Did you ever notice: Tubes are CLEANER than transistors!

[Steben]

At least triodes. We are speaking of single-unit non-feedback circuits.
If one looks at the output response graphs, its clear to see that a triode has about 80% almost perfect linear response of its swing (output = x times input, a line). This means a total harmonic distortion of... zero! That's why it is popular in the Power amp Tube HiFi world.

Transistors are logarithmic to te bone, which means there is ALWAYS assymetry and a bunch of second order harmonics. So you always need feedback in HiFi purposes or in other words, complexity and opamps.

So especially a CLEAN Triode Tube amp is therefor superior to solid state, not because of its warm harmonics, no, because its cleaner! When clipped, it happens gently, when not, there is NO distortion.

Its the same reason why transistors (not opamps) are so good at fuzzing, which is in fact making your guitar sound like something else.

[brett]

Maybe I'm missing something.
Don't our transistors mostly operate in a linear region?  At least for our usual class A designs. 
My concept has always been: Bias into linear region, put input into the base, take output from the collector, gain = Rc/Re, THD = almost nothing. 

[Joe]

Look at this chart:
http://www.analogalchemy.diystompboxes.com/pedals/waves.gif

The transistors are only linear before saturation. When saturated the BJT stays on for a while, then switches off abruptly. Yuck! The plain JFET is a little better, but still not great. I think many of the problems with transistors are due to their misapplication. The input signal requires some sort of pre-shaping, so that when modified by the transistor it comes out right. The other stages try to do that, though not perfectly.

[R.G.]

The vacuum triode is the most inherently linear device we've ever made. In its linear region, it gets distortion numbers under 1%, and the distortion it does do sounds relatively benign, and is almost totally harmonic distorion, only tiny amounts of intermodulation. This is due to the gain changing only slowly when you force it out of its linear region *except* when you drive it into grid clipping. I suspect that this is due to the triode being literally a ballistic device, the electrons being thrown on Newtonian paths through the vacuum and let pass by openings in the grid field.

The FET has a similar, albeit slightly less benign characteristic. The basic linear region is a touch more distorted, and being a low impedance device, the distortion region is more abrupt.

The bipolar transistor is a bit less linear yet.

Note that all of these comments are on the raw device, with no local or global feedback. Feedback is the great equalizer. When feedback is used, differences of a few percent of distortion get swallowed up to almost nothing. We have to drive the amp-plus-feedback out of the linear region to deplete the forward gain to hear distortion again, and when it happens, the distortion is the distortion caused by being well into the nonlinear region of the device(s).

It's all about what happens at the edges.

[George Giblet]

>Note that all of these comments are on the raw device, with no local or global feedback.

Yes.   

There's an element of comparing apples to oranges with the raw device - this is mainly directed to the original poster.

The gain of a single stage varies in each case.  Also running circuits at higher voltages tends to give more gain.

If you take a transistor stage and operate it at a relative high voltage the gain is very high.  If you put a resistor in series with the emitter to make the transistor stage have the same gain as the tube I have little doubt it will be far more linear than the tube.

[mac]

My uncle, once violin concertino and elect. technician, used to laugh at those "new" 70's ss hi-fi amps that claimed less than 0.01% distortion. He told me when I was a kid that the 1% distortion of his home-made valve amp was not only acceptable but DESIRABLE. Of course I am talking about audio amps; he never considered a guitar to be an instrument
So, why a device having more harmonic distortion "sounds" better? It is because the way the human ear works is closer to valves than ss.
I believe that any ss stomp box, ie joe's jfet vulcan, will sound better running at 220V than at 9V. I strongly suggested the use of high voltage in a recent topic "attempt at tubish mosfet amp". But it is a lost war, ears love valves. Even worse, maybe our ears will still love valves if they were non-linear!
As R.G. said valves, and ears, are newtonian... well, at least the behaviour of electrons after leaving the cathode is more classical. The thermionic emission is of quantum statistical nature, Richardson-Dushman Law. Once an e- has left the cathode it is influenced by others e-, the gas in the tube and the the electric field, which depends on the shape of the cathode and the anode. Fooling with the cathode and the anode you can have a voltage vs. current plot as linear as you want. This is not possible in a ss diode. They obey to the Ebers-Moll model, which is again of quantum nature. Nothing to tune inside a 2N5088 to make it more linear
Didn't read much about fets. Can the tunnel and/or the drain-source path be shaped to have a more linear response?
Another nice effect of valves is its own pressure-feedback. The speakers generate waves that hit the glass of the valves a few millisecods later. This movement is transmitted to the gas inside the valve changing the pressure and so the temperature which controls the emission. A analog delay & distortion stomp box in just one unit?
The more I think about valves and ss this photo  ( http://geofex.com/FX_images/realtube.jpg ) makes more and more sense to me

mac

http://geocities.com/guitarfxs