I planned a post on the μProcessor regulated hybrid fixed / cathode bias of my Project 1 KT77 push-pull tube amplifier. A major failure with the high-precision cathode resistors interrupted things. Please do not laugh, I know I had it coming…
Update uncovering an unexpected failure mode of cathode resistor
I was installing an update to my Project 1 amplifier to measure the cathode current of tubes of my Project 1 . In anticipation, I reserved easily accessible contacts on the PCB, one on each side of the cathode resistors. The photograph and schema below shows how.
A low-pass filter attenuates the residual AC across the cathode resistors and attenuates any noise from the bias measurement and the cabling back into the amplifier. Also the resistors, in combination with a TVS diode on the μController’s board, protect the ADC circuitry during a catastrophic failure
In order to reassure myself of which side of the contacts was ground (right side of cap C1, pin J2), I measured the resistance to ground. So the cathode side was supposed to be 100Ω, the ground side (close to) 0Ω.
One channel measured fine and I connected the quiescent measurement interfaces for both tubes. The other channel confused me. Both resistors where about 90Ω and with a few ohms difference between them. That is odd for precision resistors. They were 0.1%.
What was going on here?
The challenge was that the PCBs are quite high density for a tube amplifier and had components on both sides.
Aside: I did not minimize the PCB to the max, but compared to its father, the UL40-S2, the PCB are smaller while the caps are larger.
These resistors were on the tube side and, as you can see in the photograph, hidden from view. That implied dismantling the PCB. Besides reconfirming the amplifier was not designed for easy access of components, what I saw was a set toasted resistors.
I installed 0.1% resistors, with a temperature coefficient of 50ppm, enabling accurate quiescent current balancing. Not so accurate any more… The 100Ω cathode resistors are actually 250/250/500Ω parallel. This allows a smaller resistor size, better fitting the PCB layout I wanted. This 250Ω resistor measures 203.7Ω.
I was expecting resistors to fail open, but clearly not these flame resistant ones…
Why did this happen?
14 months ago
While experimenting earlier (14 months ago), I accidentally left the negative bias wire unconnected. That meant for a short time the amplifier worked with only cathode bias. A resistor of 100Ω per tube pushed the tubes into an excessive current of about 400mA! I was there and measuring the current and hit the power switch in seconds.
The B+ is controlled through a MOSFET. Switching off such a high current so quickly caused a surge. This surge was drained by carefully placed TVS diodes, so no sparks nor any OPT damage. But the current had to go somewhere and tripped the fuses on the primary and secondary windings of the power transformer.
After connecting the wire and replacing the fuses all worked normal again. At least that is what I thought. In hindsight, the event might have damaged the tubes severely. The amp worked fine for a while but severe microphonics developed over a matter of weeks playing otherwise normally. At first, I could not find the reason and suspected a lose contact. However, nothing to be found.
Eventually the situation deteriorated into thunder from the speakers. I ran to the amp to hit the power switch. One of the tubes proved bad and had also a visual clue: compared to the other tubes, the cathode seemed a bit out of place. I cannot be sure if that was a manufacturing intolerance or that it was caused by the overload. Anyway, remarkable…
I replaced the tubes and all was fine. I left the other channel’s tubes in place as these worked fine and were not part of the disconnected negative bias supply wire incident. The necessary bias voltage between both channels differed a bit, but by replacing only one channel’s tubes, the tubes between the L and R channel were not paired like they initially were (I bought at the time a paired set of 4, now only a paired set of 2).
I did not further worry about the difference in negative bias voltage. I should have though. Now I know the cathode resistors were quite off because of this severe overload. And I was running one channel’s tubes 10% hotter and no longer balancing the quiescent current as intended. The balancing is very important for the reproduction of micro details (more about that in a later post).
Awaiting new high-precision resistors
Now replacing the resistors. Oeps no stock…. Worldwide shortage of components. Luckily Mouser could help with a few resistors. The replacements have a slightly higher wattage (1.75W instead of 1W, nominal load is 0.144W). But otherwise the same series, accuracy and temperature coefficient. All should be fixed in a week or so when Mouser delivers and I find a moment to repair. I am so eager to the amplifier playing some good music again.
Update: The resistors arrived and are installed. All works like a charm.
I enjoyed a few hours of music again, but now with the μController regulating the fixed bias. More about that later…