The LTP (or PI, if you like).
To understand this circuit
The first stage is the voltage amplifier. It is single-ended (only one output, which is inverting).
The second stage is actually two stages combined. It is a phase splitter with two outputs in opposite polarity (push-pull). It is also the push-pull driver stage for the output tubes' grids.
So I think of the second stage as a combined phase-splitter/push-pull driver stage. It's often called a "long-tailed pair" or "Schmit inverter".
If you look at the Williamson amp, the first stage is a single-ended voltage amplifier stage, the second stage is a phase splitter (cathodyne or "concertina", in this case) and the third stage is a push-pull driver stage. The Mullard 5-20 simplifies this by combining the 2nd and 3rd stages into a single LTP stage.
--
To understand this circuit
The first stage is the voltage amplifier. It is single-ended (only one output, which is inverting).
The second stage is actually two stages combined. It is a phase splitter with two outputs in opposite polarity (push-pull). It is also the push-pull driver stage for the output tubes' grids.
So I think of the second stage as a combined phase-splitter/push-pull driver stage. It's often called a "long-tailed pair" or "Schmit inverter".
If you look at the Williamson amp, the first stage is a single-ended voltage amplifier stage, the second stage is a phase splitter (cathodyne or "concertina", in this case) and the third stage is a push-pull driver stage. The Mullard 5-20 simplifies this by combining the 2nd and 3rd stages into a single LTP stage.
--
Last edited:
Hi Ron,
I replaced the bias resistor on the 6DJ8 with a 1K, and swapped the plate resistors on the 6CG7s for 18K.
I now have 151V on the plates of the 6DJ8 and 4.4V on the cathodes
163V on the 6CG7 plates and 4.2V on their cathodes at the CCS.
Tested it today and got full bandwidth within 1db from 20-35Khz at 1W under .8% and under 2% full bandwidth at 12.75W.
this was using a 10.2ohm load, so let's play it safe and say it is a 13W amp.
I do have the HV supply too close to the left channel, so I am going to move it because it is injecting a but if noise into that channel.
Blair
I replaced the bias resistor on the 6DJ8 with a 1K, and swapped the plate resistors on the 6CG7s for 18K.
I now have 151V on the plates of the 6DJ8 and 4.4V on the cathodes
163V on the 6CG7 plates and 4.2V on their cathodes at the CCS.
Tested it today and got full bandwidth within 1db from 20-35Khz at 1W under .8% and under 2% full bandwidth at 12.75W.
this was using a 10.2ohm load, so let's play it safe and say it is a 13W amp.
I do have the HV supply too close to the left channel, so I am going to move it because it is injecting a but if noise into that channel.
Blair
Hey, lookin' good there!
Is the plate resistor on the 6DJ8 a 27k?
+4.4V across a 1k resistor means you've got 4.4mA going through that 6DJ8. While there's nothing wrong with that, I find that 6DJ8's sound better with closer to 10mA (or even 15mA) going through them.
Is that with or without NFB?
--
Is the plate resistor on the 6DJ8 a 27k?
+4.4V across a 1k resistor means you've got 4.4mA going through that 6DJ8. While there's nothing wrong with that, I find that 6DJ8's sound better with closer to 10mA (or even 15mA) going through them.
Is that with or without NFB?
--
Hi Ron,
The 6DJ8 could be biased a bit harder, but it is clean sounding so I left it alone. I did think about dropping a red LED as the CCS for it, but haven't yet.
I have not installed the feedback yet either. I do not know how to calculate the values? What is the equation?
Now that I have it up an running, my curiosity about the 7533 is kicking in, but the extra few watts is probably not worth $100
I also rearranged the PS last night to install a 500 ohm between the plates and the screens to drop the voltage so the plates are higher than the screens. It sounded a bit better with the screens at a lower voltage than the plates.
Thanks,
Blair
The 6DJ8 could be biased a bit harder, but it is clean sounding so I left it alone. I did think about dropping a red LED as the CCS for it, but haven't yet.
I have not installed the feedback yet either. I do not know how to calculate the values? What is the equation?
Now that I have it up an running, my curiosity about the 7533 is kicking in, but the extra few watts is probably not worth $100
I also rearranged the PS last night to install a 500 ohm between the plates and the screens to drop the voltage so the plates are higher than the screens. It sounded a bit better with the screens at a lower voltage than the plates.
Thanks,
Blair
Looking better there. Those measurements you sent are pretty good for a pentode amp open loop.
A few observations...
