imbalanced 12ax7's due to 12.6v filament

[black_labb]

i built an amp a few weeks ago. its a fairly traditional amp with a 5f6a order of stages but with a different tone stack, a pentode instead of a triode preceding the cathode follower, a post pi MV (dual gang pot type, cant remember the name) and an output section along the lines of a ac30.

i originally had issues with a lack of power/volume. i thought it was related to the pair of output tubes being almost finished (which i had run for a while with with rectifier diodes bypassing the cathode bias resistor instead of the zener diodes i planned to have in the spot. didnt notice the redplating for a long while as the chassis was upside down so i could measure voltages.)
i ordered some new tubes and just started to test the amp again now and realised that the issue wasnt the (amazingly alive) tubes.

when testing the amp with the new tubes found that the issue was that the PI was seriously imbalanced (one putting out ~100vp-p somewhat clean, the other put out 25vp-p of very not clean). i checked the plate resistors and found them both at 100k as expected. before doing further tests and checking the tube i remembered that the first stage was also oddly imbalanced (*see description below). i had run the 12ax7's on 12.6v as the pt has 2 filament windings and doing it this way should keep the filament windings a bit simpler. i floated the 12.6vac filament by connecting pin9 to a voltage divider off of one power supply nodes giving it 45v.
testing it shows that the voltage across the filament for each triode is balanced at 6.3v each, so it doesnt seem to be an issue. seems a coincident that both 12ax7's (the only tubes that are running off the 12.6v supply, the other filaments are supplied by the 2 6.3v supplies that make up the 12.6v)


anyone have any insight into this? sorry if this is kinda in the wrong forum, but i think i know the problem, i just want to understand the theory behind the issue. it seems to be a pretty odd situation imo.

*in my earlier tests i noticed that the plate voltage on the 12ax7 on the input triodes were very different. the input has 2 triodes in parrallel with different cathode bias systems and a different coupling cap out (one is designed to be a bright style channel while the other is for a normal channel). one uses a red led for biasing and the other uses a 1k5 resistor (with a .82uf cap bypassing it). the one with the 1k5+cap has a plate voltage around 95v, while the other is around 150v. both are running from a power supply node with a voltage of ~210v. this did not correct itself with a tube change.

[Wilder Amplification]

*in my earlier tests i noticed that the plate voltage on the 12ax7 on the input triodes were very different. the input has 2 triodes in parrallel with different cathode bias systems and a different coupling cap out (one is designed to be a bright style channel while the other is for a normal channel). one uses a red led for biasing and the other uses a 1k5 resistor (with a .82uf cap bypassing it). the one with the 1k5+cap has a plate voltage around 95v, while the other is around 150v. both are running from a power supply node with a voltage of ~210v. this did not correct itself with a tube change.

Either a bad grid leak resistor (doubtful but possible), or a leaky bypass cap, making the 1k5 appear to be close to a 330 ohm resistor and as such biasing the stage hotter, which drops the plate voltage accordingly.

Based on what I'm seeing you probably have about 1.5V at the cathode of the LED stage while having 0.5 at the cathode of the 1k5/0.82uF stage? On the 1k5 stage you should have about 1.3"ish" at the cathode, which would set your plate voltage at 130V.

[ThermionicScott]

A 5F6A PI should have 82K and 100K plate resistors -- that won't cause the amount of imbalance you're seeing, though. I guess I'd debug it by disconnecting the heaters from your 12.6V source and just hook them up to a 6V source temporarily.

[black_labb]

the PI is pretty much a ac30 PI with a couple differences. the amp doesnt have much to do with a 5f6a but the order of the stages is identical, but the stages themselves are different.

i'll check the bypass cap. i assumed it was good as the multimeter measured 1k5. will check though. the voltage across it is ~.8v, the voltage across the led is 1.7v for comparison.

just checked the voltage on the plates of the pi and they are both at ~185v, so i'd say its a different issue and not just a common 12.6v will keep looking.

