Question re: Distortion in triode gain stage

[rongon]

I noticed something in the discussion on distortion in medium mu triodes in Morgan Jones' Valve Amplifiers 3rd Ed.

There's that part where he compares 6SN7 and its cousins against 12AU7/ECC82, E182CC (7119), and 5687.

The test circuit has a triode wired D3a as an active load (mu follower) for the lower triode under test. The plate current for all the tubes under test is set to 8mA.

Now, I understand that 8mA plate current is a nearly ideal condition for getting the most out of a 6SN7 or similar triode. I noticed that on the graphs for 6SN7 ra vs. Ia, 8mA puts the tube in a place where its ra remains constant through much of the output swing.

When you look at a 5687 with Ia = 8mA, the picture is not so rosy. The 5687 never has as constant an ra vs. Ia line as 6SN7 does, but it is much better at 15 to 20mA than it is at 8mA. The same holds for E182CC.

Isn't it unfair to compare 5687 at Ia = 8mA to 6SN7 at Ia = 8mA? Wouldn't it be a more fair comparison if Ia = 15mA for 5687 and Ia = 8mA for 6SN7?

--

[SY]

You can get comparable results to his by loading the tube with a CCS, then measuring through a buffer or high impedance input; that's what I did when I repeated and extended slightly his 6SN7 survey. I'd be interested in what you find with the 5687.

[rongon]

Well, I lack the gear, and probably the skill, to make comparable measurements.

The basic question is: Will a 5687 with a CCS plate load, or as a mu follower, and 15mA Ia, have comparable distortion to a 6SN7 in a similar circuit, but with 8mA Ia?

I suspect the 6SN7 will still have lower distortion (its ra curve looks impressively flat from about 7.5mA and up), but if you look at the 5687 data sheet, the 5687 looks like it's in a pretty bad spot with only 8mA of plate current.

Here are a couple of pics to illustrate...

To me, it looks like 5687 doesn't really come into its own until it has 18mA of plate current.

[SY]

Remember that this circuit is running constant current, so variations in rp with current are second order issues. The main issue is variation in mu with plate voltage.

[rongon]

I always learn something new from you (and many others)!

Ah, that is very interesting...

From those charts, the 6SN7 with 150V on the plate and at least 9 or so mA of plate current has an impressively flat curve for mu.

The 5687 mu curve is always ever so gently sloping, from lower currents to high.

Mu for 6SN7 with Va of only 50V is not so constant at all. Actually goes down with higher current. 5687 mu curve stays pretty flat all the way down into cutoff. Is that due to 5687's "high perveance"?

Is a very flat mu curve in the above charts predictive of low distortion performance?

Also... When using a resistor as the plate load, does the variability of rp become more important? Or as important as variability of mu?

--

[SY]

Well, again, that's mu versus current, not mu versus voltage- you're running constant current so the former doesn't matter. At the current you choose, you can pick off the points at different voltages, but for most datasheets, it will be three points spaced very far apart in voltage (like 50, 150, 250V).

[rongon]

I guess what you're saying is that running the tube with a CCS in the plate makes it run in a linear fashion at a wide variety of plate currents, because mu will be fairly constant. So the voltage can do what it will and the mu remains fairly constant, as the current is forced to be constant.

Therefore, the test is a valid one and the 5687 at plate current of 8mA should be as linear as it is at plate current of 16mA, so long as it is loaded with a CCS (and not called on to swing outside of its linear window of operation).

Right?

--

[SY]

A tube will generally show different distortion at different currents, even with a CCS. But... at any given current, the distortion will be minimized FOR THAT CURRENT with a CCS load.

Pick your current, draw a vertical line. It will intercept the mu curves at a few different voltages (depending on how kind the datasheet writers were). The closer those mus are to one another, the lower the distortion.

[rongon]

Pick your current, draw a vertical line. It will intercept the mu curves at a few different voltages (depending on how kind the datasheet writers were). The closer those mus are to one another, the lower the distortion.

Hmmm....

So if the mu lines are very close to each other, you're looking at a very linear spot to run a tube?

I've marked up the charts I uploaded before, but now with red lines showing what you were saying there (I hope).

If I got it right, a 6SN7 running really hot at Ip = 15mA will be extremely linear, because its mu will be almost exactly the same at Vp = 150V as it is at Vp = 250V.

Looking at the 5687, its mu is always different at different plate voltages, no matter which plate current you choose.

The 5687's difference between mu from one plate voltage to another stays constant no matter what plate current you choose.

However, the more current you put through a 6SN7, the less difference in mu at different plate voltages. So the 6SN7 gets more linear the more current you put through it.

No?

--

[Michael Koster]

"Isn't it unfair to compare 5687 at Ia = 8mA to 6SN7 at Ia = 8mA? Wouldn't it be a more fair comparison if Ia = 15mA for 5687 and Ia = 8mA for 6SN7? "

"A tube will generally show different distortion at different currents, even with a CCS. "

Right?