1) You have the 6L6's bias set so that the tubes are drawing only 45mA each. With a B+ of 300V, that means each 6L6's plate+screen dissipation is only 13.5W. The 6L6GB (not GC) is rated to 19W plate dissipation. You can safely draw 55mA to 60mA through each 6L6 in this amp (0.55V to 0.6V at each 6L6 cathode). That will make the output tubes easier to drive, and they'll be operating more in class A, so should be more linear. Since this amp is never going to make a lot of power, why not tweak it for nicest sound? (= lowest distortion and widest bandwidth open loop, and least need for NFB)
*Question: What is the bias voltage at the 6L6 grid that gives you 45mA plate+screen current?
2) The 6CG7 with only 6mA plate current per triode is still not in its most linear region. The reason is that the 6CG7 rp will be higher at that low a plate current, and more variable (with more current the rp goes down, with less current the rp goes up). Once you get to about 8mA the 6CG7 rp gets more constant, so varies less with signal. 9mA or 10mA is even better. With the rp varying less, the stage will operate with lower distortion.
3) The math for calculating the NFB resistor value is way too complex for me. The easiest way I know to calculate the feedback resistor value is to start with something like 100k, then parallel other resistors to make lesser values as needed (220k in parallel with 100k = 69k, 100k in parallel with 100k = 50k, etc.).
Find a convenient output voltage to work with. Let's say 3V out, as measured from the OPT secondary into a 10 ohm dummy load (or 8 ohm).
Measure the input signal voltage that gives you 3V out. Let's say it's 100mV.
Now, add the 100k feedback resistor. Let's say it now takes 200mV to drive the amp to 3V out (the whole amp has become half as sensitive). That means the amp now has half the gain it had when open loop, so adding that 100k feedback resistor added 6dB of negative feedback. 6dB of gNFB is a good target to aim for (IMO, YMMV). Give it a listen, measure the distortion, look at the clipping behavior and bandwidth on a 'scope.
To adjust for more/less gNFB: The larger the value of the feedback resistor, the less NFB there will be around the amp. The smaller the value of feedback resistor, the more NFB.
Remember that this amp has the feedback applied to the second (LTP) stage, so that is the stage that will get less sensitive, and will need to be driven harder. That's a distinct disadvantage. However, the advantage is that you will have only one RC time constant and the OPT within the feedback loop, thus increasing the stability margin. Your first stage will be a fairly linear SE triode running open loop, so should contribute a nice sound to the amp.
You can also use the 6DJ8's 1k cathode resistor as the bottom leg of the gNFB voltage divider. If you have Morgan Jones' "Valve Amplifiers", there's a really good discussion of this type of arrangement in the chapter on his Beavois Valley amplifier.
--
PS - the Common on the OPT secondary winding needs to be grounded.
A few observations...
1) You have the 6L6's bias set so that the tubes are drawing only 45mA each. With a B+ of 300V, that means each 6L6's plate+screen dissipation is only 13.5W. The 6L6GB (not GC) is rated to 19W plate dissipation. You can safely draw 55mA to 60mA through each 6L6 in this amp (0.55V to 0.6V at each 6L6 cathode). That will make the output tubes easier to drive, and they'll be operating more in class A, so should be more linear. Since this amp is never going to make a lot of power, why not tweak it for nicest sound? (= lowest distortion and widest bandwidth open loop, and least need for NFB)
*Question: What is the bias voltage at the 6L6 grid that gives you 45mA plate+screen current?
2) The 6CG7 with only 6mA plate current per triode is still not in its most linear region. The reason is that the 6CG7 rp will be higher at that low a plate current, and more variable (with more current the rp goes down, with less current the rp goes up). Once you get to about 8mA the 6CG7 rp gets more constant, so varies less with signal. 9mA or 10mA is even better. With the rp varying less, the stage will operate with lower distortion.
3) The math for calculating the NFB resistor value is way too complex for me. The easiest way I know to calculate the feedback resistor value is to start with something like 100k, then parallel other resistors to make lesser values as needed (220k in parallel with 100k = 69k, 100k in parallel with 100k = 50k, etc.).
Find a convenient output voltage to work with. Let's say 3V out, as measured from the OPT secondary into a 10 ohm dummy load (or 8 ohm).
Measure the input signal voltage that gives you 3V out. Let's say it's 100mV.
Now, add the 100k feedback resistor. Let's say it now takes 200mV to drive the amp to 3V out (the whole amp has become half as sensitive). That means the amp now has half the gain it had when open loop, so adding that 100k feedback resistor added 6dB of negative feedback. 6dB of gNFB is a good target to aim for (IMO, YMMV). Give it a listen, measure the distortion, look at the clipping behavior and bandwidth on a 'scope.
To adjust for more/less gNFB: The larger the value of the feedback resistor, the less NFB there will be around the amp. The smaller the value of feedback resistor, the more NFB.
Remember that this amp has the feedback applied to the second (LTP) stage, so that is the stage that will get less sensitive, and will need to be driven harder. That's a distinct disadvantage. However, the advantage is that you will have only one RC time constant and the OPT within the feedback loop, thus increasing the stability margin. Your first stage will be a fairly linear SE triode running open loop, so should contribute a nice sound to the amp.