[black_labb]

quick update

disconnected the bypass cap on the input valve that had the low plate voltage and it had no effect. voltage is 89v so no change. bias across the 1k5 resistor is .8v

i noticed on the PI that the voltage on the grids are 17v each, but the voltage at the top of the tail resistor (where the bias resistor and grid resistor meet) is 32v. thats putting 15v across the grid resistors which doesnt seem right to me. i am using 330k grid resistors to closer aproximate the impedance matching from the output from the hiwatt tone stack as this seems to effect the workings of the TS (modeled it in LTspice).
am i wrong in thinking that the grid resistors should not have voltage across them?
what would cause a voltage across the grid resistors on the PI? there is no NFB so its not an issue with the circuitry linked to the grids leaking voltage. (otherwise it would be different on the 2 sides).

[Wilder Amplification]

quick update

disconnected the bypass cap on the input valve that had the low plate voltage and it had no effect. voltage is 89v so no change. bias across the 1k5 resistor is .8v

i noticed on the PI that the voltage on the grids are 17v each, but the voltage at the top of the tail resistor (where the bias resistor and grid resistor meet) is 32v. thats putting 15v across the grid resistors which doesnt seem right to me. i am using 330k grid resistors to closer aproximate the impedance matching from the output from the hiwatt tone stack as this seems to effect the workings of the TS (modeled it in LTspice).
am i wrong in thinking that the grid resistors should not have voltage across them?
what would cause a voltage across the grid resistors on the PI? there is no NFB so its not an issue with the circuitry linked to the grids leaking voltage. (otherwise it would be different on the 2 sides).

On that 90V stage...are you sure it's not using a 150K plate resistor? Have you measured it?

There will be voltage across the grid resistors. This is due to the grid being only 1-2 volts more negative than cathode voltage, which is a bit higher than the tail resistor voltage.

[black_labb]

i have measured it.
i have an odd situation here, but the 2 triodes seem to be linked in their odd behaviour. i'll try to get the heater voltage to normal 6.3v heating. wished i didnt have to as i paid alot of attention to the heating windings. i put them all under copper foil tape before starting any other wiring. i may try lifting the float from pin 9 and connect the centrepoint somewhere else to see if that helps before ripping out my heater wiring.

[Steve Conner]

You can't measure the grid voltage of a LTPI with an ordinary meter. The impedance there is very high due to bootstrapping, and the meter loads it enough to change the voltage you're measuring.

[black_labb]

didnt realise that that you can't measure the grid voltage steve. my grid resistors are lower value at 330k, which should make the impedance 660k i thought i read, or would it be higher? either way if the measured but innacurate voltage isnt showing a problem i will look elsewhere.
thanks

[R.G.]

In reading this it struck me that you didn't say "the PI tube" anywhere, you only referred to "triodes". This is a long shot, probably not the issue, but just to eliminate it, did you by some chance make a PI out of parts of two different tubes?

If so, I would almost expect bad imbalances, as the two sections inside one tube envelope should be better balanced than two parts of separate tubes.

Again, probably not the case, but best to eliminate it as an issue.

[black_labb]

not the case rg, but important to clarify.

[Joey Voltage]

didnt realise that that you can't measure the grid voltage steve. my grid resistors are lower value at 330k, which should make the impedance 660k i thought i read, or would it be higher? either way if the measured but innacurate voltage isnt showing a problem i will look elsewhere.
thanks

can be much larger

[Merlinb]

The heaters are almost certainly not the problem- I wouldn't waste your time on them.

Since you are sure of the anode resistors on your PI, I would measure the grid leaks on your power valves. If one of them is abnormally low in value it would unbalance the PI.

As for the input stage, either one or more resistors are not what you think they are, or there is DC on the grid (or one of the triodes is just bad). I suspect the first option since 0.8V across 1.5k is only 0.5mA, which is unusually low.

[loudthud]

Although I have seen imbalance between sides in dual triodes related to series and parallel heater connections, it's small, generally less than 5%. Imbalance in the LTPI is a function controlled by the value of the tail resistor.

[mooreamps]

*in my earlier tests i noticed that the plate voltage on the 12ax7 on the input triodes were very different. the input has 2 triodes in parrallel with different cathode bias systems and a different coupling cap out (one is designed to be a bright style channel while the other is for a normal channel). one uses a red led for biasing and the other uses a 1k5 resistor (with a .82uf cap bypassing it). the one with the 1k5+cap has a plate voltage around 95v, while the other is around 150v. both are running from a power supply node with a voltage of ~210v. this did not correct itself with a tube change.

and you have a coupling cap between these two parallel plates ?

-g