[KurtW]

At the recent bay area meet I had a chance to measure the distortion of a bunch of amps. For all amps I measured THD at 1v output at 1 kHz into both a 30 ohm and 100 ohm resistive load, and many amps I also recorded the levels of the 2nd, 3rd, 4th and 5th harmonic. All of the solid state amps had higher distortion with the 30 ohm load than the 100 ohm load. Here are the measurements for the 30 ohm load for various amps, but keep in mind my lower measurement limit is around 0.03%:

Sugden HeadMaster, 0.06%
HeadRoom Max, 0.03%
Grado RA-1, 0.03%
Berning MicroZOTL, 0.18% (tube amp)
Meier-Audio Corda HA-1, 0.03%
Antness Gilmore, 0.04%
META42 #1, 0.20%
META42 #2, 0.19%
META42 #3, 0.08%
META42 #4a, 0.03% (4x EL2002 per ch.)
META42 #4b, 0.03% (1x EL2008)
META42 #4c, 0.07% (1x HA5002)
META42 #4d, 0.03% (2x BUF634)
META42 #4e, 0.11% (2x EL2001)

The first three META42s were built by three different people, but I believe all three used 2 EL2001 per channel (I know for sure that at least two of them did). META42 #4 was measured with different output buffers, but everything else was unchanged. There is certainly a wide variation in the various META42 amps, much more than any of the commerial solid state amps.

I can say that in some cases amps with lower distortion didn't sound as good as ones with higher distortion, and obviously you wouldn't judge an amp just on one criteria. It is also obvious from the data that there can be a lot of variation from META42 to the next.

 

[morsel]

Thanks for doing those tests, Kurt.

 

[jamont]

Thanks for this very interesting data! Do you have any speculation about why there is so much variability in the META42 samples? Power supply differences?

 

[Voodoochile]

Very enlightening information once again, Kurt.

Thanks so much for your truly valuable contributions here!

 

[morsel]

There are many reasons why there is so much variation with the META42. Number of buffers, choice of buffers, resistor values, opamp, power supply, capacitors, etc.

This is a reminder that the META42 is a board, not an amp. The amp is what an individual makes of it. It was designed to be flexible to give lots of choices to the builder. Also note that higher distortion readings don't automatically mean you won't like the sound as much, although all things being equal, lower distortion is better.

It's too bad we didn't have distortion meters last spring. If we did we could have given more specific advice on choosing components, especially on buffers and resistor values which Kurt has shown can drastically change THD measurements.

Not wanting to put words into his mouth or make promises on his behalf, but I speculate that Tangent will eventually get a distortion meter and update his recommendations. In the mean time, if you have read Kurt's posts you know that local loop gain needs to be higher for best results.

Kurt's distortion meter is a nifty device, partly because it is portable and lends itself well to field use (read: headphone amp gatherings), but it is expensive. A PC with a decent sound card and appropriate software can do the same thing. I predict DIYers, especially the custom amp builders, will soon be using this sort of setup regularly.

 

[earthling]

It would be nice if we could gather up some additional info on the meta42s that were tested. Component choices, choices concering optional part population, power supplies, etc.. Some of the variables which help to dictate the overall sound or measurements may have little or everthing to do with the final results.

Even with basic info about the tested units we could probably draw some reasonable conclusions as to how and why the variances exist.

 

[aos]

Sound card + software works very well, especially concerning component values. I changed inner feedback loop resistors as kurtw's findings suggested and got a nice big 2X reduction in distortion. So even precision of measurement put aside when using soundcard, it can still be useful - just as even an imprecise multimeter is good enough for matching resistors, so is the soundcard in comparing two configurations.

Local loop gain findings however need to be taken with caution because they're likely extremely dependent on the opamp choice (AD8620).

 

[tangent]

Quote:

It would be nice if we could gather up some additional info on the meta42s that were tested.

I agree, but since most people use Vishay-Dale RN55 resistors in META42s (all the DIYFSE-ers, at least) you can't just read the resistor body to find out the values. There's a 1 in 4 chance the value label on the resistor isn't visible for most of the resistors, and a 1 in 2 chance for the others. And, you can't exactly go desoldering someone else's amp to measure them with a multimeter, now can you?

I also doubt that the individual builders kept data on the particular resistor configuration for all the amps they've built. I know I don't.

No, what we need to do is just start experimenting with "META42 breadboards" to start formulating some rules of thumb that will guide future builders.

