Hey, I'm looking for some of that "tube sound"!

[Saurav]

Let's assume that "tube sound" is defined as just being the distortion spectrum of the output, without looking at impedance or any other factors. Does this depend on the topology of the tube stage? For instance, if I have a 12AU7 set up as a common cathode amplifier, and then use the same 12AU7 in a cathode follower stage. The tube is biased to the same operating point in both cases (that's doable, right?). The heater supplies are the same, everything else is the same (this is a hypothetical situation, after all). Let's also assume that the input impedance of the destination/load device is high enough that both the common cathode and cathode follower can drive it with negligible degradation (I think that's a reasonable hypothetical assumption to make too).

If someone ran a distortion analysis on these two setups, would they behave differently?

Where this question comes from: There are people who mod CD and SACD players to put tubes in their analog stages. Some people only replace the final buffer stage opamp with a tube, others replace the voltage amplifier stage too. If we restrict the comparison to just distortion spectra, and ignore the fact that the second option would remove more opamps from the signal chain (i.e., assume that the voltage amplifier opamp adds no distortion of its own), is there a difference between the two? Does a tube add more "tube sound" as a voltage amplifier than as a cathode follower/buffer?

When used as output power devices, I know things change, because then you have the transformer, you have the interaction of the tube with the reactive speaker load, you have damping factor issues, and so on. But if we restrict the comparison to 'line level stages', and compare only distortion, how much effect does the stage's topology have on distortion?

Thanks, and I hope this isn't a really dumb question.

Saurav

[SY]

If someone ran a distortion analysis on these two setups, would they be different?

Yes. You've got two very different gains and two very different feedback levels. This is one more reason you have to be suspicious when someone talks about a device's "inherent" linearity.

Not a dumb question at all.

[Saurav]

But that doesn't mean that there is no such thing as a tube's inherent linearity, does it? A tube's plate curves don't care what topology you use them in. But I see what you're saying, the topology affects more than just the operating point.

I have a vague understanding of how/why feedback affects distortion. How does gain affect distortion?

[SY]

Well, gain is what you trade off for distortion. As you increase negative feedback, gain decreases in the same proportion as the distortion (a simplification, but 99% true).

But that doesn't mean that there is no such thing as a tube's inherent linearity, does it?

Well, yeah, it does. let's take an easier case, a bipolar transistor versus a tube. If you run the transistor common emitter, the voltage transfer function (VTF), in the absence of degeneration or other feedback, is exponential- highly nonlinear. Take a tube and do the same thing and its VTF will be MUCH more linear. Now, put the transistor in common collector configuration and the tube in the analogous common anode, and suddenly the transistor's VTF is MUCH more linear.

So which device has better "intrinsic linearity"?

[Saurav]

Well, gain is what you trade off for distortion.

Duh. Of course. Thanks.

As you increase negative feedback, gain decreases in the same proportion as the distortion (a simplification, but 99% true).

But the distribution of that distortion across the different harmonics also changes, right?

So which device has better "intrinsic linearity"?

I see what you're saying. So when we look at plate curves and try to pick a linear operating point... do the plate curves only apply for common cathode stages then? Actually, I think I know the answer to that, because the gain and output impedance predicted by the plate curves do not apply in a cathode follower topology, which means that the plate curves aren't valid any more, and so neither are their linearity predictions.

[fdegrove]

Hi,

A tube's plate curves don't care what topology you use them in.

Plate curves are only valid under the conditions they were made as SY points out.

A much more interesting approach to the use of tubes is to trace the actual topology and see how that fares on a plotter.

To give you an example, you can build an outputstage using SEPP as a topology with low µ triodes and I can assure you that for a good range of operating points it's going to be so linear you'd hardly believe it...

When you actually think it through, this is to be expected.

Actually, I think I know the answer to that, because the gain and output impedance predicted by the plate curves do not apply in a cathode follower topology, which means that the plate curves aren't valid any more, and so neither are their linearity predictions.

First, you can hardly predict Zo of a tube just by looking at plate curves....Actually I can, but that's more to do with experience and knowing thousands of tubes' data by heart.

Secondly, a CF is based on 100% local FB which is one way to achieve even higher linearity and lower Zo by the same token...but what happened to the gain?
Vanished into the FB loop...no free lunches here either I'm afraid.

As for the sound of tubes, well yes, this is another one of those myths...I can assure you that you can build tubed stages that distort as nastily as any half decent transistor with pretty much the same harmonic content.

Making transistors sound or distort like tubes is only possible to some extent, fact is a tube will distort in a gentler manner ( to the human ear) than a transistor when driven into clipping, which is one reason why people often think, just think, that tube amps are more powerful than transistor amps.

That is yet another myth that can't possibly hold water, can it?
It simply can't.

Cheers,

[SY]

A much more interesting approach to the use of tubes is to trace the actual topology and see how that fares on a plotter.

Truer words never spoken.

But the distribution of that distortion across the different harmonics also changes, right?

Yes, distribution AND magnitude. There's a terrific and entertaining discussion of that going on in the Solid State forum area. A lot of interesting simulations, but I'd sure like to see some real measurements as Frank suggests.

[Saurav]

First, you can hardly predict Zo of a tube just by looking at plate curves

Right. I was thinking of Rp, which is just one of the things that determines Zo.

To give you an example, you can build an outputstage using SEPP as a topology with low µ triodes and I can assure you that for a good range of operating points it's going to be so linear you'd hardly believe it...

Oh, but that won't be a SET, so how can it possibly sound good?

which is one reason why people often think, just think, that tube amps are more powerful than transistor amps.

The 1 tube W = 2 SS W thing? I've seen that one before.

There's a terrific and entertaining discussion of that going on in the Solid State forum area.

8 pages long, the subject line is exactly what I'm asking about, and I hadn't noticed it. I guess I should visit the other forums on this board too, every now and then.

[fdegrove]

Hi,

Oh, but that won't be a SET, so how can it possibly sound good?

The only half decent one I ever heard costs close 50.000$, put out 8W into equally expensive and dreadfully coloured Lowther speakers and was still nowhere near to what can be achieved regardless of that pricetag.

I'll skip the details on the cost of the rest of that system but if you, or anyone else, think that designing a fine* SET is simple, I can tell you upfront it's not just as "simple" as building an excellent PP stage, quite to the contrary.

The 1 tube W = 2 SS W thing? I've seen that one before.

A mere subjective expression, not a real one.

*Let me restate that: amongst the finest amps in the world.

Cheers,

[Saurav]

My attempts at sarcasm have been falling a little short lately I'm aware of the "anything SET must be magic" and "anything magic must be SET" myths. I'm also aware of the physics bending 1W of type X = 2W of type Y myths.

but if you, or anyone else, think that designing a fine* SET is simple, I can tell you upfront it's not just as "simple" as building an excellent PP stage, quite to the contrary

What if we change the conditions. Let's say I'm not trying to build a world class amp, I'm just trying to build the best amp I can. For small budgets and limited skills, is it easier to design a halfway decent amp by going SET, or PP? Or is that even a meaningful question. Which one has more things that a newbie designer could get wrong?

I decided to build a DHT SET for my first amp mostly because I just wanted to see what the hype was all about, and it looked like it would be simpler and cheaper to do. I'll probably try PP next. Many people who's opinion I respect have had favorable things to say about transformer coupled PP DHT amps, to use one specific example. But that's going to be a lot more than the $500 I spent on my current amp.