When you changed the circuit, you reduced the tail resistance from 20k (10k in series with 10k) to only about 11k (10k in series with 1k, with that 1k shunted by 100k). Reducing the tail impedance will worsen the balance and worsen performance -- gain will go down, THD will go up.
With the 82k and 100k resistors you had to compensate for the imbalance, the outputs from the 12AX7 plates would still be severely unbalanced. Try 91k and 100k.
Merely reducing the voltage drop across the tail resistance by reducing the resistor value may help with biasing the 12AX7s, but it worsens balance by a lot.
Really, the cure for this is to use a CCS in the tail of the LTP and matched (equal) plate load resistors. The ideal would be to have an infinite impedance in the tail, but you only have a meager 20k ohms, which is less than the rp of a 12AX7 by a factor of more than 4. A constant current sink will have an impedance of 500k to 1M ohms. That would be an improvement of over 50X to 100X and you would only need to drop about 50V across the CCS. Win, win.
I showed the NFB connection exactly the same way as in the schematic in post #5. I didn't include the 100k series feedback resistor, though.
The amount of NFB is almost always defined by a voltage divider.
I don't like how the NFB is implemented in the post #5 schematic, but since it's there, I guess you'll want to keep using it. But it doesn't have to be that way...
Last edited:
Using the 82k and 100k resistors the plates were identical in output using the 12BZ7.
Not sure what caused either the rise in output near 20kHz or the dip in output at 10kHz.
Not sure what caused either the rise in output near 20kHz or the dip in output at 10kHz.
Patching the balance by intentionally unbalancing the circuit is a band-aid, not a cure. The balance will vary from tube to tube or as the tubes age.
So with the circuit in post #34 I'd just use a resistor from the 4 ohm secondary to where it shows the NFB connected, right? Do I need a capacitor to isolate the DC voltage?
Yes, for the schematic in post #34 you would need to block DC with a capacitor.
If the value of Rs is 1k ohms, then the value of Rf might be 22k ohms (maybe more, maybe less, depending on how much NFB you want to apply here).
If Rf is 22k ohms, then Cg2 may need to be large, as much as 6.8uF.
If the value of Rs is made to be 10k ohms, the value of Rf could be something like 220k ohms, in which case Cg2 could be 1uF and would be plenty large enough in value.
This will take some finagling.
The other way to apply NFB is the Fender guitar amp style, with a small value resistor placed in series with the tail and negative feedback applied on top of that Rs from the OPT secondary.
I know the first one should work. I'm not 100% positive the second one will work, as I haven't seen this done before with a CCS in the tail. But it should work, as far as I know. The Rs is only adding a little bit of resistance to the tail, and forms the necessary voltage divider with Rf.
You could change U1 and U2 to 12BZ7 and everything else should stay the same.
I haven't put together a sim of this yet. I'll try to get to that soon.
If the value of Rs is 1k ohms, then the value of Rf might be 22k ohms (maybe more, maybe less, depending on how much NFB you want to apply here).
If Rf is 22k ohms, then Cg2 may need to be large, as much as 6.8uF.
If the value of Rs is made to be 10k ohms, the value of Rf could be something like 220k ohms, in which case Cg2 could be 1uF and would be plenty large enough in value.
This will take some finagling.
The other way to apply NFB is the Fender guitar amp style, with a small value resistor placed in series with the tail and negative feedback applied on top of that Rs from the OPT secondary.
I know the first one should work. I'm not 100% positive the second one will work, as I haven't seen this done before with a CCS in the tail. But it should work, as far as I know. The Rs is only adding a little bit of resistance to the tail, and forms the necessary voltage divider with Rf.
You could change U1 and U2 to 12BZ7 and everything else should stay the same.
I haven't put together a sim of this yet. I'll try to get to that soon.
Last edited:
I had that issue even with the 100k and 10k resistors used for feedback.
In a normal long tailed pair phase splitter the .47uF cap and 10k tail resistor would go to ground if feedback was not used.
Where the feedback is used was where someone told me it should go.
Altering the 10k resistor to 1k and the feedback resistor from 100k to 10k didn't affect the balance at all.
Both the schematics in post #40 have equal outputs from the 12AX7 sections as verified by a dual channel scope.
It looks like you're getting an idea of what you want to do with this. That's a good thing!
All I can say is that the second circuit in post #45 will have much better balance and lower distortion than the second circuit in post #46, which is taken straight from an old Fender guitar amp schematic from the late 1950s and was also used in the Fender Blues Junior guitar amplifier.
Those two circuits are almost exactly the same thing, except for the use of the CCS in the tail and the matched plate resistors in the circuit from post #45. They are drawn somewhat differently...
All I can say is that the second circuit in post #45 will have much better balance and lower distortion than the second circuit in post #46, which is taken straight from an old Fender guitar amp schematic from the late 1950s and was also used in the Fender Blues Junior guitar amplifier.
