[R-390] R392 - potential easy cure for horrible audio

[Charles Steinmetz]

The 26A7 (push-pull pentodes with common cathodes, screens, and 
suppressors), used as the final audio tube in R392s (V608), was 
designed to be run with 25-50 V on the plates and -4.5 to -7 V of 
grid bias.  The R392 runs it with ~28 V on the plates and only about 
-0.25 V of contact bias developed across the 470k grid 
resistors.  The 26A7 datasheet contemplates zero-bias operation,  but 
that is clearly abuse and results in about 60 mA of no-signal plate 
current per side rather than the normal 20 mA or so (i.e., it steams 
away at 1.5 watts plate dissipation per side, sounding awful and 
killing itself).  When punished this way, a 26A7 needs to be driven 
by an interstage transformer because the grids need to be driven 
positive (class A2 operation).

The R392 drives the 26A7 grids through coupling capacitors from 
split-load phase inverter V607, NOT through an interstage 
transformer.  Thus, when there is audio the 26A7 grid current charges 
the coupling capacitors on positive signal peaks (to the extent that 
it can, given the limited current available from V607), which 
develops a bit of additional negative grid bias.  The good news is 
that this reduces the 26A7 plate current somewhat when there is 
audio.  The bad news is that this bias is generated by rectifying the 
audio, so the audio performance of the R392 is worse than 
horrible.  The time constant of the 26A7 grid circuit is relatively 
short (t = ~47 mS), so the bias drains off rapidly between audio 
peaks and pumps all over the place with the varying audio signal.

To top it off, because the ability of the split-load phase inverter 
(V607) to supply positive grid current is asymmetrical, the 
corresponding grid bias and grid distortion in the two halves of the 
26A7 are also asymmetrical, leading to additional distortion from the 
resulting unbalanced DC in the output transformer primary.  It works, 
sort of, and it provided fidelity sufficient for intelligibility in 
the military context, but it is fugly in the extreme and makes using 
an R392 today a painful experience.

A transistorized "audio module" was used in some R392s in place of 
the 26A7.  It is shown in Figure 20 of the TM-11-5820-334-35 manual 
and comprises 4 ancient (3-digit) transistors, two transformers, some 
resistors, and a capacitor.  Hardly state of the art, but better than 
the 26A7.  It should not be hard to design a simpler solid-state 
replacement for the 26A7 (one without any internal transformers) 
using modern solid-state devices.

Alternatively, I have sketched out a super simple mod to the existing 
circuit that converts V608 to a Class A2 amplifier and should be a 
huge improvement in both sound and tube life (it's completely and 
easily removable).  I haven't tried it, because I've never owned an 
R392.  However, analysis indicates that it should perform as intended.

I'm looking for one or two R392 owners to beta test the 
mod.  Volunteers should:

1)  have a properly-working R392 that they are willing to take apart 
and experiment on relatively soon (please do not respond if your R392 
has other problems)

2)  have the skills required to troubleshoot and repair an R392

3)  have access to junk box parts or can buy a few cheap parts ($10 or less)

4)  are able to measure the plate or cathode current of the 26A7 
(V608) before and after

Bonus points for:

5)  are able to measure the distortion before and after

Please let me know if you are interested in being a beta tester, and 
confirm that you have the capabilities listed above.

If the test results are satisfactory, I will post the mod.

Thank you,

Charles