The Dynaco circuit decouples them with 20uF/22K between supply points A and B. It might solve yours, hopefully so.
I incorporated an RC decoupling network between the VA and PI stages, and retained the AC coupling between the two stages.
THD is reduced by 4X (a lot).
Gain is also reduced by about half.
Operating points chosen for the VA are oriented more towards linearity and wider bandwidth, as opposed to as-much-gain-as-you-can-get.
(6BL8 and 6U8A are near equivalents. You could drop a 6U8A in and that would work too. I don't have a SPICE model for 6GH8A, but 6BL8 and 6U8A characteristics are pretty close to that.)
The Scott LK72 circuit still looks pretty good to me. Higher gain than what I proposed, but higher THD comes with that. The LK72 cathodyne is a well thought out circuit, though.
--
THD is reduced by 4X (a lot).
Gain is also reduced by about half.
Operating points chosen for the VA are oriented more towards linearity and wider bandwidth, as opposed to as-much-gain-as-you-can-get.
(6BL8 and 6U8A are near equivalents. You could drop a 6U8A in and that would work too. I don't have a SPICE model for 6GH8A, but 6BL8 and 6U8A characteristics are pretty close to that.)
Yes, that is quite possible. It might still be oscillating at infrasonic frequencies, but you can't hear them, so you may not notice them. However, the extra power required by the circuit to try to amplify them would negatively impact linearity and power available for amplifying music signals.
- Cap voltage rating: I'd say 400V minimum (it should be able to withstand a short from B+ to 0V).
- Cap value: I don't think going larger in value will be a problem. The cap value is chosen in combination with the screen grid resistor value. The combination of the screen grid load resistor and its bypass cap (screen grid to cathode) will create a high-pass filter. That capacitor's job is to bypass the screen down to 1Hz or lower. Too small a value and low frequency response may be rolled off. Too large a value and it will take a long time for that cap to charge up (very long time constant).
The Scott LK72 circuit still looks pretty good to me. Higher gain than what I proposed, but higher THD comes with that. The LK72 cathodyne is a well thought out circuit, though.
--
Last edited:
The problem with the values chosen by Dynaco for that RC network is that it was designed for current-starved operation, so the voltage drop won't be too severe. If you bias the 6GH8A pentode into a more linear area of operation (with about 2.5mA to 3mA plate current), that 22k resistor will drop a significant number of volts. They didn't have 100uF or 220uF 450 rated electrolytic capacitors back then, so they had to use larger resistor values to get low enough time constants.
Nowadays we can get around that by using a smaller value of R with a correspondingly larger value of C to arrive at the same time constant, but with less voltage dropped. \
22k and 20uF could be replaced by 2.2k and 220uF and the filter time constant would be exactly the same, while the voltage drop would be much less.
4.7k and 100uF would have a similar time constant.
If you really want to solve that issue, you could use a capacitance multiplier using a MOSFET. But that's veering off topic...
Last edited:
Now since I do not have a 100uF 450V cap I might try the version without the interstage coupling cap and see what I get, but I'll make it to where I can easily add the cap and resistor later.
Digging into my simulation collection I found something similar, a 6U8 driving 4 EL84s. I replaced the output tube4s with 6V6s, adjusted the bias, this is what it looks like (usual caveats about simulations):
I've got all the resistors for the circuit without the interstage decoupling.
Will wire it up tonight.
Honestly if I could figure a way to get a proper feedback signal from the output transformer I'd wire the amp up with a balanced input. The only way I can figure out how to do it is to use an Edcor WSM15K/15K balanced/unbalanced to balanced/unbalanced transformer, but that is introducing another transformer to the mix.
That said how important is feedback if the amp is balanced from input to output?
The Edcor isn't balanced from the 4 ohm tap to the 16 ohm tap and the 4 ohm tap to ground, however I could possibly use pots to make the signal balanced or would that not work as expected? I'd then have to work out compensation which might be difficult given the tests I've done so far have proven the output taps don't all present a good square wave.
To produce the balanced signal I'd likely use one channel of the DRV134 board.
I wonder if it's possible to add the feedback to the DRV34 board?
Will wire it up tonight.
Honestly if I could figure a way to get a proper feedback signal from the output transformer I'd wire the amp up with a balanced input. The only way I can figure out how to do it is to use an Edcor WSM15K/15K balanced/unbalanced to balanced/unbalanced transformer, but that is introducing another transformer to the mix.
That said how important is feedback if the amp is balanced from input to output?