Quote:

Local loop gain findings however need to be taken with caution because they're likely extremely dependent on the opamp choice (AD8620).

Indeed. You wouldn't want to use an inner loop gain of 200 with an OPA2132, because that's probably high enough that you can do bad things to upper frequencies. (Theory says you can't get more than 40 kHz out of this setup, and that's probably conservative.) As always in engineering, there are no pat answers, only choices.

 

[antness]

Tangent, you mean on your amps you don't bend the leads so the value label is always perfectly visible?

 

[tangent]

Are you saying that you do?

 

[KurtW]

antness, I do that too. If nothing else if helps in debugging if for some reason it doesn't work when its first powered up. I swapped a R1 and R2 resistor once, and it was easy to find because of this.

tangent, I agree about the rules of thumb guide. Change one resistor per channel and the distortion drops by a factor of two or more. This is the downside of having such a flexable design where you can put in whatever op amps, buffers, and resistor values you want...potentially that's very good as it allows for a lot of experimentation, but most people don't bother to do it. That's one big difference between the commercial amps from companies with a R&D budget and the typical DIYer. Your website has provided some guidelines but it's clear there is more work to be done in order to optimise a design. That task doesn't need to and shouldn't fall on one person. And people shouldn't complain about the big companies ripping people off because they charge more for an amp than a DIYer does. When people buy something from a DIYer they are getting a lot of R&D for free, but it may well be less R&D than they might get from the commerial outfits.

I think a lot of the basic information about what's happening here is understood. One problem is how to apply that to the many different variations of the design that are used. Maybe the thing to do is derive and publish one proven high end design with all of the details. People can do varaitons on that and publish their results; they may be cheaper, lower power, or sonically better. Then people can modify their designs to meet their own objectives. This is basically what was done for the META42, but I think we can see now that there is a need for even more details to be thrashed out.

 

[eric343]

Quote:

Originally posted by antness Tangent, you mean on your amps you don't bend the leads so the value label is always perfectly visible?

*raises hand*

I do, for the same reasons as KurtW, and I also like little anal-retentive neatnesses like that...

 

[earthling]

Quote:

Originally posted by tangent I agree, but since most people use Vishay-Dale RN55 resistors in META42s (all the DIYFSE-ers, at least) you can't just read the resistor body to find out the values. There's a 1 in 4 chance the value label on the resistor isn't visible for most of the resistors, and a 1 in 2 chance for the others. And, you can't exactly go desoldering someone else's amp to measure them with a multimeter, now can you? I also doubt that the individual builders kept data on the particular resistor configuration for all the amps they've built. I know I don't.

But I bet a lot of those can tell you that they "followed info from x thread or from your website, or.... " Even if they can't measure their own values they can point at *something* that they used as a guideline. Also, there are a lot of other just gross variables. Did they populate certain positions, did they go for a certain size or brand of cap for the power supply. I would *think* that even these clues could add up to something useful. I like KurtW's idea of publishing one proven design, perhaps one geared towards Grados and one towards Senn 580/600s.

Quote:

No, what we need to do is just start experimenting with "META42 breadboards" to start formulating some rules of thumb that will guide future builders."

I think thats part of the problem. There are some rules of thumb already for those willing to do a little reading but I would bet that a LOT of the people that have built this kit have no idea what to do even with the rules of thumb.

If you (I mean anyone) wants to approach the consistancy of the commercial amps then there needs to be a 'golden' design or designs. Anyone that wants to experiment beyond that can probably figure out everything else for themselves.

 

[goon-heaven]

quote:
--------------------------------------------------------------------------------
Originally posted by antness
Tangent, you mean on your amps you don't bend the leads so the value label is always perfectly visible?
--------------------------------------------------------------------------------

Well then you must send out your rejects. I recall being slightly annoyed that a few resistors in the Gilmore kit arrived pre-bent (and cut down ready for fitting) so that the label would have been obscured.

 

[aos]

>Tangent, you mean on your amps you don't bend the leads so >the value label is always perfectly visible?

Me too

.

But then, I haven't been building in volume.

Also, distortion for sure isn't the answer to everything, as all the audiophiles already know (there wouldn't be so much voodoo if it were). As I mentioned, I have LMH6718 that doesn't flinch distortion-wise when you attach headphones (stays at 0.004% or so) but the sound is nowhere near EL2001 transparency.