A 10kHz squarewave shows the amp stability is worse than it was with the other circuit and if I place a cap large enough to start making the squarewave better, I see more instability like the amp is just about to break into oscillation. Maybe that's why the compensation cap went from the plate to the 47 ohm resistor in the amp the driver and phase splitter was taken from as otherwise the amp would be more unstable with the compensation cap across the feedback resistor. Also placing the compensation cap (tried 390pF and 195pF) where it was between the 1uF caps just caused the amp to oscillate and with a squarewave at 10kHz even 47pF caused the squarewave to show the amp had become more unstable, however I don't think it should have and don't know why it did.
Also the amp is no longer flat to 20kHz as without any compensation the output is slightly lower than it is at 1kHz. Not sure what the problem is there as I don't see anything that can be causing the slight drop in output unless it's something in the driver and phase inverter stage.
The very low frequency oscillation is still there at times as well and might be the key to the other issues I'm having.
The only thing I can see that might be affecting the HF response is the 270k resistor unless miller capacitance doesn't come into play with a split load phase inverter.
The amp not being flat to 20kHz is pretty much a dealbreaker to me especially since I know the amp was flat to 20kHz before I changed the driver and phase inverter.
Could I need to lower the 1 meg screen resistor or is it fine as is since I basically duplicated the circuit posted?
Also the amp is no longer flat to 20kHz as without any compensation the output is slightly lower than it is at 1kHz. Not sure what the problem is there as I don't see anything that can be causing the slight drop in output unless it's something in the driver and phase inverter stage.
The very low frequency oscillation is still there at times as well and might be the key to the other issues I'm having.
The only thing I can see that might be affecting the HF response is the 270k resistor unless miller capacitance doesn't come into play with a split load phase inverter.
The amp not being flat to 20kHz is pretty much a dealbreaker to me especially since I know the amp was flat to 20kHz before I changed the driver and phase inverter.
Could I need to lower the 1 meg screen resistor or is it fine as is since I basically duplicated the circuit posted?
Last edited:
Moving the feedback connection to the 4 ohm tap helped a lot.
With a 47pF cap from the bottom 6V6 plate to the 47 ohm resistor the 10kHz squarewave looks much nicer and the output sinewave at 20kHz is now a little higher so I know that I need a little more than 47pF of capacitance, however adding more increased the output slightly at 20kHz. So maybe I need to put a small cap across the feedback resistor.
With a 47pF cap from the bottom 6V6 plate to the 47 ohm resistor the 10kHz squarewave looks much nicer and the output sinewave at 20kHz is now a little higher so I know that I need a little more than 47pF of capacitance, however adding more increased the output slightly at 20kHz. So maybe I need to put a small cap across the feedback resistor.
Have you calculated how much negative feedback you're applying? It takes a really fine OPT to withstand 18db or 20dB of NFB. Most can only be made stable with 12dB of NFB.
To calculate the level of NFB:
1. Disconnect the feedback resistor (run the amp 'open loop').
2. Connect a 4 ohm 10W resistor load across the OPT 4 ohm secondary and ground.
3. Turn on the amp.
4. Apply a 10mV 1kHz sine wave to the input.
5. Measure the AC voltage across the 4 ohm 10W load resistor.
Let's say the output voltage is 10X the input voltage, or 100mV. That means the amp yields 20dB (10X) gain from in to out.
5. Turn off the amp.
6. Reconnect the feedback resistor.
7. Turn on the amp.
8. Measure the AC voltage across the 4 ohm 10W resistor with the feedback connected (closed loop).
Now let's say the output voltage is 3X the input voltage, or 30mV. That would mean the amp supplies 9.5dB (3X) gain.
Since the closed loop gain is 11.5dB less than the open loop gain, we'd say that you had applied 11.5dB of negative feedback.
What happens if you do this with your current amp circuit?
- Pentodes have less Miller capacitance than triodes. However, if you have the pentode biased to make over 100X gain, the Miller capacitance could still be considerable.
