference. Spectrograph? Something. Gotta be. And if we can "measure" that, then certainly we can "measure" the difference in the reproduced sound of a viola playing the same note as a violin. That timbre difference should also be "measurable". And if we can distinguish that, then maybe we can "measure" the difference in the reproduced sound of two different violins playing the same note. As subtle as it may be, there is going to be a difference from instrument to instrument. Different strings. Different wood. Different bow. And IF we CAN "measure" that sort of difference then maybe the same methodology can be used to distinguish between the 'sound' of different amps. Of course, the speakers are going to be important, here. I think these sorts of differences will probably be revealed more with certain speakers than others. So pick the speakers and "measure" the differences in the 'sound' of the amps. IF you can.

That is my goal exactly. Looking at the spectral content of a pulse is something I had not thought of yet. Good lead.

A tale of 5 amps.
Cleaned up the HK A300 so I thought I would conduct a test using the most precise measurement tool I have access to. My wife's hearing when it comes to trumpets. Don't know why. I hope some objective measurement will shed some light, but she is hyper sensitive. In seconds she can lay a yes/no on an amp or speaker.
OK, 100W B&K 140. Mosfet, slightly modified.
Rotel 35 W850bx intigrated amp,
Harmon Kardon A300 15W, all original,
Gainclone
6P1 10W tube amp.
The test was the first track of The King James Version. Sheffield CD.
I may have my bias, I know what I paid, and I know what they "should " sound like. She knows if it grates her teeth.

The HK was highly respected in its day. I have been listening to it in my office, and found very musically engaging. It still gets a lot of waxing on the WEB. I thought if tubes were the answer, this would be a ringer, but it got a quick NO. Plugged in the B&K. Much better but a slight edge. Not surprised, this is my standard amp. Then the gain clone. Could not stand it long enough to get to the hard part. Then the 10W China wonder. Not bad, but not quite as good as the B&K. "Different balance". Then the little Rotel, which is my test bench amp. Surprise! This is the first amp that could play this CD without causing her to flinch.
A small very conventional bi-polar easily trumped everything else. I was quite surprised, I actually agreed on all accounts. Mind you, these are all well respected in their price ranges. The B&K clearway trumped a lot of other very expensive amps when I bought it. I did not drag any of my Haflers out. My Knight 35W is mono, so it stayed on display.

So now I have a cadre of amps I can test to see if I can find something measurable different. Audible differences were not as great as differences in speakers ( Used my Paradigm 20's for this test), but I was quite surprised in how big they were. I still have not had any offer of a really good amp for the test. Naim, Aragon, Krell, Acrus, VTL, Cary. I can dream can't I?

I still have not had any offer of a really good amp for the test. Naim, Aragon, Krell, Acrus, VTL, Cary.

Anyone local willing to let you demo one?

Did you perform any measurements of what you listened too?

Just did some basic sweeps on the HK A300 while waiting for breakfast to bake. Pronounced shelf from 600 up. Good old days of LOUDNESS, rather than gain controls. Actually, this puppy seems to work as well as when it was new, some 40 years ago. Just watching the sweep, the second order HD at 10W must have been 10%! My resistors were getting hot, so I quit playing. It is now for sale.

There is enough difference audibly between the "modern" amps that if the pulse test is valid, I should be able to see it. All of modern ones have ruler flat frequency response, so that is out of the way.

Another subjective observation. The applause on Clapton Unplugged and the strings on Bream Plays Segovia "evil" effects were exactly consistent with my wife's trumpet blaring test. So we have three different effects both of us can hear. ( But she is far better at it than I) This would confirm that whatever the issue is, it is the one that most characterizes an amp's sound.

I was re-reading David Manley's old book on why his amps (VTL) were so great. Actually, a lot of valid technical info on tubes which makes having to buy the 200 page advertisement worth while. He talks a lot about the stored energy. ( High B+ of course makes it easier for smaller caps) I was very much suspecting this but the Rotel has no where near the caps as the B&K and a lower rail voltage to boot! Proximity to the outputs still could be.

Time to get scientific.

Comment: Whatever this is, it seems to be a upper mid or higher effect. So all the arguments on which cheap "pro" monster amp to drive your 18" subs with should not be worried about this. I leave that for some one else.