The Edcor isn't balanced from the 4 ohm tap to the 16 ohm tap and the 4 ohm tap to ground, however I could possibly use pots to make the signal balanced or would that not work as expected? I'd then have to work out compensation which might be difficult given the tests I've done so far have proven the output taps don't all present a good square wave.
To produce the balanced signal I'd likely use one channel of the DRV134 board.
I wonder if it's possible to add the feedback to the DRV34 board?
Last edited:
Rongon
I like your schematic from post # 122 very much , but IMHO coupling cap C3 (0,68uF) is with way to big value , in that position 0,1uF is more appropriate since P.I. 1M grid resistor effective value is actually much bigger due to positive feedback from P.I. cathode ,so effective VAS loading is very low ,
also VAS g2 decoupling cap (C1=4,7uF) is with way to big value ...
I like your schematic from post # 122 very much , but IMHO coupling cap C3 (0,68uF) is with way to big value , in that position 0,1uF is more appropriate since P.I. 1M grid resistor effective value is actually much bigger due to positive feedback from P.I. cathode ,so effective VAS loading is very low ,
also VAS g2 decoupling cap (C1=4,7uF) is with way to big value ...
I do have an Edcor transformer I can use to convert the unbalanced signal to balanced, however I still have the problem of how to add feedback unless I can do it just from the plates of the output tubes to the cathodes of the dual triodes.
That value of 0.68uF is there because the OP has it there now. I agree with you that 0.1uF would be more appropriate. I don't know if the OP has any 0.1uF caps in his possession.
NFB is needed in a pentode amplifier to lower output impedance in order to provide sufficient damping and control over the speaker voice coil and crossover network, reduce THD to reasonable levels, and bring gain down to sane levels. It has nothing to do with unbalanced vs. balanced operation.
So much magic has been attributed to balanced operation. When people fully understand what it does, it becomes less like magic and more like plain old engineering.
The currently discussed pentode VA+triode PI has a single-ended (unbalanced) input only.
A differential amplifier with balanced inputs would be more appropriate for balanced operation.
Applying NFB becomes more complicated with a balanced input, but many here have used local feedback from the plates of the output tubes to the grids of the output tubes, leaving the OPT out of the NFB loop. But now we're going waaaaay off-topic. (Look up "Schade" feedback.)
The RCA RC-20 amplifier design uses a 12AU7 cathodyne to drive a pair of 6AU6 pentodes in a differential pair driving a quad of 6V6s, with NFB taken from the plates of the 6V6s to the cathodes of the 6AU6 diff pair.
That is a more complicated amplifier than what you have. Would require two more tube sockets too.
One thing good engineering requires is that before you start designing things, you decide what your design goal is.
What is it that you actually want from this circuit?
Last edited:
Got some very good news.
I built this circuit.
And got this very pretty 10kHz square wave.
I needed very little compensation only requiring 5pF from the lower 6V6 grid to the feedback connection and 47pF in series with a 500k pot to ground from the junction of the .47uF cap and 1K resistor.
The 500k pot was tweaked until the amp was flat to 20kHz.
I had a slightly better square wave with just the initial overshoot only, but that was without the 47pf cap and just a clip lead connected there which I knew wouldn't work unless I kept the amp on its side and never moved the lead. Perhaps I needed slightly more than 5pF of capacitance. If not for the cap being connected to a plate I'd connect a 1.5-20pF variable cap there and tweak it.
The 3k resistor in series with the 50k gain control came about as I needed to span the 2" distance between the gain control and pin 7 of the 6GH8. I found that 3k is the minimum resistance I can use and the amp stay stable on the 4 8 and 16 ohm taps. That way I cannot make the amp go unstable by adjusting the gain pot to 0 ohms.
The amp is finished for now until I can find the Acrosound transformer at which point I'll install it and re-compensate, however the squarewave looks so good that I might want to leave well enough alone and just enjoy the amp.
Also with no signal being fed to the amp it is dead quiet.
A big thanks to all who have helped with this amp. Without y'all I wouldn't have gotten it right.
I built this circuit.
And got this very pretty 10kHz square wave.
I needed very little compensation only requiring 5pF from the lower 6V6 grid to the feedback connection and 47pF in series with a 500k pot to ground from the junction of the .47uF cap and 1K resistor.
The 500k pot was tweaked until the amp was flat to 20kHz.
I had a slightly better square wave with just the initial overshoot only, but that was without the 47pf cap and just a clip lead connected there which I knew wouldn't work unless I kept the amp on its side and never moved the lead. Perhaps I needed slightly more than 5pF of capacitance. If not for the cap being connected to a plate I'd connect a 1.5-20pF variable cap there and tweak it.