- If you have a lot of NFB applied, the amp could be resonating and/or oscillating due to phase characteristics and poles and zeroes of the OPT, and how those might interact with poles and zeroes of the driving circuitry. Compensation capacitors will then be necessary, or you will need to employ less negative feedback to get the circuit to run stable enough at all frequencies.
The falling high frequency response could be due to a pentode's very high plate resistance (output impedance).
Have you compared the voltages of the pentode plate and the split load inverter's cathode?
Is the plate voltage at least 1.5V lower than the split load inverter's cathode?
If the split load inverter is zero-biased, it would be drawing grid current, and all sorts of bad things could be happening.
Also, I'd recommend putting a 1k ohm resistors in series before the phase splitter's grid. If it's oscillating then that should fix it.
That's a compensation network.
Applying lots of negative feedback to a tube amp with multiple time constants and the (imperfect) output transformer gets complicated.
--
Last edited:
Not sure how much gain I have, but I set it to where it's 1V in for 25 watts out which is where it was set at for the other phase inverter.
Looks like the amp wants 247pF where the 47pF cap is as I put a 200pF cap in parallel with the 47pF cap and the 10kHz squarewave looks better.
I also forgot the cap to ground on pin 9. That should take care of the rising response closer to 20kHz. Gotta figure what cap value will work there.
My line voltage varies some so I have the amp running on a variac with a DMM measuring the AC voltage and I keep it set to 120Vac for all testing.
I tried two 390pf caps and a 180pF cap in series for 93.6pF and it looks like the amp wants a little less than that on pin 9 to ground.
Looks like the amp wants 247pF where the 47pF cap is as I put a 200pF cap in parallel with the 47pF cap and the 10kHz squarewave looks better.
I also forgot the cap to ground on pin 9. That should take care of the rising response closer to 20kHz. Gotta figure what cap value will work there.
My line voltage varies some so I have the amp running on a variac with a DMM measuring the AC voltage and I keep it set to 120Vac for all testing.
I tried two 390pf caps and a 180pF cap in series for 93.6pF and it looks like the amp wants a little less than that on pin 9 to ground.
Last edited:
I goofed on the schematic. Did the feedback wrong.
Here's what a 10kHz squarewave looks like now. Looks like that may be as good as I can get it unless I use a better output transformer. On another forum it was recommended to use the Acrosound TO-330 and that the Edcor output transformers aren't as good as people like to think.
This is how it looked with the long tailed pair. One thing I noticed with the current phase splitter is the top and bottom half of the squarewave are a lot more similar than they were with the long tailed pair.
Here's what a 10kHz squarewave looks like now. Looks like that may be as good as I can get it unless I use a better output transformer. On another forum it was recommended to use the Acrosound TO-330 and that the Edcor output transformers aren't as good as people like to think.
This is how it looked with the long tailed pair. One thing I noticed with the current phase splitter is the top and bottom half of the squarewave are a lot more similar than they were with the long tailed pair.
Last edited:
That seems part of the biasing circuit. I have never seen that used before.
Would it be worthwhile to try it to see if the squarewave performance improves? I can make a 300k resistor which should be close enough.
I wonder if I lower the screen grid resistor would performance improve. There's only one way to find out.
Here's how the Scott LK-72 does the same circuit.
I wonder if I lower the screen grid resistor would performance improve. There's only one way to find out.
Here's how the Scott LK-72 does the same circuit.
Last edited:
0.435V on the cathode of the 6GH8A pentode seems awfully low. That's very close to zero. The pentode might be running with appreciable grid current. No bueno.
That 6GH8A pentode is running at a very cold operating point. 0.435V/667 ohms = 652uA (0.652mA) plate current.
Also, the 6GH8A triode is running with only about 42V plate-to-cathode (p-k).
Each 50k plate/cathode load resistor is dropping 87V, so the plate current of the 6GH8A triode is only 1.7mA.
I think the reason the 6GH8A is running at such 'current starved' levels is that the Dynaco driver stage was designed to run on something like a 400V B+.
This circuit used in this 6V6 amp has only 217V B+ running to it. (What's draining away all that voltage?)
