[PaulFolbrecht]

I didn't know we were limiting the discussion of feedback to global feedback. I was referring to simple emitter degeneration, which is just a resistor at the transistor's emitter. In other words, it's local.

We weren't limiting the discussion at all.  I was pointing out that, at the least, the very small-scale triode feedback - if it exists - is not exactly directly comparable to the global feedback almost always used to linearize SS circuits - which is what you originally implied unless I misunderstood.

Not sure I follow. The only thing I cited from the paper was their saying that the notion of tubes clipping more softly than transistors was a myth.

So, I read the Hamm paper and now do have better context.  Sorry about that. 

You're correct that the article points this out, but it also points out that the distortion spectra at clipping of tubes is typically much friendlier than transistors.  We could talk about a lot of things in the article, but overall the article rather heavily "favors tubes", as you know.  A quote from the last section:

'Vacuum-tube amplifiers differ from transistor and operational amplifiers because they can be operated in the overload region without adding objectionable distortion. The combination of the slow rising edge and the open harmonic structure of the overload characteristics form an almost ideal sound-recording compressor. Within the 15-20-dB "safe" overload range, the electrical output of the tube amplifier increases by only 2-4 dB, acting like a limiter. However, since the edge is increasing within this range, the subjective loudness remains uncompressed to the ear. This effect causes tube-amplified signals to have a high apparent level which is not indicated on a volume indicator (VU meter). Tubes sound louder and have a better signal-to-noise ratio because of this extra subjective head room that transistor amplifiers do not have.'

But that's neither here nor there.  Now I understand your original point better: that tubes and transistors *as output devices* are more similar than conventional wisdom says, but in practice transistors aren't implemented in ways that display this.  Let me know if that is off the mark.

** Of course.  The point is that, in the real world, in real amplifiers, tubes and transistors have all kinds of important differences - many of them related to the fact that transistors *need* NFB in order to operate anything close to linearly.

Well there's evidence that this is not necessarily the case when the comparisons are more apples to apples.

But I still don't know what this has to do with crossover distortion.

Nothing.  That was a tangent.  I thought for a moment you might be hinting at an "all amplifiers sound the same; a watt is a watt" type of argument, but I am sure you are much brighter than that.

**Of course.  And in the real world, transistors need lots of feedback and tubes don't.  Do you agree with that?

Not necessarily, no. Again, there is evidence that when you run transistors at similar voltages to tubes, they virtually identical distortion characteristics as tubes. And in this comparison, while the overall distortion was a bit higher, it wasn't dramatically so.

Also as I pointed out in a previous post, cascoding can dramatically linearize a transistor (or a tube for that matter) without using any negative feedback.

Once again this is appearing to me as academic.  If transistors can be run this way, and apparently operate better doing so, why does nobody do it?  Why is it not happening in the real world?

I think you are the first person I've ever heard say that, in general, tubes don't require less negative feedback than transistors.  I think just about every staunch proponent of sand amps would acknowledge that. 

I don't quite know enough about cascoding to comment on your last point.  Your technical knowledge certainly surpasses my own in a lot of these areas. 

Anyway, this notion that transistors behave more like tubes (than thought) at higher voltages is truly very interesting, and new to me.  Maybe there is something valuable there.

Ah, ok. Well suffice to say that I think the whole "single-ended characteristic of air" thing isn't really an argument. By the time we would start experiencing the single-ended characteristic of air, our eardrums would be slammed six inches into our skulls.

I'm not prepared to go there (at least not at the moment).  It's been a long time since I've read that article.  I found it very intriguing, if nothing else.

[PaulFolbrecht]

Here are some more thoughts.

I listen to jazz and I re-calibrate my ear regularly with live, unamplified acoustic music, and single-ended tube amps are able to recreate such a live event for me more convincingly than other amplifier topologies.  I really wanted to know what kind of amps I liked best, so I bought over 15 of them in a two-year or so span to hear them in my own home for extended periods, and listened to dozens more as well, of all topologies and technologies.  Those were my conclusions. 

Secondly, I found that what comes closest to SET realism (I say SET but single-ended pentodes and tetrodes are very similar) is not any sort of conventional amplifier, but instead digital one - Tripath at that.  Tripath amps, especially with the extra voodoo the maestro at RWA works, have almost everything they have (and in fact I could live with one happily if I had to).  Yet, these amps have vanishingly small distortion when not overloaded (and a distortion spectra completely unlike SET when overloaded).  The fact that these amps sound a lot like SET and yet are dissimilar in terms of distortion in every way answers the question "Do SETs sound the way they do because of or in spite of their relatively high levels of low-order HD"?  (In any case I couldn't buy that increased presence, a bigger soundstage, and more detail could possibly the result of distortion products.)
 