The 3k resistor in series with the 50k gain control came about as I needed to span the 2" distance between the gain control and pin 7 of the 6GH8. I found that 3k is the minimum resistance I can use and the amp stay stable on the 4 8 and 16 ohm taps. That way I cannot make the amp go unstable by adjusting the gain pot to 0 ohms.
The amp is finished for now until I can find the Acrosound transformer at which point I'll install it and re-compensate, however the squarewave looks so good that I might want to leave well enough alone and just enjoy the amp.
Also with no signal being fed to the amp it is dead quiet.
A big thanks to all who have helped with this amp. Without y'all I wouldn't have gotten it right.
Hopefully that circuit works out to be a keeper.
Just one thing... It may have been a drawing mistake only, but you want to make sure you have a decoupling capacitor from the 3k ohm decoupling resistor to ground. It's circled in red below:
47uF would be the minimum value. 100uF would be a little better, but might not make a huge difference.
400V would be the minimum DC rating for that capacitor. 450V would be better (safer).
Just one thing... It may have been a drawing mistake only, but you want to make sure you have a decoupling capacitor from the 3k ohm decoupling resistor to ground. It's circled in red below:
47uF would be the minimum value. 100uF would be a little better, but might not make a huge difference.
400V would be the minimum DC rating for that capacitor. 450V would be better (safer).
So far so good. Playing some jazz and the amp sounds good. I'm thinking it sounds slightly better in the upper treble as well.
The only other thing I really need to do is get a matched quad of 6V6 tubes. I'll then move one 120k grid resistor to one of the bias pots as I will no longer need one pair of 6V6 tubes at a different bias point.
That might help the amp even more.
What I might do this weekend is do a 10kHz squarewave test using 8 and 16 ohm non-inductive loads and I'll also do a test with a 20kHz sinewave to see if the amp is still flat to 20kHz.
I'd love to do a gain chart of 20-20kHz, however that would require two RMS voltmeters that are good to 20kHz which I don't have.
The way I do that is I do 20Hz to 100Hz in 10Hz steps, 200Hz to 1kHz in 200Hz steps then 2kHz to 20kHz in 2kHz steps.
I record the input voltage and output voltage at each step and then calculate the gain.
The only other thing I really need to do is get a matched quad of 6V6 tubes. I'll then move one 120k grid resistor to one of the bias pots as I will no longer need one pair of 6V6 tubes at a different bias point.
That might help the amp even more.
What I might do this weekend is do a 10kHz squarewave test using 8 and 16 ohm non-inductive loads and I'll also do a test with a 20kHz sinewave to see if the amp is still flat to 20kHz.
I'd love to do a gain chart of 20-20kHz, however that would require two RMS voltmeters that are good to 20kHz which I don't have.
The way I do that is I do 20Hz to 100Hz in 10Hz steps, 200Hz to 1kHz in 200Hz steps then 2kHz to 20kHz in 2kHz steps.
I record the input voltage and output voltage at each step and then calculate the gain.
Last edited:
12AX7 and DN2540 concertina give great results , gain of 70 at each output.
500mV should be enough to drive 6V6 outputs to full power.
500mV should be enough to drive 6V6 outputs to full power.
I set the gain to where 1V equals full power of 25 watts as the Magnavox amp was set to an input sensitivity of 1V stock.
Also my Schiit Saga S preamp given it uses relays for the volume control doesn't go to full zero output at the minimum volume control setting so if I set the gain to where 500mV is full output the audio would still come through the speaker at an easily audible level.
With 1V input sensitivity I just can hear the audio from the speaker at a real quiet level which is a real comfortable level for when I'm listening to music while going to sleep.
Also my Schiit Saga S preamp given it uses relays for the volume control doesn't go to full zero output at the minimum volume control setting so if I set the gain to where 500mV is full output the audio would still come through the speaker at an easily audible level.
With 1V input sensitivity I just can hear the audio from the speaker at a real quiet level which is a real comfortable level for when I'm listening to music while going to sleep.
Last edited:
Decided to do a test with the 10kHz square wave at 4 8 and 16 ohm loads. Signal measured across the load. Here's the results.
16 ohm load.
8 ohm load.
4 ohm load.
I also moved the bias connection of the other pair of 6V6 so that I now have a single bias adjust pot. That required swapping 6V6 tubes around until I got the cathode currents as equal as possible within reason.
16 ohm load.
8 ohm load.
4 ohm load.
I also moved the bias connection of the other pair of 6V6 so that I now have a single bias adjust pot. That required swapping 6V6 tubes around until I got the cathode currents as equal as possible within reason.