[John L Stewart]

In a previous post John Byrnes asked why I used a resister rather than an MJE340 for the long tail in a phase splitter. The short answer is that it works & very well at that.

The question has gotta be, ‘why does it work so well’? For success the circuit surrounding the splitter stage needs to satisfy the relation, Rk(mu + 1) rp + Rl.

In some of the splitters I’ve built using the 6SL7 family the Rk has been 75K sourced from –150 volts. That could just as easily been a 12AX7. The result is about one ma in each triode. Checking the 6SL7 curves shows rp to be ~58K & mu to be ~70 at that point. The plate resistors (Rl) used are 100K each.

Plugging the numbers into the relationship we get-

Rk(mu + 1) rp + Rl

75 ( 71 ) 58 + 100 (all resistances are K)

and 5325 170

Here is the Mullard 520. The ECC83 (12AX7) is running about 0.5 ma in each section. So rp is ~80K & mu is ~100. Neglecting the following grid resistors, Rl is 180K. Rk is 82K.

Plugging the numbers into the relationship we get-

Rk(mu + 1) Rp + Rl

82 ( 101 ) 80 + 180

So 8282 260

You might think that an N-Fet or MJE340 tail would be infinitely superior. But in practice that may not happen. Balance depends just as much on the two Rls being equal which can be done with good accuracy. The two sides of the twin triode may not track as well over their operating range. Could the difference be heard? Perhaps.

Here is an example of what does not work well, for fidelity anyway. Some circuit designers in the past (30s, 40s) have tried to get self-inversion in the power stage using a long tail. A typical stage is PP 6V6s, loaded by 8K, p-p. So each Rl is 4K. The longtail in one of these circuits is a series connexion of two resistors of 0.27 K each. The grid bias is taken off at their midpoint. The Class A operating point of the 6V6 results in rp of 50K & gm of 4.1 ma/v, so mu is 205.

Again, plugging the numbers into the relationship we get-

Rk(mu + 1) rp + Rl

0.54 ( 206 ) 50 + 4

55.35 54 Ouch! The relationship fails

If it made sense that could have been easily fixed by using a small iron cored choke, for example, the Hammond 157M. Notice I’ve picked this one since it is 259R, so can serve as the cathode bias resister as well. The 8H will still looks like 2.5K at 50 Hz. The 157M would have been too expensive but does demonstrate what is possible.

Today another fix would be an N-Fet. No low frequency problems in that part of the circuit anyway.

Some folks still like the self-inverting output stage. It is useful as a musical instrument accessory for the even order harmonics (2nd, 4th, ..Etc) it generates. See one at this link-

http://ax84.com/media/ax84_m276.gif