System distortion and Single Ended Tube Amps

[Gerrit Boers]

Some time back I came across a paper by Eduardo de Lima on the subject of harmonic distortion cancellation. The idea was that harmonic distortion components of single ended tube amps and loudspeakers can cancel each other out.
I have been able to confirm by measurements that the harmonic distortion for my single ended tube amp is lower when loaded with a loudspeaker then when loaded with a 8.2ohm resistor.

The results can be found on my System distortion page.

An example:
1173Hz sine wave:
SPL @1m on axis: 111.2 dB
THD @1m on axis: 1.366%
THD on amp while loaded with LS: 1.449%
THD on amp while loaded with 8.2ohm resistor: 2.366%

I also compared THD and IMD in room @1m for my tube amp and a NAD C 320. I used 4 two-tone pairs at 2 sound pressure levels. This yields 16 measurements. The score was 11-5 for the home team.

You are all invited to replicate/repudiate(with measurements please) these results.
If you can find any flaws in my procedure please let me know.

[AKN]

Hi Gerrit Boers,

If you arrange so that output impedance on both amplifiers being equal, what result will you get then?

[SY]

Thanks for posting this. There was obviously a lot of hard work here, and I'm delighted when people stop speculating and go and actually measure things.

My I have the temerity to make a couple of comments? It might be helpful to dig into the repeatability and reliability of the measurement. Here's something to try: do the same measurement you did on your web page, but repeat the measurement pair, i.e., instead of AB, do ABAB with full connecting-disconnecting, and see how well the repeat measurement corresponds with the first pair. It may be that everything is fine and consistent, which you will now have data to support. But it may be that there's a systematic error (order, for example) that needs to be tracked down.

The IMD chart is interesting. You state that the levels of the two tones are the same, yet the graph shows them as different.

The data need to account for differences in source impedance. A series resistor on the output of the NAD can do a great deal to equalize things. Mills and Hawksford have shown some interesting results for distortion with current source drive, and to a lesser extent, that could be the case with the high output Z amp you used here. Well, if output Z reduces this speaker's distortion, we have a simple and cheap mod to the NAD. (edit: I see that 4fun said the same thing while I was typing)

The last thing I'd caution about is something I think you already have an idea about- frequency choice and binning.

[arnoldc]

Thank you very much for opening up a discussion on this topic. While I have downloaded that and converted to PDF for easy reading, I have not really immersed myself into it due to my rather pathetic technical capability.

This thread could really enlighten the likes of me about the subject.

I'm not sure if related, but I've seen a "distortion cancellation" circuit from one of Mr. Borbely's creation or is this a totally different approach?

[jon_010101]

Very interesting!

I am more curious about the presence of the 5th and 7th harmonics in-room with the SS amp, where they are unmeasurable at the amplifier's output (is that the output while driving the speaker???). Those high-order components are not visible in the SET, suggesting they aren't intrinsic to the driver. The reduction of 2nd harmonic distortion in the room via cancellation makes sense to me -- but these 5 and 7 components in the SS plot suggests that attempts to reduce distortion at the speaker terminals via feedback may produce more (and higher order) nonlinear distortion from the driver into the room.

link

[Gerrit Boers]

If you arrange so that output impedance on both amplifiers being equal, what result will you get then?

Good question I'll have to try that.

It might be helpful to dig into the repeatability and reliability

Definitely, it was my intention to repeat the series at different levels. Your remark on frequency choice and binning (the frequencies are FFT bin adjusted) is also spot on.
The level differences for the two-tone is the result of frequency response deviations of my system+room. I will add the pink noise measurement which will show the peaks and dips. It will also explain the choice of frequencies for the two-tones.

I'm not sure if related, but I've seen a "distortion cancellation" circuit from one of Mr. Borbely's creation or is this a totally different approach?

No, they're not related. This is something that 'just happens' if you connect a single ended tube amp to a loudspeaker. Note: I used a single ended amp for these series, I do not know if this is a SET thing or just a result of the high output impedance of tube amps in general.

I am more curious about the presence of the 5th and 7th harmonics in-room with the SS amp

You should not put too much weight in the first measurement, this particular result could not be reproduced.
Also keep in mind that the Fostex has some nasty whizzer peaks at 10KHz (working on that) which will probably skew the results.


The most striking result was the difference in distortion for the tube amp loaded with a resistive load and loaded with the loudspeaker. This reminds me of the 'iron in the spinach' story (the iron content was determined by chemical analisys, but Fe3+ is not absorbed by the body). Same thing here, we measure the distortion of a tube amp with a resistive load and get high levels of distortion. However, these high levels of distortion are not to be found in the room.

I also still have my trusty Fisher KX90 integrated tube amp (EL84 PP), will have to try that one to.

[arnoldc]

That's a really interesting result.

Will the quality of the speaker determine the distortion results?

[Gerrit Boers]

Will the quality of the speaker determine the distortion results?

It certainly will affect the in-room distortion. In how much the loudspeaker affects the differences between amplifiers I wouldn't know, this is the subject of another investigation.
If you compare my results to those of others, pay special attention to the sound pressure level and mike distance. As an example, the measurements on Zaph's site were made at 50cm with a level for the fundamental of 92dBSPL.

One thing I will certainly investigate is the influence of a passive crossover. I'm contemplating the idea of adding a Raal ribbon to the system but first I want to be sure I do not loose the benificial effect on tube amp distortion by adding a crossover.

[jon_010101]

Originally posted by Gerrit Boers

You should not put too much weight in the first measurement, this particular result could not be reproduced.
Also keep in mind that the Fostex has some nasty whizzer peaks at 10KHz (working on that) which will probably skew the results.

The most striking result was the difference in distortion for the tube amp loaded with a resistive load and loaded with the loudspeaker.

I'm glad we can ignore that first measurement -- increased high-order distortion would be a troubling finding.

So, that aside, it does appear that the 2nd harmonic distortion of the tube amp may cancel slightly with the 2nd HD of the driver. The total distortion of the Fostex driver is, of course, much higher than either amp. I'll be curious to see how the Fisher performs here. In any case -- very interesting!

[Jeb-D.]

Excellent work!
I've always had my suspicions regarding this matter. "why does my amp that benches at 2%THD sound cleaner with these speakers than my amp with .01%THD?". But without an acoustical measurement setup there wasn't really any way to prove anything.

However I did build a line stage that would intentionally inject a variable ammount of 2nd harmonic. While listening I'd tweak it, there would be an noticable spot where the sound started to clear. Going with too little 2nd harmonic sounded ok, and going over would sound progressively worse. The sweet setting seemed to vary with betweens speakers, which makes sense.