Large difference in 6V6 plate voltages; 5A3 Deluxe

[akulikow42]

Hi everyone...

Working on a 5A3 Deluxe that has ~ 20vdc difference in plate voltages on the 6V6s with one plate being equal to HT / center tap voltage. Thus, the bias is wildly different for each tube. It will be ok with equal voltages for ~ 30 sec once off the current limiter, then voltages start to fluctuate and eventually settle with the imbalance. Problem stays with the socket (i.e. swapping OT leads, tubes ect. doesn't affect which tube has the high voltage that = CT) and I can instigate the behavior by probing the grid of the tube with the higher voltage with my multimeter.

Things I've investigated with no luck:
- swapped in three sets of tubes & rectifiers
- confirmed cathode connections and appropriate bias voltage on each tube - stays balanced throughout but voltage drop across the cathode resistor doubles when behavior starts
- Disconnected grids to rule out something going on under the board after the PI
- Swapped reservoir cap
- Jumped in a new cathode bypass cap
- Tried two different OTs (one with a near perfectly balanced CT)
- Swapped out the socket with the higher voltage / grid probing behavior.

Frustrated since this is the second time this simple circuit has thrown tough stuff my way... any thoughts are much appreciated!

 

[Pete Farrington]

Maybe oscillation, try grid stoppers, eg 10k at each 6V6 socket lug 5.

 

[Helmholtz]

- Swapped out the socket with the higher voltage / grid probing behavior.

I'd rather suspect the other socket with the higher plate current.

What is the cathode voltage in problem mode?
What are gid voltages?
Do you have a scope?

 

[akulikow42]

@Helmholtz cathode voltage jumps from ~ 20v to 40v in problem mode. Grid-ground voltage on the higher-plate-current tube is < 0.01v. Grid-cathode is the negative of above cathode voltage. On the other tube… the grid voltage appears ~ +5v at first but then wildly fluctuates in my DMM. I do have a scope but still a novice… I’ve been reluctant to run a signal through given that one tube is already idling super hot.

I’ll try grid stoppers as @Pete Farrington suggests too and report back…

 

[Helmholtz]

@Helmholtz cathode voltage jumps from ~ 20v to 40v in problem mode. Grid-ground voltage on the higher-plate-current tube is < 0.01v. Grid-cathode is the negative of above cathode voltage. On the other tube… the grid voltage appears ~ +5v at first but then wildly fluctuates in my DMM.

Imagine a leaky socket or conductive board raising the grid voltage of one tube.
The cathode (and plate) current of the tube would increase, raising the common cathode voltage.
Consequently the other tube would be biased much colder (probably even into cutoff with a Vgk of -40V) significantly lowering its plate current .
A common cathode resistor tends to increase power tube imbalance.

 

[akulikow42]

Unfortunately, @Helmholtz I swapped out the other socket and no dice. Also tried the 10k grid stoppers al la @Pete Farrington and no difference...

Unsoldered the PI coupling caps to test for DC leakage and run leads directly to the grids to omit the board. Observed the following behavior:

- the higher plate current/lower voltage follows the grid of the inverted (left side); this grid has ~ +18vdc on the grids when probed which quickly falls to ~ 400mv as the meter drags it down
- the higher voltage that equals CT has the expected Vgk and no DC on grid
- The plate voltages but not the grid voltage behavior switches sides when leads of the OT are swapped
- no measurable DC on the grid side of the caps but I swapped out the suspect side of the PI cap anyways...

Any thoughts on what could be pulling the grid of one tube so positive when the caps / tube sockets aren't leaking and the board connections are taken out of the picture?

 

[peteb]

one plate being equal to HT

This means no conduction.

Any thoughts on what could be pulling the grid of one tube so positive

The grids are not supposed to go positive.

 

[akulikow42]

So, I disconnected the grids from the PI to look for oscillation before the output tubes... waveforms out of each side of the PI look fine. With the grids disconnected, both tubes were conducting evenly with appropriate voltages on the plates and no +voltage on grids but... after a minute the hum changed and the behavior started again with an increase in all voltages. Once it stabilized, there was 24v on one grid and ~ 1v on the other...

