Hi guys, Yesterday I've finished building the 6336 SET amp based on this website:
https://www.cascadetubes.com/the-6336-set-amp/
After that I did some testing on the amp I found some strange wiggling on 1W output frequency response graph at around 14 - 30 Hz:
What can cause this wiggling? BTW is it even something to get worried about?
Here is some pictures from square wave responses and also power vs. THD if they can help:
Thanks
https://www.cascadetubes.com/the-6336-set-amp/
After that I did some testing on the amp I found some strange wiggling on 1W output frequency response graph at around 14 - 30 Hz:
What can cause this wiggling? BTW is it even something to get worried about?
Here is some pictures from square wave responses and also power vs. THD if they can help:
Thanks
1. Do you test on a resistive load?
2. Does this happen with generator only test of the output transformer, excluding the amplifier stage?
2. Does this happen with generator only test of the output transformer, excluding the amplifier stage?
Hi 50AE, thanks for answering, Yes I'm using power resistors in series to get 8 ohms nominal load.
No, I tested the OPT before and the response was smooth.
In this case we can suspect the amplifier stage. What about, if cathode bias, RC values? Also, power supply decoupling and impedance vs frequency?
Is the behavior dynamic at different power levels? 0.25, 0.5 and 2W for example.
I'd suspect some LF filtering action in the power stage with two mixed poles.
Try doing an unloaded frequency response for comparison. That would ease the loading of the power supply, depending on the primary inductance value. If you're afraid of open secondary test, use a high value resistor, 100R for example. Although with a triode, unloaded stage without cutoff mode should be no problem.
Is the behavior dynamic at different power levels? 0.25, 0.5 and 2W for example.
I'd suspect some LF filtering action in the power stage with two mixed poles.
Try doing an unloaded frequency response for comparison. That would ease the loading of the power supply, depending on the primary inductance value. If you're afraid of open secondary test, use a high value resistor, 100R for example. Although with a triode, unloaded stage without cutoff mode should be no problem.
Last edited:
The Edcor GXSE15-8-2.5K lacks sufficient magnetic flux to support a waveform below 20Hz. The wiggle at 30Hz is nothing significant, the small inperfection at 14 Hz is interesting.
Any such change in the magnitude response must be matched with an appropriate phase deviation. Since the phase deviation is relatively clean (only a small deviation at 14Hz) but the magnitude response is not (a transition which approaches a small level passband interval around 14Hz), there is something unexplained in the data. Such a magnitude change would require a pole in close proximity to the jω axis around 14Hz. This would cause a significant phase deviation which in not present in the plot.
I would retake data manually in 2Hz increments between 10Hz and 30Hz and replot that data. I suspect the mathematics of the sweep generator program you are using. It is doing something that not quite right to produce such mismatched plots.
Would like to see the schematic of the amp including the power supply which will help us to see what could be the problem. Also, it is better to monitor the B+ voltage with oscilloscope and see if it is stable when operating at 20Hz.
Thanks again for the reply, here is a link to the schematics: https://www.cascadetubes.com/the-6336-set-amp/
I also tested Freq. Response at 0.1W, 0.5W, 2W, 4W and 6W max and as you can see wiggling gets a little better with higher powers:
Last edited:
I suspect sooner roll-off due to saturation at higher power level might be masking the "wiggling".
Thanks for the reply, Actually I'm using my own handmade transformer which here is its Freq. Response at 1W:
Exactly, I too suspect my Focusrite 2i2. Could it be the line out of the soundcard causing this issue?
Yes that's exactly my thought on this. Also the other thing that you said about unloaded frequency response; I will do some testing tomorrow, but if it does work, what does it say?
Different wiggling under less loaded conditions might indicate problems from the power supply, because less current will be drawn from it. That can change the Q from potential resonances of the power supply LC filter networks with audio band cutoff. Let's see the power supply schematic? However, this thesis will be somehow limited by the amount of primary inductance.
I believe 12AU7 at 1.6 mA 300 V is inadequate driver for 6336. Low driver current limits output slew rate, which shows in 10K square wave. Low driver stage plate voltage limits undistorted signal output, causing relatively high THD. I would use a dissimilar triode like 6CS7, low power section voltage amplifier as of schematic and high power section as driver running at 360 V B+ 15 mA, choke-loaded.
Yes you are right about the slew rate, it's definitely better to use higher current in at the gate of 6336 to cope with the input capacitance and improve the slew rate, but my main issue right now is the low F instability that is appearing at around 14-30 Hz.
Correct me if I'm wrong, but I don't know that it's "unstable." Thta term indicates imminent oscillation, which I doubt is a problem here. It's a bit lumpy, but only by a few fractions of a dB. There's no feedback, so I doubt there's a stability issue, but you could test with a pulse and see how it recovers. I'm not sure I would worry about it, personally. Probably an artifact of the output transformer.
I tested the amp with 100 ohms load and here is the results compared to 8 ohms load:
So could it be the power supply causing the problem? The 10H choke plus 47u reservoir cap resonate at around 7.3 Hz if my calculations are not wrong, but how can I test if it's the resonance causing the issue?
How about the second LC tank, where L=1H and C=100u? Try increasing C to 400u and measure again. You could also increase choke Rdc.