Any basic paraphase "tricks" (for smoothness...)

[EFK]

I tried the split cathodes unbypassed as well as bypassed (both) and the shared cathodes both ways as well. In all cases, when I add the bypass cap/caps across the cathode R, the balance goes way out of whack and no adjustment of the divider R will bring it back to near balance.

Why is the lower output grid R above so much larger? That is almost identical to my version, although my cathode is currently shared and both output grid R are the same value (270K).

[octal]

For symmetry, sometimes the designer will make the combination of the divider string equal the grid resistor on the other phase. Obviously, 270K + 12K does not equal 470K in this case. It's not super critical, but the mismatch in the sample schematic is larger than normal for whatever reason.

You absolutely should be able to achieve balance unless something is broken or you've got a measurement error of some kind. When you bypass pin 8, the voltage at "B" will increase to be higher than "A." If you reduce Rx, the voltage at point "B" must go down. Perhaps you never tried an Rx small enough. You made reference to "raising" the value of the 4.7K... that's the wrong direction to go!

Nathan

[EFK]

I had to raise the resistor as when bypassing, the input triode may gain a volt or two, but the paraphase triode drops a whole pile of output (this is w/ shared cathode R). So that R to ground would have to be raised to feed the paraphase grid more signal, right?

Still experimenting here, don't understand it all. Here's one: I watched the VAC on the output tubes really carefully while *very* slowly rotating the volume. Above, I mentioned this weird little "dip" in the 1Khz signal as the volume was raised - it's happening between @ 9 o clock and 10 to 11 o'clock on the volume rotation, at which point the outputs are at @ 40 VAC. What is happening is that before I can hear the little volume dip, from say 0 to about 9 o'clock, the paraphase output (measured on the output tube grids) is a volt or two higher than the input triode output, but when the volume begins to dip, they become equal, then they flip-flop, at which point from there up to full volume the input triode output is higher than the paraphase triode output. Does this mean anything?

[EFK]

Well here is some more weirdness. Measuring voltages throughout the amp, everything has a pretty consistent voltage drop from idle to full volume EXCEPT the plate voltage at the paraphase input triode; that one drops from 233 V at idle waaaaay down to 136 V at full volume. That is much, much more of a voltage sag than anywhere else in the amp (typical in this w/ 5V4 is b/t 30-40V). Coincidentally - or not - the power tube which that triode is feeding shows a steady climb to -18 VDC on the grid, but the other power tube (fed by the paraphase triode) only goes to -3 VDC. Just for the fun of it I flipped the power tube grid leads and the noted effects follow whichever power tube is fed from the PI input triode.

Meanwhile, as I wind up the amp, the voltage across the shared cathode resistor for the paraphase goes from 2.42 VDC at idle, to a drop down to 2.30 when it hits the little flip-flop at @ 9 oclock volume, then steadily climbs back up to 3 VDC at full volume.

Does any of this mean anything? I bet if I could get this amp on a scope, and actually knew how to use it, I'd see some weird stuff.

[pdf64]

I'm wondering if there's oscillation going on; are there grid stoppers on the power tubes? Does this (ie the strange paraphase Vdcs) still happen if the power tubes are removed?Pete

[EFK]

Currently I have 27K stoppers on the output tubes. IF I yank the tubes, with no signal injected, all voltages in the amp remain stable (albeit higher) EXCEPT the paraphase triode plate voltage, which drops a total of 41 VDC between idle and full volume. Once I inject signal, still w/ no power tubes, both PI plate voltages drop (input triode and paraphase triode), this time the input triode plate drops more. VAC on the output grids w/ no power tubes is much more out of balance, 118 VAC on the input PI side but only 89 VAC on the paraphase side.

I suspect all of this means there is an oscillation sucking up current, but I'm not sure where it is other than that it's in the PI somewhere. ???

[pdf64]

How about if 22k grid stoppers are fitted on the paraphase tube grids?

[jpfamps]

I use a splitter very similar to the schematic in octal's post in our amps. I nicked it from a particularly good sounding Valco built Gretsch amp that came in the shop.

I doubt very much whether your problems are caused by oscillation.

In my experience absolute balance isn't crucial to getting a good overdrive with this inverter, however when the non-inverting phase output is significantly higher than the inverting phase, you get some fairly nasty distortion artefacts. My thinking behind this is that the non-inverting drive output goes into grid conduction much earlier than the inverting phase results in half wave rectification of the signal peaks. This doesn't sound good.

Conversely if the if the inverting phase output goes into grid conduction, then non-inverting phase output will also be clipped as it is simply an inversion of the inverting phase output.

Did that make sense?

Thus you may want to try skewing the inverter to giving slightly more output on the inverting phase.

You also need to be careful that neither triode stage in the inverter is clipping hard assymetrically. The two main causes of this will be grid conduction and inappropriate biasing.

[EFK]

Pete the two channel mix resistors (470K and 220K) junction right on the input of the PI input side, so aren't they effectively grid stoppers for that triode? They aren't right on the socket, but it's a very short run from there to the socket - @ 2.5 inches - and being as it was pretty sensitive I used a piece of shielded wire right to the socket. The other input to the paraphase triode seemed less sensitive, and that wire is only @ 2 " long. If I put a stopper on that, it's going to drop the balance on that side a bit I bet as there probably will be a tiny signal reduction. Couldn't hurt to try though!

jpfamps - that is some very interesting information! To be sure we are on the same page, the stage you are calling the 'inverting phase' is the second-half of the 12ax7, the side getting fed from the (in my case) 270K/6.8K divider, correct? As I mentioned, I don;t have a scope. Is there a way I can check for grid conduction w/ just a DVM? I know I'd have no way of knowing whether it was clipping asymmetrically w/o a scope, but if I could check for grid conduction that might be helpful.

