Long Tail vs Paraphase?

[loudthud]

The feedback to the bottom of the tail resistor makes the tail look more like a constant current source until the amp clips. In two bass amps I built, I connected the tail resisitor (30K or 33K) to the -60V bias supply. That allowed the grids of the LTPI to be at ground level. I can't say what it sounds like for a guitar amp.

[Gingertube]

I use cascode bipolar transistor current source (about the best CCS you can get) tail in my HiFi desgns. I think that for guitar the arguement would come back to how balanced do you want it. Perfect balance (in a differential amp) tends to cancel even order harmonics whilst leaving the odds. Usually we want enough imbalance such that the even order harmonics balance the remaining odd order harmonics, that is perfect balacnce is not what we want.
There may well be other ways to get this even / odd harmonic balance right and I'm experimenting with a HiFi variation now.

Cheers,
Ian

[ReadyTeddy]

That's true.

I get nervous about using ICs in the signal path of a tube amp when they're DC connected. I think the law of diminishing returns applies here - you could get most of the value of a high performance CCS by using a simple CCS instead of a resistance tail.

Something as simple as a MOSFET, a zener, and a few resistors would do a good-enough CCS and be more voltage-tolerant.

It's not hard to get a CCS with more internal impedance than the resistor string of a typical LTP in a tube amp.

Diffamps do have smoother entry into clipping, even with bipolars. It's worth experimenting with.

I'd have to think about what happens to the feedback path on amps using the cathode string as a target for output feedback.

The LM334 performs about as well as a FET current source, but is cheaper (in general), uses fewer components than some FET configurations, and requires a smaller burden voltage. It will withstand 40 volts, which is way more than is needed. Here is a very nice article comparing current sources, http://waltjung.org/PDFs/Sources_101_P1.pdf

[ReadyTeddy]

I use cascode bipolar transistor current source (about the best CCS you can get) tail in my HiFi desgns. I think that for guitar the arguement would come back to how balanced do you want it. Perfect balance (in a differential amp) tends to cancel even order harmonics whilst leaving the odds. Usually we want enough imbalance such that the even order harmonics balance the remaining odd order harmonics, that is perfect balacnce is not what we want.
There may well be other ways to get this even / odd harmonic balance right and I'm experimenting with a HiFi variation now.

Cheers,
Ian

Of course, you can get any degree of balance you want out of such an LTP by adjusting the relative values of the plate resistors.

[R.G.]

Of course, you can get any degree of balance you want out of such an LTP by adjusting the relative values of the plate resistors.

So long as the plate resistors are significantly lower impedance than the grid leak resistors and grid impedance on the output tubes, yes.

Long tail pairs driven non-differentially, on only one grid, push signal into the second tube section by cathode follower action pushing the second cathode up and down while the second grid is held constant(ish). The lower the impedance on the joined cathodes, the lower the effectiveness of this action. So the second tube can't have as high an output signal on its plate with equal plate resistors. offsetting the plate resistors fixes this somewhat; it's less effective on triodes than on things with higher output-node impedances.

Here's another random thought. As mentioned, perfect balance may or may not be desired, just as matched AC gains on the output tubes may or may not be desirable. Dinking with the plate resistors can tune in - or out! - the gain of one tube versus another to get good AC balance in the OT, or just the right degree of mismatch desired, in spite of how well matched, or un-matched the output tubes are.

And finally, getting a PI that always produces equal-and-opposite signals to the output tube grids is made less useful by the non-matched-ness of output tubes, as perfection is invalidated by tube variation. And a fixed mismatch on the PI output signals may work better or worse by the output tubes and their drift.