Phase splitters: Constant current sources

[BlackUnikorn]

Anyone experimented with a constant current source at the cathodes of a phase splitting differential amp? Its seems to be a much better alternative to using resistors, providing better balance.

[Circlotron]

I made a similar comment some time ago about using a single unbypassed CC source from the two cathodes to earth of a p/p class A output stage. Sounds like a good idea; haven't tried it though.

[Brett]

Originally posted by BlackUnikorn
Anyone experimented with a constant current source at the cathodes of a phase splitting differential amp? Its seems to be a much better alternative to using resistors, providing better balance.

Yes. Works much better than a resistor.

[James D.]

I completely agree with Brett, it works much better than a resistor.

However there are ways of making the resistor approach work better than others. Here is a reference for making it work well Aiken Amps LTP design .

hope this helps

James

[stigla]

As far as I understand it (from reading Morgan Jones's book), is that a pure differential stage with a CCS on the common cathodes (and one grid grounded) is not the perfect phasesplitter.

It comes from the fact that this type of phasesplitters need a very well balanced load, whitch is a fact with a Class A output stage, but with a class AB outputstage, you've got another case. When one of the tubes closes cut-off and the other is going conducting-city, the miller-capacitance will change on the tubes i.e. making the loads unbalanced. (at least the way I understands it)

This can be cured though, by isolating the phase splitter stage f.ex. by another differential stage between the ph.splitter and the output stage, or by cascoding the ph.splitter. (as in allen wright's PP1).

this is just my understanding of it, though

stigla

[richwalters]

Yes, despite doing many mods and botch up to long tail followers I could never get them to sound right, despite what I thought were ideal tubes and configurations for it. The Radford ST100 (100W amp) uses EF184 pentodes which I've tried but always found poor harmonic performance between with the two outputs. Yet so many designs use it with high value resistors that it does "nothing" to quench Miller effect. .

The "concertina" recommended by RCA with the 7199 tube application note is the best configuration I've tried that gives symmetry, sound and simplicity. The 3 's' .
Anyone else cobbled it in an amp ??

rich

[PRR]

> The "concertina" recommended by RCA with the 7199 tube application note is the best configuration I've tried that gives symmetry, sound and simplicity. The 3 's'. Anyone else cobbled it in an amp ??

Who hasn't?

What 7199? The split-load (or concertina) goes back to 2A3 days. Langford-Smith argues for it in Radiotron Third. When I recently went back to P-P triodes but threw a computer at the math, I decided that for a given supply voltage, no diff-pair could make as much grid-swing, as-clean, as a split-load. That seems absurd, but the 100% local feedback really-really helps linearize the tube.

There are objections. Output impedance isn't matched. That does not seem to do any harm, just looks bad on the dual-trace scope. If the grid driven by the cathode goes into grid-current, the other grid is slammed negative-- for this and other reasons you want grid-stoppers in the 10K-47K range.

[SY]

Other objections: lousy gain. Limited swing.

A diff amp with time constants and operating points wisely chosen, along with a tail CS or resistor returned to a good, solid, negative rail, can work very well with the right tube- or with a cascode. That said, I'm partial to crosscoupling or the SYclotron for balance, gain, voltage swing, and good overload characteristics.

[PRR]

> Other objections: lousy gain. Limited swing.

Come on, Sy, you know better.

The diff-pair needs two tubes, so compare the concertina plus a plain-amp.

And for example, say 250V B+, same for both.

The well-designed concertina will swing 40V peak per grid at 1% THD. With a medium-high Mu driver, the total gain from one input to one output will be roughly 0.8*0.9*Mu or around 40 for a 12AT7.

That same 12AT7 as a diff-pair will give a gain from one input to one output will be roughly 0.6*Mu/2 or around 18. (Gain is less in part because if the diff-connection, but also because the gain stage also drives a load, while the cocertina's input impedance is about infinity so the previous stage gets highest gain.) Assuming you do have a negative supply or a low-loss current source, the diff-pair can make about 50V peak to each grid, but at THD approaching 5% each-side (there will be some cancellation if you take the output push-pull, i.e. assuming your output stage never goes into cut-off).

Yes, you can do worse. A 12AX7 concertina driving fixed-bias outputs usually isn't a happy critter, because the usual fix-bias 100K grid resistor is a heavy load on the 12AX7's space resistance. 6C5 is a real good choice; 6SN7 or 12AU7 is a modern approximation.

The diff-pair sounds different, and overloads different. That's why Fender used it in their big bottle amps, in a clever self-bias affair that eliminates a negative supply (even though they had one for bias). But especially with a significant cathode voltage, the swing out of a diff-pair is too small for triodes and low-gain pentodes, you need to be up in the 6L6 class.

The split-load concertina does have low PSRR. That was about my first audio trouble-shoot: big old 15W ArKay was motorboating. After pondering, my dad told me to check the split-load resistors. Supposed to be 47K; one was 52K and the other was 70K and drifting before my eyes. Ran too hot too long. Got a ride to Lafayette, couple 47K 2W resistors, worked real sweet.

[fdegrove]

Hi,

Reading the first thread it does say "phase splitting service" , not necessarily "phase splitter" as in PP service.

Not quite the same.

Cheers,