2. tech-talk
  3. 147133-20253

The Carver Amp Challenge and the 21st Century and it's Failure

Some of you may be old enough to remember this article from Stereophile. Bob Carver claimed he could make an amplifier audibly indistinguishable from some of the best from Conrad Johnson. A high efficiency (not class D), solid state linear amp vs. a linear tube amplifier.


https://www.stereophile.com/content/carver-challenge


Carver's approach was to feed a speaker via both amps at the same time using opposite terminals. The speaker itself was the measure of accuracy. Any difference in output between the two amplifiers would cause audible output.


What's super important here is Carver invented a new way to measure the relative difference of amplifiers with a real load.


That's kind of revolutionary from the standpoint of commonly published measurements of amplifiers before. Steady state, frequency sweeps, THD, IM and S/N all failed (to my ears) to express human experience and preference. I remember a reviewer for Audio, I think Julian Hiirsch, who claimed that these primitive measures were enough to tell you what an amplifier sounds like. The man had no ear at all, in my mind.  More here:


https://www.soundandvision.com/content/reconsidering-julian-hirsch

And here was Carver in 1985 cleverly showing that two amplifiers which measured reasonably well, sounded differently. We should also be in awe of Carver's ability to shape the transfer function on the fly. That's pretty remarkable too but not the scope of this post.


My point is, really, Carver showed us a revolutionary way to examine differences between gear in 1985 and yet ... it did  not become widespread.  << insert endless screaming here >>


As far as I know (and that is very little) no manufacturer of any bit of kit or cable took this technique up. We are still stuck in 1985 for specifications, measurements and lack of understanding of what measures cause what effects and end up cycling through cables and amps based on a great deal of uncertainty.


My points, in summary:

  • Most of what we consider state-of-the-art measurements are stuck in the 1970s.
  • There are a number of ways to improve upon them
  • No one has.
  • We should be a little more humble when asserting if it can't be measured it isn't audible because our measurements are not nearly comprehensive
  • I look forward to manufacturers or hobbyists taking modern equipment to pursue new measurement and new insights into our hobby.


Best,
E


erik_squires
rixthetrick's avatar
rixthetrick
1,060 posts
 
Of course it is understood that Bob Carver used a measurement device to tune his amplifier to match the two amplifiers, right?
He didn’t sit down with an oscilloscope, he used the output of the loudspeaker to identify the differences. He did use his ears, not his eyes, and he did use a loudspeaker to do this.

Like for example a waveform that has low distortion however isn’t a match to the original despite not having more or less distortion isn’t going to sound the same. And yes a computer can analyze this difference. This difference can be shown graphically or like the speaker did, be represented audibly.

I am agreeing that despite the sound reproduced by a device having less noise, than another with more noise where the wave form is closer to the original can and often will sound better than the former in resolving as an approximate to the original. This is measurable.

Is it possible that science hasn’t discovered all of the parameters that we can perceive with our hearing? Maybe not, maybe we've been sold on the wrong metrics. Perhaps there’s bigger fish to fry.
atmasphere's avatar
atmasphere
11,849 posts
 
One vast, gaping hole in Bob's idea way back then is the output impedance of the two amps involved in the test. If not identical, the one with the lower output impedance will dominate the results.


And if its not obvious, another serious issue is simply that neither amp will be operating correctly if connected to another amp in a null test! This is the sort of thing you don't do at home, lest ye 'let the smoke out'.  
bigkidz
3,392 posts
 
@atmasphere - exactly plus who cares anyway the end sound is what is important
erik_squires OP
16,200 posts
 
So, I kind of let this thread go in preference to newer threads talking about the same idea, that we’re stuck, but I wanted to take this up:

One vast, gaping hole in Bob’s idea way back then is the output impedance of the two amps involved in the test. If not identical, the one with the lower output impedance will dominate the results.

@atmasphere

Au contraire, mon ami. The clever EE trickery Bob does forced him to alter the output impedance of his amp to match the other. Using the complex impedance of the speaker in between two amps is how it happened.


Amp1 (+) =====> (+) Speaker (-) =====> (+) Amp2


The goal being to null the voltage at the speaker terminals with music, which, as you allude to, could only happen if the impedance of the amps matched.
Now, thinking of this, Carver claimed he manipulated quite a bit, but it may have been all he did was change his amp’s impedance curve. Indeed in future amps, that was the only difference, whether you wanted low or hi Z outputs.  I think the story if not the legend is he was able to change the distortion profiles.  What if the latter part was bunk.   His claim was to match the "transfer function" of the amps.  That means voltage into a given load. I've never seen a transfer function that incorporated distortion.


So, arguably, and with some induction (i.e. guessing) Carver proved distortion didn't matter.  If your distortion is low enough what matters is the output impedance.  Well, I've taught myself something, thank you guys. :)
almarg's avatar
Ag insider logo xs@2x
almarg
9,670 posts
 
Ralph ( @atmasphere ), I believe Erik is correct on this, in the first part of his post just above (but see below re the second part of his post). Note the description of the null test I provided in an earlier post (dated 3-3-2019):

The null test, as described by Aczel, was performed by connecting Carver’s amp and the target amp in the normal manner to speakers that were at inaudible locations, and connecting a monitor speaker and/or a meter at an audible location between the + terminals of the two amps, both amps being provided with identical inputs.

Also, note that the two amps were not connected directly together, but rather through a monitor speaker or a meter, presumably some sort of voltmeter having a wide bandwidth.

As I mentioned in that earlier post, though, two major flaws in Carver’s approach were inability to maintain a deep null in production (which he pretty much admitted many years later, in the TAS interview I referred to), and the fact that the deep null he obtained with his modified prototype was obtained with just one particular speaker load.

I’ve never seen a transfer function that incorporated distortion.
So, arguably, and with some induction (i.e. guessing) Carver proved distortion didn’t matter.

No, I don’t think that is correct, Erik. The transfer function of an amp is the relation between output and input, not just in terms of voltage gain or any other single parameter, but in terms of everything. If the transfer functions of two amps are precisely matched, or in the case of Bob’s prototypes matched to a null of something like 70 db with a musical test signal, then everything about what the amps are doing (gain, distortion of all kinds, noise, etc.) is presumably matched very closely. But as we’ve both indicated, just under the particular test conditions, including the characteristics of the particular speaker load.

Best,
-- Al

More to discover