Those two circuits are almost exactly the same thing, except for the use of the CCS in the tail and the matched plate resistors in the circuit from post #45. They are drawn somewhat differently...
Chapter 9 of Merlin Blencowe's "Designing Tube Preamps for Guitar and Bass" contains an excellent overview of designing a long-tailed pair phase splitter, complete with different ways to couple its input (you can AC couple the input as in the circuits mentioned above, or you can DC couple the input to the plate of a common cathode voltage amp stage), as well as two different ways to apply negative feedback from the speaker/OPT secondary. There's also an excellent passage on the effects of different bias points for the LTP, which I found especially helpful.
NB: Merlin Blencowe also wrote the ValveWizard website.
PS - Dang! "Designing Tube Preamps for Guitar and Bass" is out of print! It seems it's not available anywhere. I hope Merlin's planning on a second printing...
NB: Merlin Blencowe also wrote the ValveWizard website.
PS - Dang! "Designing Tube Preamps for Guitar and Bass" is out of print! It seems it's not available anywhere. I hope Merlin's planning on a second printing...
I do see a little imbalance as I increase the signal which is likely about close to where the output stage starts to distort.
It would be awhile before I get the necessary stuff to try a CCS.
In the meantime could I try a split load phase inverter like this as I have the resistors for it.
It would be awhile before I get the necessary stuff to try a CCS.
In the meantime could I try a split load phase inverter like this as I have the resistors for it.
That should work fine.
Gain will be about 27X (per output).
THD looks OK.
Balance will be excellent until the output stage gets driven into grid current (at high power).
The split load inverter is only biased to 890uA (0.89mA), which means it will probably slew limit into any appreciable load capacitance. You're driving 6V6 in Pentode? Ultralinear? Or wired Triode?
You might want to reduce the value of the 56k plate and cathode load resistors to 47k, but that might not make enough difference to be worth the bother. Your call.
Gain will be about 27X (per output).
THD looks OK.
Balance will be excellent until the output stage gets driven into grid current (at high power).
The split load inverter is only biased to 890uA (0.89mA), which means it will probably slew limit into any appreciable load capacitance. You're driving 6V6 in Pentode? Ultralinear? Or wired Triode?
You might want to reduce the value of the 56k plate and cathode load resistors to 47k, but that might not make enough difference to be worth the bother. Your call.
Last edited:
Four 6V6 in push pull parallel pentode.
What needs to be changed to bias the 12AX7 better and would the 12BZ7 perform better with that circuit or will the THD be higher?
Also is the feedback arrangement ok or can it be done better?
If I needed more gain I can increase the 100k plate resistor, right?
Now If I had an extra socket I'd do the version I posted initially on the first page as I've built that one for a 6BQ5 amp before and it works great. Also with an extra socket for another tube stage I'd need to either invert the feedback or leave the first stage out of the feedback loop.
Unfortunately I don't have 47k resistors and had to make the 56k by paralleling a 120k and 100k.
The cap for the NFB will need to be quite large though. I figure greater than 1uF in order for there to be zero phase shift down to 20Hz.
What needs to be changed to bias the 12AX7 better and would the 12BZ7 perform better with that circuit or will the THD be higher?
Also is the feedback arrangement ok or can it be done better?
If I needed more gain I can increase the 100k plate resistor, right?
Now If I had an extra socket I'd do the version I posted initially on the first page as I've built that one for a 6BQ5 amp before and it works great. Also with an extra socket for another tube stage I'd need to either invert the feedback or leave the first stage out of the feedback loop.
Unfortunately I don't have 47k resistors and had to make the 56k by paralleling a 120k and 100k.
The cap for the NFB will need to be quite large though. I figure greater than 1uF in order for there to be zero phase shift down to 20Hz.
Last edited:
If you want to bias the split load inverter so it sinks more plate and cathode current, you can reduce the value of the 56k resistors. I'd try 47k and see if it does any better.
Using 12BZ7, THD will be higher. That's not as linear a tube as 12AX7.
6V6 pentodes are an easy load. What you have should be good enough. At least that's what I think. Someone may disagree. That usually happens around here.
The NFB layout looks fine. It does not need a DC blocking capacitor, unless you find you really need it. This is a very commonly used circuit, and I don't remember anybody putting a capacitor in series with the feedback resistor.
If you need more gain, you can:
1) Split the cathode resistor of the input 12AX7 into two parts. Maybe 1.3k and 200 ohms. Then bypass that 1.3k resistor with a 100uF capacitor in parallel. Apply the negative feedback to the junction of the 1.3k resistor//100uF cap and 200 ohm resistor (the other end of which connects to signal ground). Start with a 2.2k ohm feedback resistor, then try some different values until you get the sound you like. Some prefer less NFB for a looser sound, others prefer more NFB for tighter bass and a more damped sound.
2) You can also re-bias the input 12AX7 with a higher value plate load resistor (perhaps 150k ohms, or even 220k ohms) but you will want to keep the plate current of that first stage 12AX7 above about 600uA (0.6mA) to avoid 'starved' operation. You will probably need to change the value of the cathode load resistor too.