 

[2L man]

So, I disconnected the grids from the PI to look for oscillation before the output tubes... waveforms out of each side of the PI look fine. With the grids disconnected, both tubes were conducting evenly with appropriate voltages on the plates and no +voltage on grids but... after a minute the hum changed and the behavior started again with an increase in all voltages. Once it stabilized, there was 24v on one grid and ~ 1v on the other...

Look like Coupling Capacitor leak current from PI HV and screw the bias. If you have completely cut the connection to PI then possibly tube socket (dirt, carbon, oxide/moisture etc) cause the leakage? Or circuit board conductivity?

You could short both Control Grids and if combined voltage move somewhere between separate CGs then it is leakage problem. If voltage come and stay correct then problem side had conductive problem in Grid Leak Resistor where bias voltage come. Possibly resistor itself?

If this is new build verify that OT Center Tab connect to the HV! There should be voltage loss on both primary coils!!! Functioning coil should have normal loss what its anode current cause. Bad side should have much higher loss which very hot bias would cause!

If you measure coil half resistances you can calculate what the voltage loss should come for the bias (cathode) current you have measured deducting 5% what Screen idle current usually is!

Screen voltages are essential to have and know!

 

[Helmholtz]

Symptoms could indicate a missing ground reference at the grid which goes positive.
Measure resistance from each grid to ground and check the resistor string between the power tube grids (250k-7k-250k) for values and connections.

 

[Jasonpatrick]

This kinda sounds like when I’ve had OTs laced up wrong. In other words, the red CT wire is switched with say the blue wire internally.

 

[Wally]

This kinda sounds like when I’ve had OTs laced up wrong. In other words, the red CT wire is switched with say the blue wire internally.

Interesting. A quick check of the resistances between those leads would quickly reveal an unexpected factory error.

 

[peteb]

I would make sure the control grids have a reliable ground reference.

Isolate the grids by disconnecting the coupling caps.

Run the tube.

verify conduction by measuring the cathode voltage and measure plate current at the OT.


Then if there is still unwanted voltage on the control grids, figure out where it is coming from. It has to come from some where.

 

[Helmholtz]

This kinda sounds like when I’ve had OTs laced up wrong. In other words, the red CT wire is switched with say the blue wire internally.

How could such wrong OT connection affect the bias?

 

[Jasonpatrick]

How could such wrong OT connection affect the bias?

Well, I think, wouldn’t one side have more resistance if you hooked the plate to the ct and the ct to the plate? I think mine was off by like 20 volts. Bias was off by -10 between the tube and it had a horrible noise floor.

 

[Jasonpatrick]

Interesting. A quick check of the resistances between those leads would quickly reveal an unexpected factory error.

That’s how I discovered it.

 

[Helmholtz]

Well, I think, wouldn’t one side have more resistance if you hooked the plate to the ct and the ct to the plate? I think mine was off by like 20 volts. Bias was off by -10 between the tube and it had a horrible noise floor.

Plate load has close to zero influence on plate idle current with a pentode/tetrode.
Plate current is determined by Vg1-k and Vg2-k.

 

[GilmourD]

What happens if you flip the leads from the OT between the tubes? Does it follow the lead of the OT or stay with the socket?

 

[akulikow42]

Thanks everyone. @Helmholtz and @peteb called it. The ground reference for one of the grids was jumpy and I found that it's grid leak resistor was bad. It was a new metal film that I measured before putting it in the amp so I just wasn't "seeing" it as a potential issue. Well, a whole Sunday down the tubes but the knowledge that comes with the "Doh!" moment is priceless, I guess.

Again, very appreciative to have you all as a resource while I improve my skills.

 

[Jasonpatrick]

Plate load has close to zero influence on plate idle current with a pentode/tetrode.
Plate current is determined by Vg1-k and Vg2-k.

Just telling ya what I saw on mine