[jpfamps]

jpfamps - that is some very interesting information! To be sure we are on the same page, the stage you are calling the 'inverting phase' is the second-half of the 12ax7, the side getting fed from the (in my case) 270K/6.8K divider, correct? As I mentioned, I don;t have a scope. Is there a way I can check for grid conduction w/ just a DVM? I know I'd have no way of knowing whether it was clipping asymmetrically w/o a scope, but if I could check for grid conduction that might be helpful.

No. The inverting phase is after the 1st triode stage. The signal after the second triode stage is in phase with the input to the phase inverter as a whole.

I would try to balance the phase inverter so there is slightly more signal on the power valve grid after the 1st stage than the second. To be sure you have avoided grid conduction us a smallish signal or remove the power valves.

Investigating grid conduction/ clipping without a scope is hard.

You can try measuring the signal on the power valve grid with/without the power valves in situ. You should achieve higher maximum signal swing without the power valves in place, cf the power valves in.

You should also notice that the signal on the power valve grids reaches a maximum value (which is less than the value without the power valves) as you increase the gain. This is usually due to grid conduction.

[EFK]

This has been very informative and I appreciate all the help! FWIW, I've temporarily left pots on both the PI shared cathode and the lower resistor of the divider feeding the non-inverted grid. I also have different 'values' clearly marked on the pots so I can consistently test different settings while both using my ears as well as measuring VAC on the power tube grids. What I have found is (1) this circuit DOES NOT LIKE BALANCE and it does not sound good. As jpfamps has noted, it is best if the inverted stage feeds a stronger signal to the power tube than the non-inverted stage. Like 10 to 12 VAC difference at max power! The closer I get them, the worse it sounds and the more prominent the weird little "phase shift" as I'm calling it on the volume knob is. I find that if I balance them closely at max power, what happens is that when the volume is lowered they get even closer together and there is point where they 'cross' and change positions. This creates a harsh sound if breakup is occuring. 10 VAc apart at max volume reduces to only about a 2 volt difference at idle. (2) Depending on value of cathode R and divider R value, there was a point where the VAC would increase as I raised the volume, then hit a plateau (usually near about 3-4 oclock) before taking a little dip backward as I raised the knob up to full. I *think* this might have been indicating grid conduction? All I know is that it sounds better if I adjust values until that does not happen.

I also experimented w/ different tubes in the PI. This PI design was actually originally based around a 6SC7 tube, which I think is a little less gain than a 12ax7, and Valco did not use this design of inverter w/ 12ax7 for more than @ 2 years before going over (late 1959-1960) to the more common split cathode 2.2K bypassed/3.9K unbypassed PI that they used on most amps through the 1960s. My guess is that they ran into problems too. 5751, 12at7 and 12ay7 all are useable in the circuit, although I notice that with those tubes (not so much the 5751 but def. the at7 and ay7) the plate voltage on the PI plates at idle drops dramatically and I have to use a much higher cathode R to bring it back up so that there is leeway for it to sag under full load.

(Can someone briefly mention why these tubes drag the PI voltages down?) *Edit: did some reading, figured that one out*

The at7 and ay7 do clean/tighten things up quite a bit, although despite being "lower gain" tubes they do not drop the VAC on the output grids by more than a few volts at most and the amp doesn't lose hardly any volume, just saturation.

I always come back to the stock 4.7K lower divider resistor - despite the tube or cathode R, that consistently sounds the best so there would be no use in this particular case leaving a pot there. I could see tacking a pot onto the shared cathode R, though, as it is necessary to adjust that depending upon the PI tube used and it helps when going to different pairs of power tubes.

[soundmasterg]

The 6SC7 is basically a big 12AU7....gain about 20, high current. They were used in the Silvertone 1484 and 1485 as a cathodyne inverter and driver, among other places. Very nice sounding tube for hifi applications, and sounds ok in the Silvertone, but I think there are better tubes to use in guitar amps. You might check out the Vox AC100 phase inverter circuit....a floating paraphase using a 12AU7. I think it sounds pretty good myself, and works fine in that application, though it is driving higher gain EL34's instead of 6L6's or something like that.

Greg

[MWJB]

a 6SC7 has a mu of 70, not 20. The 1484 & 1485, as well as other 60's Silvertones/Danos typically used a 6CG7.

The 6SC7 is a twin triode, but both triodes share the cathode...this is what would steer me away rom using it in a paraphase PI (though Fender did just this in the very early 50's).

[soundmasterg]

a 6SC7 has a mu of 70, not 20. The 1484 & 1485, as well as other 60's Silvertones/Danos typically used a 6CG7.

The 6SC7 is a twin triode, but both triodes share the cathode...this is what would steer me away rom using it in a paraphase PI (though Fender did just this in the very early 50's).

Ah woops...got my numbers confused there, thanks for the correction! I shouldn't post when I'm tired...haha.

Greg