Matchless Spitfire clone cathode bias issue

[mxrshiver]

the issue is with the cathode biased push-pull 6BQ5 power amp, with a fully bypassed shared cathode resistor. under static conditions, the bias seems just right, but as signal is applied, a substantial negative DC voltage is appearing at the power tube grids for some reason. this is in addition to the positive voltage increase on the cathodes, making the bias extremely cold and drastically limiting the output. when plugged into a speaker it barely feels like a 1W amp, and indeed, the voltage measurement at the 8 ohm output with the amp at full tilt is only about 600mV.

here are my measurements. signal applied to the input is a 400Hz sine wave at around 400mV. controls all the way up unless otherwise noted. control grid measurements are actually taken before the grid stopper, as this junction is much easier to reach due to a dense build.

DC measurements with signal applied and master volume all the way down:
plates: 319.8V, 319.2V
screens: 319.5V, 319.6V
grids: 0.01V, 0.012V ***
cathodes: 12.39V

DC measurements with signal applied and master volume all the way up:
plates: 294.3V, 289.6V
screens: 292.9V, 292.6V
grids: -7.3V, -13.7V ***
cathodes: 22.22V

AC (signal) measurements with all controls at max:
plates: 40.6V, 41.8V
screens: 1.06V, 0.96V
grids: 38.9V, 28.3V
cathodes: 0.125V

one last measurement of note is, when the MV is all the way down, there is still a 6VAC signal on each power tube grid.

other notes:
- results are the same with the MV disconnected
- note the small positive DC voltage on the grids with the MV all the way down. i feel like this may be indicitive of the issue
- i did install an adjustable NFB loop with a 500K switched pot to break it. the loop is open for these measurements, and is fully decoupled with 100uF anyway, but i did notice that the switch's resistance actually only seems to be around 1M, so i thought it worth noting.
- also worth noting with the NFB loop: the .01uF that used to go from pin 2 of the PI to ground, now goes through 4.7K to ground, as the 4.7K is the bottom half of the NFB loop's voltage divider.
- i have also installed a very simple dummy inductive load circuit to simulate the impedance response of an 8 ohm guitar speaker, to be switched in between the 8 ohm tap and ground when the line out is being used. worth noting, but this loop is also open while the measurements were taken
- the sound output is actually fairly well behaved, despite this issue - no squealing, absurd levels of distortion. it's just not very pretty distortion, a far cry from how this elegant amp should sound.

[tubeswell]

Must be something in the way the MV is wired. Disconnect the MV. What happens?

[mxrshiver]

Must be something in the way the MV is wired. Disconnect the MV. What happens?

that was my first thought, too, but no dice - symptoms are unchanged.

my other thought was perhaps one of the post-PI coupling caps. i tested them before installing, but i'm admittedly not great with my dwell time yet and it's possible i overheated it. would it make sense that the leakage is tiny at idle, and increases with signal level? and could these symptoms be caused by just one coupling cap? that would potentially explain the drastic difference in the AC signal voltages between sides. (power tubes are NOS and curve tracer matched within 5%, so that's not it)

[tubeswell]

Must be something in the way the MV is wired. Disconnect the MV. What happens?

that was my first thought, too, but no dice - symptoms are unchanged.

my other thought was perhaps one of the post-PI coupling caps. i tested them before installing, ....

But did you test the coupling caps for DC leakage? - See Fig 4-4 P57 (It could be one cap is leaking and the leaking DC is crossing the MV pot resistance, thus affecting both output tube grids, but you should actually test the caps for DC leakage first before going holus bolus and changing everything)

[Kaaskop]

DID YOU DOUBLE CHECK THE VALUE OF THE SCREEN RESISTOR? I BUILT THIS AMP AND ACCIDENTALLY PUT A 100K INSTED OF 100 OHMS RESISTOR IN AND I THINK I HAD THE SAME VOLUME ISSUE UNTIL I FIXED IT.