None of this is exactly new or original, by any means.  This is what my path has been, and it's a path a lot of others have followed as well.  If there is something better (for me) in the future than SE tubes, I'm betting it's Tripath or something like it, not transistors.  Still a very young technology with a lot of maturing to do, I am sure.

[Steve Eddy]

We weren't limiting the discussion at all.  I was pointing out that, at the least, the very small-scale triode feedback - if it exists - is not exactly directly comparable to the global feedback almost always used to linearize SS circuits - which is what you originally implied unless I misunderstood.

I wasn't implying global feedback, I was implying local feedback in the form of simple emitter degeneration which I had mentioned in a couple of previous posts.

So, I read the Hamm paper and now do have better context.  Sorry about that.

No problem.

You're correct that the article points this out, but it also points out that the distortion spectra at clipping of tubes is typically much friendlier than transistors.  We could talk about a lot of things in the article, but overall the article rather heavily "favors tubes", as you know.  A quote from the last section:

'Vacuum-tube amplifiers differ from transistor and operational amplifiers because they can be operated in the overload region without adding objectionable distortion. The combination of the slow rising edge and the open harmonic structure of the overload characteristics form an almost ideal sound-recording compressor. Within the 15-20-dB "safe" overload range, the electrical output of the tube amplifier increases by only 2-4 dB, acting like a limiter. However, since the edge is increasing within this range, the subjective loudness remains uncompressed to the ear. This effect causes tube-amplified signals to have a high apparent level which is not indicated on a volume indicator (VU meter). Tubes sound louder and have a better signal-to-noise ratio because of this extra subjective head room that transistor amplifiers do not have.'

But that's neither here nor there.  Now I understand your original point better: that tubes and transistors *as output devices* are more similar than conventional wisdom says, but in practice transistors aren't implemented in ways that display this.  Let me know if that is off the mark.

Fair 'nuff.

Though I would note that I saw no apples to apples comparisons in the Hamm article either. As I said in a previous post, some of the transistors used to make comparisons were 25 volt devices. And I'm sure they weren't running their tubes at 25 volts.

Nothing.  That was a tangent.  I thought for a moment you might be hinting at an "all amplifiers sound the same; a watt is a watt" type of argument, but I am sure you are much brighter than that.

Oh. Nope. Wasn't hinting at any thing like that.

Once again this is appearing to me as academic.  If transistors can be run this way, and apparently operate better doing so, why does nobody do it?  Why is it not happening in the real world?

Beats me.

I think you are the first person I've ever heard say that, in general, tubes don't require less negative feedback than transistors.  I think just about every staunch proponent of sand amps would acknowledge that.

Dunno what to tell you. I would probably have acknowledged it too if I hadn't seen the evidence supporting it. Which by the way I finally located and will post at the end of this post.

Anyway, this notion that transistors behave more like tubes (than thought) at higher voltages is truly very interesting, and new to me.  Maybe there is something valuable there.

Ya never know.

I'm not prepared to go there (at least not at the moment).  It's been a long time since I've read that article.  I found it very intriguing, if nothing else.

Well, it was intended to intrigue. It's called marketing.

Anyway, here's the piece I was referring to previously but wasn't able to locate until this evening.

http://www.spectrum.ieee.org/archive/1640/distortion

Here's a 6AU6A pentode and a 6SN7GTB triode:



The triode has a good amount of second, a bit of third, and virtually nothing after that. Meanwhile, the pentode shows products all the way out to the ninth.

Here's a small signal 2N2222 bipolar transistor and a small signal 2N5457 JFET both run at 20 volts:



The distortion characteristics of the 2N2222 are remarkably similar to that of the pentode, again with products all the way out to the ninth. The JFET's a bit better than the bipolar up above the sixth.

Now here's where it gets interesting. An MJE2362 bipolar transistor and an IRF822 MOSFET transistor, both run at 180 volts, same as the triode and pentodes were:



The distortion characteristics of both devices are now remarkably similar to that of the triode. A good amount of second, a bit of third, and nothing after that.

Admittedly the noise floor's rather higher here because the transistors have a lot more voltage gain than the tubes (the measurements were made with the input voltage adjusted to give 2 volts output from all of the circuits). However just look how much lower the third is compared to the second here versus the results of the 20 volt transistor circuits.

se