It looks like you can't simply cut and paste this Dynaco circuit into this 6V6 amp. The pentode and triode will need to have their operating points adjusted for the much lower available B+ voltage.
I would do that and then test the voltages open loop (before adding feedback).
Make sure the voltages are correct and the driver stages are working optimally.
Then add negative feedback and test for stability.
One step at a time.
What happened when you tried the circuit in post #49?
That 6GH8A pentode is running at a very cold operating point. 0.435V/667 ohms = 652uA (0.652mA) plate current.
Also, the 6GH8A triode is running with only about 42V plate-to-cathode (p-k).
Each 50k plate/cathode load resistor is dropping 87V, so the plate current of the 6GH8A triode is only 1.7mA.
I think the reason the 6GH8A is running at such 'current starved' levels is that the Dynaco driver stage was designed to run on something like a 400V B+.
This circuit used in this 6V6 amp has only 217V B+ running to it. (What's draining away all that voltage?)
It looks like you can't simply cut and paste this Dynaco circuit into this 6V6 amp. The pentode and triode will need to have their operating points adjusted for the much lower available B+ voltage.
I would do that and then test the voltages open loop (before adding feedback).
Make sure the voltages are correct and the driver stages are working optimally.
Then add negative feedback and test for stability.
One step at a time.
What happened when you tried the circuit in post #49?
Missing the pin #9 filter network. Then, for GNFB a phase cap is usually added across the primary FB resistor to shape up the square wave. There are still different FB line options and cathode resistor choices without splitting the resistor and just running the FB straight to the K.
Actually, if you're not going to bypass the pentode stage Rk, you might as well insert the FB straight to the pentode's cathode and get rid of that 47R.
However, the value of the pentode Rk will need to be changed in order to bias the pentode correctly (along with the Rp and Rg2 load resistors).
I still think both tubes in the 6GH8A are not biased correctly with the values shown and with such a low B+ voltage.
Why did the B+ voltage sag to only 217.1V? It was 275V before. What happened?
When you design a circuit, you will encounter problems to be solved. Often solving one problem will cause a new problem. Then you have to balance the benefits/deficits of the changes you made to get the best possible result given the limitations of your circumstances. Everything ends up being a compromise.
What were the problems when you tried the circuit in post #49?
However, the value of the pentode Rk will need to be changed in order to bias the pentode correctly (along with the Rp and Rg2 load resistors).
I still think both tubes in the 6GH8A are not biased correctly with the values shown and with such a low B+ voltage.
Why did the B+ voltage sag to only 217.1V? It was 275V before. What happened?
When you design a circuit, you will encounter problems to be solved. Often solving one problem will cause a new problem. Then you have to balance the benefits/deficits of the changes you made to get the best possible result given the limitations of your circumstances. Everything ends up being a compromise.
What were the problems when you tried the circuit in post #49?
Last edited:
It's connected there I think. I just drew it wrong.
I couldn't get proper operation with the 12AX7 and better operation, but not enough gain with the 12AT7. The 12BZ7 worked, but the higher THD it was mentioned as having was a no-go for me.
Tried adding a cap there and it made the 10kHz square wave look worse.
The power supply resistor feeding that stage is 50k and the other phase splitter likely provided a much lighter load. I can maybe use a 1k resistor in place of the 50k resistor to get more B+ voltage. It's likely the low B+ is why I had such an issue with the compensation. I can get the B+ up to 339V though.
That's usually the case, however trying that just introduced more stability which is likely why the compensation cap was connected to the screen grid in the original circuit.
It also takes a resistor in series to create the correct rolloff point for the filter.
Your square waves don't look as bad as I think you feel they are. There is very little ringing and the bit of overshoot can be tempered by a phase cap across the FB resistor. But it's the amount of FB you need to get a handle on. Start with a modicum and add as it can hold it.
I'll up the B+ to that stage today and see where I am.
The pip at the end of the waveform might be due to the low B+.