By my reckoning, if you use a 220k ohm plate resistor to the first stage 12AX7, you will need to change its cathode load resistor to 2.3k ohms. You could conveniently put 2.2k ohms with a 47uF capacitor in parallel, and put that on top of a 100 ohm resistor. Then the feedback resistor from the OPT secondary would connect to the junction of the 2.2k//47uF and 100 ohm resistors.
Here's the load line for that:
There are many things that can be done to modify the circuit this way or that. Just remember that every change brings both benefits and defects. Everything is a compromise. Nothing is perfect. That's just the way it is.
--
Using 12BZ7, THD will be higher. That's not as linear a tube as 12AX7.
6V6 pentodes are an easy load. What you have should be good enough. At least that's what I think. Someone may disagree. That usually happens around here.
The NFB layout looks fine. It does not need a DC blocking capacitor, unless you find you really need it. This is a very commonly used circuit, and I don't remember anybody putting a capacitor in series with the feedback resistor.
If you need more gain, you can:
1) Split the cathode resistor of the input 12AX7 into two parts. Maybe 1.3k and 200 ohms. Then bypass that 1.3k resistor with a 100uF capacitor in parallel. Apply the negative feedback to the junction of the 1.3k resistor//100uF cap and 200 ohm resistor (the other end of which connects to signal ground). Start with a 2.2k ohm feedback resistor, then try some different values until you get the sound you like. Some prefer less NFB for a looser sound, others prefer more NFB for tighter bass and a more damped sound.
2) You can also re-bias the input 12AX7 with a higher value plate load resistor (perhaps 150k ohms, or even 220k ohms) but you will want to keep the plate current of that first stage 12AX7 above about 600uA (0.6mA) to avoid 'starved' operation. You will probably need to change the value of the cathode load resistor too.
By my reckoning, if you use a 220k ohm plate resistor to the first stage 12AX7, you will need to change its cathode load resistor to 2.3k ohms. You could conveniently put 2.2k ohms with a 47uF capacitor in parallel, and put that on top of a 100 ohm resistor. Then the feedback resistor from the OPT secondary would connect to the junction of the 2.2k//47uF and 100 ohm resistors.
Here's the load line for that:
There are many things that can be done to modify the circuit this way or that. Just remember that every change brings both benefits and defects. Everything is a compromise. Nothing is perfect. That's just the way it is.
--
Last edited:
Oh ok no cap is needed due to the DC voltage on the cathode being so low, right?
My thinking was what would the effect be if I made the feedback resistor variable and adjust it lower as that would lower the cathode resistance some.
My thinking was what would the effect be if I made the feedback resistor variable and adjust it lower as that would lower the cathode resistance some.
Right, and especially if you split the cathode load into two parts, the top R with a C in parallel (bypassed) and below that the feedback Rseries, which will probably be only 100 ohms or so. With 600uA across 100 ohms:
100R(0.0006A) = 0.06 V
If you connect the NFB resistor directly to the cathode of the input 12AX7, the secondary of the OPT will have about 1.4V DC across it. That still shouldn't hurt the speakers.
If you apply too much negative feedback (by making Rf a small value in the circuit above) that's likely to provoke oscillation, which you can see as a spike on a 1kHz square wave. If that happens you'll need to install compensation capacitors to damp that oscillation. That's a whole other can of worms...
100R(0.0006A) = 0.06 V
If you connect the NFB resistor directly to the cathode of the input 12AX7, the secondary of the OPT will have about 1.4V DC across it. That still shouldn't hurt the speakers.
If you apply too much negative feedback (by making Rf a small value in the circuit above) that's likely to provoke oscillation, which you can see as a spike on a 1kHz square wave. If that happens you'll need to install compensation capacitors to damp that oscillation. That's a whole other can of worms...
Last edited:
Here's a more detailed illustration of what I was writing about...
That will give you the most possible gain, open loop.
You can then adjust the value of Rf to set the overall gain.
That will give you the most possible gain, open loop.
You can then adjust the value of Rf to set the overall gain.
I'll try that circuit today.
I'll wire it up for the 12AX7 then I may drop in a 12AT7 or 12AU7 and see how much the gain drops and if it's possible to make up that gain with just the one tube if I cannot adjust negative feedback to where 1V equals full output.
Now If I lower the value of the 50k resistor supplying the B+ to the driver and phase splitter I can get up to 339Vdc B+.
I'll wire it up for the 12AX7 then I may drop in a 12AT7 or 12AU7 and see how much the gain drops and if it's possible to make up that gain with just the one tube if I cannot adjust negative feedback to where 1V equals full output.
Now If I lower the value of the 50k resistor supplying the B+ to the driver and phase splitter I can get up to 339Vdc B+.
The design of the power supply is every bit as important as the audio circuit design.
Yes, please do. That would be very helpful.