[PRR]

> making the bias extremely cold and drastically limiting the output.
> DC measurements with signal applied and master volume all the way down:
>> plates: 319.8V, 319.2V
>> screens: 319.5V, 319.6V
> DC measurements with signal applied and master volume all the way up:
>> plates: 294.3V, 289.6V
>> screens: 292.9V, 292.6V

That's not "cold". Plates drop 25V, screens drop a little more--- the amp sucks MORE current. As the cathode voltage is telling you.

The drop at grids and rise at cathode is expected if you are OVER-driving the amp. ("Blocking")

I would re-check that speaker side path. It may be bad and sucking-up your power, forcing you to beat the beast to hear anything.

And check the actual load impedance. Primary? Secondary? Shunt loads? My guess (without any schematic!) is you may be driving 4K even 2K CT.

[pdf64]

Must be something in the way the MV is wired. Disconnect the MV. What happens?

That behaviour is normal for type 3 master vol. Due to the extreme loading on the LTP at low settings.
When turned down, the signal doesn’t mute because the signal there is 0, but because it’s become common mode.

[pdf64]

the issue is with the cathode biased push-pull 6BQ5 power amp, with a fully bypassed shared cathode resistor. under static conditions, the bias seems just right …

Any changes to the schematic component values?

[mxrshiver]

But did you test the coupling caps for DC leakage? - See Fig 4-4 P57 (It could be one cap is leaking and the leaking DC is crossing the MV pot resistance, thus affecting both output tube grids, but you should actually test the caps for DC leakage first before going holus bolus and changing everything)

yup, checked to 500VDC before i popped em in. so unless i ruined one with bad dwell time, they're good to go

[mxrshiver]

DID YOU DOUBLE CHECK THE VALUE OF THE SCREEN RESISTOR? I BUILT THIS AMP AND ACCIDENTALLY PUT A 100K INSTED OF 100 OHMS RESISTOR IN AND I THINK I HAD THE SAME VOLUME ISSUE UNTIL I FIXED IT.

that would sure cause some issues! they're both nice wirewounds confirmed at 100R

[mxrshiver]

> making the bias extremely cold and drastically limiting the output.
> DC measurements with signal applied and master volume all the way down:
>> plates: 319.8V, 319.2V
>> screens: 319.5V, 319.6V
> DC measurements with signal applied and master volume all the way up:
>> plates: 294.3V, 289.6V
>> screens: 292.9V, 292.6V

That's not "cold". Plates drop 25V, screens drop a little more--- the amp sucks MORE current. As the cathode voltage is telling you.

The drop at grids and rise at cathode is expected if you are OVER-driving the amp. ("Blocking")

I would re-check that speaker side path. It may be bad and sucking-up your power, forcing you to beat the beast to hear anything.

And check the actual load impedance. Primary? Secondary? Shunt loads? My guess (without any schematic!) is you may be driving 4K even 2K CT.

not sure i understand. my thought was that since the grid is becoming more negative and the cathode becoming more positive, the grid-to-cathode voltage is becoming extremely negative, about -30 to -35 volts per tube. this is a much colder grid bias voltage than the -12 it's sitting at at idle, and i believe leaves the 6BQ5's biased very far into cutoff.

either way, are you saying the negative DC voltage on the grid with signal applied is normal? i didn't think the grid voltage was supposed to move in cathode bias, just the cathode voltage. the amp is certainly being overdriven for these tests, but it's worth noting that the negative DC grid voltage begins to appear immediately as soon as the master volume is nudged upwards. so based on what you were suggesting, that seems to indicate there is something just sucking up the power somewhere, causing the tubes to be well overdriven even at low settings, without much output power.

i'm not sure it's an issue at the speaker or OT, because the point at which the measurements stop making sense is where the power tube grids are being driven with a nice 30-40VAC from the LTP, but they're barely amplifying the voltage at all, as barely over 40V is seen at the plate. as this is a 4K primary, the 0.6VAC seen at the 8 ohm secondary seems to indicate the OT is working about as it should, and the 4 and 16 ohms taps read voltages proportionate to that.

[mxrshiver]

Must be something in the way the MV is wired. Disconnect the MV. What happens?