Tried that and it made the squarewave worse with the output sinewave down a little at 20kHz. Suspect the low B+ is causing all my issues given the circuit was designed to operate on a higher B+ and with that higher B+ it's likely I'll need different compensation cap values and maybe will then be able to add a cap and resistor from pin 9 to ground.
The nice thing is about a pentode input stage is if I later decide to drive the amp with a tube preamp I can increase the value of the 47k resistor without miller capacitance issues.
Here's the corrected schematic.
Last edited:
You couldn't get proper operation in what way, exactly? What exactly happened?
Were you DC coupling the plate of the first 12AX7 to the cathode of the second 12AX7? That won't work with only a 275V B+. You'll need to AC couple the two stages so they can be biased independently of each other.
Is this the circuit you tried? Or did you change it?
The following is a collection of random thoughts and suggestions. Disregard if you wish.
I tweaked the 12AX7 circuit to get this...
The power supply decoupling R was reduced from 50k down to 22k.
The decoupling cap remained at 33uF.
B+ should then be about 298V to the 12AX7s.
Gain looks good. The simulation predicts 10V peak output with only 190mV of input.
THD looks pretty good. With 10V peak output, the simulation predicts THD will be 0.13% into those 100k grid leak resistors for the 6V6s. Second harmonic dominates.
Plate current looks good. 868uA for the first stage, 1mA for the phase splitter.
Operating points look good. Grid bias is -1.28V on the first stage, almost -2V on the phase splitter.
That should work.
One problem pentodes have is that they have almost no PSRR (Power Supply Rejection Ratio) at all. The plate supply has to be squeaky clean, or you'll get power supply noise in the pentode's output.
The 12AX7, being a high mu triode, will have much better PSRR. The 12AX7 as a voltage amp with its Rk bypassed should have good PSRR, perhaps -40dB. If the power supply filtering knocks the 120Hz ripple down by only -60dB, you still have power supply noise from the 12AX7 of only -100dB. That should be adequately quiet.
Pentodes take more work to reach a good design.
You need to choose the correct values for the plate load resistor (Rp) and screen grid resistor (Rg2), and you have to make sure the plate voltage and current and screen grid voltage and current are in their correct proportions for the pentode you've chosen.
You need to bypass the screen well, down to -3dB at 1Hz, or lower.
You need to make sure the gain isn't set so high that the high frequencies roll off within the audio band. That's because pentodes have high output capacitance and high output impedance. Fortunately, that shouldn't be a problem with proper choice of operating points because in this case the pentode is driving a cathode follower, which presents a light load.
--
Last edited:
I had an initial issue awhile back where there was slight hum in the speaker and I placed a cap across the choke to resonate it at 120Hz and that dropped the ripple voltage to a real low value with there being virtually no ripple at the filter cap for the driver phase inverter stage. Confirmed last night by placing my ear to the speaker with the amp on and hearing dead silence.
That's the 1 meg resistor and .47uF cap.
Interestingly enough I had a very low frequency (under 10Hz I believe) that would occur sometimes and gradually go away. At some point it stopped doing that. Maybe the lower B+ caused that instability?
I'll also try that 330k resistor as shown on the schematic I got the circuit from just to see what if anything it does.
Last edited:
Based on what I've found the voltage at point A on the ST-70 schematic is 271Vdc. The B+ I now have is 325.6Vdc so I need to drop it down some.
Do you?
The Stereo 70 manual (https://www.thehistoryofrecording.com/Manuals/DynaCo/Dynakit_ST70.pdf) shows +305V DC at point 'A'.
With tubes, at least generally speaking and within any device's max limits, a higher B+ allows the device to swing more volts more cleanly.
Using a decoupling resistor of 15k (or 22k) ohms with a decoupling capacitor of 33uF, what B+ do you get?
The Stereo 70 manual (https://www.thehistoryofrecording.com/Manuals/DynaCo/Dynakit_ST70.pdf) shows +305V DC at point 'A'.
With tubes, at least generally speaking and within any device's max limits, a higher B+ allows the device to swing more volts more cleanly.
Using a decoupling resistor of 15k (or 22k) ohms with a decoupling capacitor of 33uF, what B+ do you get?
Last edited: