Here's the interface I used. https://www.diyaudio.com/community/...urements-with-rew.338511/page-65#post-6972813
Pretty good for ~$25 used. I doubt its noise floor would present any issue while measuring a pedestrian tube amp - like the ones I have.
I have mine running on a W10 machine. I've abused the tar out of it and it keeps on; big loop currents through the ground, I even looked at a 350V screen supply through a cap and potentiometer in front of it. Has stereo output, mono input, so I can make a measurement in REW, close that, switch from resistor loads to speakers and listen via Foobar 2000 pretty quickly.
Just bought a spare, in case I finally do blow it up. Or wreck this one putting film caps on its output, BNC instead of the 1/4 phone, input attenuation, whatever. Right now I just use a trim pot, dial it back from clipping the input and measure; most of what I'm interested in is all relative. % THD, FR etc.
Finally, I got around to exploring what changing the feedback resistor value does, as I did change the 1st 12AX7 stage plate resistor from 470k to 390k.
Beginning with 6.8k as the original value, I stepped up and down 1k at a time, looking at measured THD at 1kHz while correcting for output level changes (keeping a consistent 8V output into 8 Ohms) due to changing the gain of the amp; of course needing more drive with more feedback.
As I increased feedback by lowering the feedback resistor value, distortion went up and as I decreased feedback by raising the resistor value, distortion went down. Ya, I understand that. The changes were miniscule, maybe doubling the 1% initial value at 4.8k, diminishing to a large fraction thereof the other way.
I just soldered the 6.8k back into place. I can take a hint... I was hoping to see distortion go down until the amp turned into a phase shift oscillator, but, obviously I understand nothing about it.
Beginning with 6.8k as the original value, I stepped up and down 1k at a time, looking at measured THD at 1kHz while correcting for output level changes (keeping a consistent 8V output into 8 Ohms) due to changing the gain of the amp; of course needing more drive with more feedback.
As I increased feedback by lowering the feedback resistor value, distortion went up and as I decreased feedback by raising the resistor value, distortion went down. Ya, I understand that. The changes were miniscule, maybe doubling the 1% initial value at 4.8k, diminishing to a large fraction thereof the other way.
I just soldered the 6.8k back into place. I can take a hint... I was hoping to see distortion go down until the amp turned into a phase shift oscillator, but, obviously I understand nothing about it.
I get really scared seeing G1 to cathode idling voltages so far under a Volt. In this range, the grid is thinking really hard about maybe an operation, and becoming an anode. If the input signal gets significantly big compared to that diode voltage, it can matter. Varies a lot with source impedance, natch.
All good fortune,
Chris
All good fortune,
Chris
There isn’t a ridiculous amount of open loop gain to start with - only one gain stage, not two. The output stage plus trafo is low gain (maybe unity to about 2, depending on bias). This is going to limit how much distortion correction you can get. You’re already getting less -a lot less- closed loop gain than just the feedback resistor ratio (which you would get with gain approaching infinity). With one more 12AU7 stage thrown in there, ala Williamson, that closed loop gain would go up closer to the theoretical 31.9 and you would have more distortion correction available. And more stability concerns.
About the G1 to K voltages being so low, it’s par for the course with AX7’s. And you can drive them into forward grid conduction when overdriven. Gate stoppers will limit any potential damage - some designers diode clamp it.
About the G1 to K voltages being so low, it’s par for the course with AX7’s. And you can drive them into forward grid conduction when overdriven. Gate stoppers will limit any potential damage - some designers diode clamp it.
The 6EU7's are identical to the 12AX7's except for filament voltage.
This amp that I built years ago performs like those "high end" amps that sell for rediculous prices.
The result is quite respectable and after years of use/age I never had any problems.
Perhaps it can give you ideas to sort out your own issues.
This amp that I built years ago performs like those "high end" amps that sell for rediculous prices.
The result is quite respectable and after years of use/age I never had any problems.
Perhaps it can give you ideas to sort out your own issues.
Chris, wg-ski, That didnt seem to be the issue with me getting the amp to work at all. The cathode resistor on that 1st 'AX7 stage is easy enough to play with, to get more drop. I was just surprised how I was all worried changing the plate res from 470 to 390K would upset things control loop wise, only to find out it probably didnt do much of a damn thing.
Perhaps I should get some more voltage drop from K to ground on that stage, then try the feedback resistor experiment again. I had to drive it harder to get the same output; maybe distortion went up in the signals driving the output tubes as a result.
WOT, thanks for taking the time to post all that for me. Lots different between yours and the hunk I'm working on; my OPTs arent as good; no UL taps for me. My grid stops on the OP tubes are double yours. I have no idea if that contributed to the eventual red-plate of one of my "matched" Sovtek 6BQ5s. Being that I drove the grid leaks (?) negative while in your amp, you drive the cathodes positive.
Well, I'm doing that now, with a 100 Ohm resistor dropping ~13.8V from taking the current of all 4 tubes. No more red plate - at least not yet. In 3x the time it took to get into that state with my other topology...
Unsure why the circuit would "care" whether the bias is applied as a negative to the grids, versus a positive to the cathodes. How does the tube know the difference, inside looking out?
I also took your advice (and that of others) to use the 6.3V winding as a buck on the AC line in, instead of to generate the negative bias. I've got a 3 wire cord in place, with a solid earth connection to chassis.
My amp design is a classic style, and can be used with non-UL transformers as well.
And no, the output tubes don't care about types of biasing as long as the control grids are negative as opposed to the cathodes.
But the best thing is that with the "fixed bias" design by David Gillespie using the LM337 regulator is that it tracks any instantanious "sag" of the power supply under heavy volume, and keeps the grids in their proper negative voltage range - which results in no annoying distortion resulting from the grids being driven positive.
I get a clean 17 watts/channel from those tubes.
Of course the transformers do help.
My amp performs solidly for years now.
And I've no need for a 3-wire power cord, the amp runs dead silent, no hum, no hiss.
Cathode bias gives the output valves some negative feedback for their idling current - if the idling current drifts higher, the larger voltage drop across the cathode resistor biases the valve(s) more negative, partially correcting the drift. If a single cathode resistor is biasing all four valves, a single valve drifting higher has proportionally less correction, so separate cathode resistors is a safer plan.
The potential issue with large value grid leak resistors is with residual G1 current, which is in the direction of positive feedback for bias stability - G1 current makes the G1 more positive, so the valve draws more current, G1 gets warmer, etc. The solution is to keep grid leak resistance to a suitably low value for modern production valves, with modern production standards of metallurgy and vacuum (not great), but with ancient ideas of B+ and esp. G2 voltages.
Restoring these old parties involves a fair amount of re-engineering whether you've wished to, or not. But can be done.
All good fortune,
Chris
The potential issue with large value grid leak resistors is with residual G1 current, which is in the direction of positive feedback for bias stability - G1 current makes the G1 more positive, so the valve draws more current, G1 gets warmer, etc. The solution is to keep grid leak resistance to a suitably low value for modern production valves, with modern production standards of metallurgy and vacuum (not great), but with ancient ideas of B+ and esp. G2 voltages.
Restoring these old parties involves a fair amount of re-engineering whether you've wished to, or not. But can be done.
All good fortune,
Chris