That behaviour is normal for type 3 master vol. Due to the extreme loading on the LTP at low settings.
When turned down, the signal doesn’t mute because the signal there is 0, but because it’s become common mode.

i see! what do you mean by type 3, is that the same as the PPIMV? it seems like this explains the 6VAC present at the power tube grids even with the master volume all the way down - even though the two sides are completely cancelling each other out in the audio output, they each still have a measurable voltage signal.

[mxrshiver]

the issue is with the cathode biased push-pull 6BQ5 power amp, with a fully bypassed shared cathode resistor. under static conditions, the bias seems just right …

Any changes to the schematic component values?

not really! i've got a bit of an underpowered PT that delivers HT+ at about 10% less than specified, that's about it. the only other changes i made were to eliminate the low impedance input, and install the negative feedback loop and dummy load, both being switched out for these tests.

incidentally, i did try the negative feedback loop and it seems to perform exactly as expected... another thing that makes me think this is an issue with the push-pull power section, not the OT or speaker side of things.

[PRR]

The rise of cathode voltage is sure sign of increased current.

Drop of grid voltage may be signal overdrive -or- oscillation.

[tubeswell]

You said the fall in grid voltage occurs even when you disconnect the MV. Therefore, the issue can't be because of the MV.


Therefore, you could experiment with (in no particular order):

1) a 47k grid stopper on each output tube grid (delays the onset of grid current limiting)
2) reducing the capacitance of the coupling caps (for a faster coupling cap discharge time), or reducing the grid leak resistance for the output tubes (for a quicker discharge).
3) lowering the screen voltage so that the output tube load line is closer to/above the knee of the Vg0 grid curve (which limits overall tube current under heavy signal conditions, and thus decreases the tendency for grids to charge up under full signal
4) check your output tube load isn't too low (as a low reflected impedance will encourage more tube current when you drive the amp hard)

[pdf64]

I don’t like the ppimv acronym. Ken Fischer provided a nomenclature for master volume types in The Trainwreck Pages, p32 https://robrobinette.com/The_Trainwreck_Pages.htm

Note that under the test conditions described, none are with the power amp idling, signal is always being applied to the control grids.
You need to remove the signal or turn down the volume control.

Control grids shift negative when overdriven due to grid rectification. Have a read of  https://www.ampbooks.com/mobile/amplifier-calculators/bias-excursion/

I suggest to avoid probing output valve anodes, it can invoke oscillation. And especially when overdriven, no regular meter will be rated to accommodate the extreme voltages there.

[mxrshiver]

You said the fall in grid voltage occurs even when you disconnect the MV. Therefore, the issue can't be because of the MV.


Therefore, you could experiment with (in no particular order):

1) a 47k grid stopper on each output tube grid (delays the onset of grid current limiting)
2) reducing the capacitance of the coupling caps (for a faster coupling cap discharge time), or reducing the grid leak resistance for the output tubes (for a quicker discharge).
3) lowering the screen voltage so that the output tube load line is closer to/above the knee of the Vg0 grid curve (which limits overall tube current under heavy signal conditions, and thus decreases the tendency for grids to charge up under full signal
4) check your output tube load isn't too low (as a low reflected impedance will encourage more tube current when you drive the amp hard)

excellent suggestions, thank you so much! i'm actually going to try 1-3 in order, because they progressively have more potential to impact tone, but the 47k grid stoppers shouldn't have much or any effect and is a good place to start.

i've also had someone suggest increasing the size of the screen stoppers, and adding 47pF in between the PI plates before the coupling caps, which both sound like minimal (if any) impact on tone. thoughts?

the impedance of 4k is pretty low in general for this setup, but that's part of the Matchless magic, as i understand it. but i'm starting to understand the picture of how multiple factors (tiny grid and screen stoppers, relatively low load, screen voltage at or even above plate voltage) could be inviting oscillation, here.

[mxrshiver]

I don’t like the ppimv acronym. Ken Fischer provided a nomenclature for master volume types in The Trainwreck Pages, p32 https://robrobinette.com/The_Trainwreck_Pages.htm

Note that under the test conditions described, none are with the power amp idling, signal is always being applied to the control grids.
You need to remove the signal or turn down the volume control.

Control grids shift negative when overdriven due to grid rectification. Have a read of  https://www.ampbooks.com/mobile/amplifier-calculators/bias-excursion/

I suggest to avoid probing output valve anodes, it can invoke oscillation. And especially when overdriven, no regular meter will be rated to accommodate the extreme voltages there.

ugh, absolutely excellent reads, i'm going to be going back to that Trainwreck script many many times. i can't imagine how hard it would be to get into this biz without wonderful folks like Rob, and yourself 💜 i can certainly see why PPIMV is not a very useful phrase, it describes 3/4 of the MV's generally used!

apologies, i seem to have missed something - which test conditions are you referring to? the ones on the schematic, or somewhere in the Trainwreck pages? and what are you suggesting removing them for? the original measurements i took are with no signal applied.

😳 good to know about probing power tube anodes... REALLY good to know. what do you suggest for taking live voltage readings there, then? would attaching the probe before i turn the amp on help avoid oscillation? i have the Fluke 87V, here's what it has to say about voltage tolerances: "All the inputs in the 87V are updated to the third edition of the EN 61010-1 to CAT III 1000 V and CAT IV 600 V. The meter is designed to withstand spikes of more than 8000 V."

it might be worth re-iterating that the main symptom/issue i'm experiencing here is a drastic lack of volume output when signal is applied. the idle voltage readings seem to be perfectly fine. but the output tubes do not seem to be amplifying correctly, even when they're only being fed a very small amount of signal. since the grid DCV seems to begin to drop drastically whenever i apply the smallest amount of signal, it seems as if the power tubes are acting as if overdriven under any drive conditions. the signal voltage does not seem to be doing the overdriving, though, as i get a very normal transition into distortion as the MV is turned up. based on my limited knowledge, this does seem to point to hf oscillation.

i think i'd like to fire up the scope and confirm the oscillation first, as i've never done that before. it occurs to me that could be done with no signal applied - is that what you were getting at before? any other tips for safe probing? planning on keeping the MV at low levels.

[pdf64]

... which test conditions are you referring to? ... the original measurements i took are with no signal applied.

I meant those in original post ("here are my measurements. signal applied to the input is a 400Hz sine wave at around 400mV. controls all the way up unless otherwise noted") where you're measuring VAC at the output valve grids with the type 3 master vol turned down. Despite the amp not putting any sound out, the power amp won't be idling, rather it's working fairly hard in common mode, as a 'pull pull' amp, rather than 'push pull'.
The signal needs removing for the amp to idle, so that idle conditions can be established.

😳 good to know about probing power tube anodes... REALLY good to know. what do you suggest for taking live voltage readings there, then?

I don't probe output valve anodes as a default thing. The anode VDC will be as near to the HT VDC feeding the OT as makes no difference, so take VDC readings there. And any AC signal at the anode will appear at the OT secondary, albeit at a somewhat reduced level.
I think it regretable that advice to take readings at output anodes has become so ubiquitous on the internet, but we are where we are 

i think i'd like to fire up the scope and confirm the oscillation first, as i've never done that before. it occurs to me that could be done with no signal applied - is that what you were getting at before? any other tips for safe probing? planning on keeping the MV at low levels.

An inadequate margin of stabilty can manifest in free running oscillation, or parasitic oscillation; the latter requires a signal to parasitise, whereas the former can usually start itself off, eg at particular control settings. Sometimes though it needs a decent signal to kick start it off,  usually background noise is sufficient.
Amps might be stable into resistive loads but unstable with reactive loads; or, more unusually, visa versa.
Parasitic oscillation is typically a power amp issue, especially class AB, as operating conditions change according to the signal level.
Obviously, if overdriven with a constant test signal, it's pretty much impossible to detect oscillation.
Low level square waves fed directly into suspect stages can be helpful in identifying instability.

the voltage measurement at the 8 ohm output with the amp at full tilt is only about 600mV

If you're only putting 600mVAC out to an 8ohm load, something is seriously wrong; eg the OT is bad, the output jacks are wired wrong.