You can also use the 6DJ8's 1k cathode resistor as the bottom leg of the gNFB voltage divider. If you have Morgan Jones' "Valve Amplifiers", there's a really good discussion of this type of arrangement in the chapter on his Beavois Valley amplifier.
--
PS - the Common on the OPT secondary winding needs to be grounded.
Last edited:
Hi Ron,
I have roughly -26V on the 6L6 grids.
I bumped it up to 50mA, but around 60mA, the tranny feels a bit hot. Could be nothing and just the filaments heating up the core.
I could swap the 100ohm resistor on the CCS for 75ohms to bump the 6CG7 current.
I see a bunch of values for NFB. Most commonly with 6L6 is between 10K and 30K. I see some as just resistors and some with small caps across the resistors in parallel. Do I need the cap?
Blair
I have roughly -26V on the 6L6 grids.
I bumped it up to 50mA, but around 60mA, the tranny feels a bit hot. Could be nothing and just the filaments heating up the core.
I could swap the 100ohm resistor on the CCS for 75ohms to bump the 6CG7 current.
I see a bunch of values for NFB. Most commonly with 6L6 is between 10K and 30K. I see some as just resistors and some with small caps across the resistors in parallel. Do I need the cap?
Blair
The two feedback resistors form a voltage divider, which controls how much of the output is fed back to the input. (These would be the 1k resistor going from the LTP second triode grid to ground and the resistor coming from the OPT secondary to that tube's grid.) The gain of the tube that is receiving the feedback voltage also enters into the feedback equation. So any two different circuits are going to use different resistor values, even if they arrive at the same gain reduction from application of NFB.
There's a relationship between the grid voltage on the 6L6 grids, the amount of voltage output needed from the driver stage, and the amount of feedback applied.
1) The grid bias on the 6L6 determines the peak signal voltage that can be accepted by the 6L6. So we aim for 26V peak input (52V peak-to-peak).
(If you add plate-to-plate ("Schade") feedback to the 6L6's, then their input sensitivity will be reduced and you will need the driver stage to deliver higher peak signal voltage to the 6L6 grids.)
2) The 6CG7 LTP will have a gain of only around 7 or 8. I'll call it 7, to be safe. That means that in order to get the 26Vpk at the 6L6 grid, we'll need the 6CG7 LTP to be able to accept 3.7Vpk at its grids. (3.7 x 7 = 25.9) With the voltage at the cathodes (grid bias) on the 6CG7 at 4.7V now, you have enough, but with little headroom. It will work.
Increasing the current through the 6CG7's will reduce the voltage at the cathodes. That will erode any safety margin you have there. So I think you'd need to also reduce the value of the plate resistors if you want to get more current going through the 6CG7.
Since the rp of a 6CG7 with 8mA plate current is still about 8k (and actually slightly higher when used in an LTP), the 18k resistors currently in there represent a very low resistance load -- but quite acceptable nonetheless.
In other words, I think you've done about as well as you can with a 6CG7 and only 290V B+. That's why I suggested a possible switch to 12BH7A. You could also use 6GU7 (cheaper, similar). If you have the current capacity, 5687 is also good. All of these can work well with plate voltage around 150V and bias of about -5V.
The other solution would be to run the 6L6's hotter, so their grid bias would be lower, and they would require a few less volts of input drive from the driver stage. You'll get a bit less power this way, but would lower open-loop distortion a little.
Yet another possibility is to reduce the voltage to the 6L6 screen grids some more, with additional filtering. The additional filtering will serve to help regulate them a little, which lowers distortion. The 6L6 will become a little easier to drive, but power output will go down a bit. Better yet would be to actually regulate the screen supply. In a tetrode or pentode, it's best to hold the screens at a constant voltage, while the plate voltage can be allowed to vary.
How do you feel about winding up with a 10W to 12W per channel pentode amp?
--
I forgot about the capacitor you see paralleled across the feedback resistor. That cap is put there to reduce the amount of negative feedback at some higher frequency.
If the OPT has a resonance up above the audio band (often at about 35kHz or so), then you'll see an overshoot on square waves. Reducing the negative feedback at that frequency of resonance will help reduce that overshoot on square waves.
It usually takes some trial-and-error to find the right value for that capacitor. Choose the capacitor value that makes the prettiest looking square wave at the output of the amplifier.
--
If the OPT has a resonance up above the audio band (often at about 35kHz or so), then you'll see an overshoot on square waves. Reducing the negative feedback at that frequency of resonance will help reduce that overshoot on square waves.
It usually takes some trial-and-error to find the right value for that capacitor. Choose the capacitor value that makes the prettiest looking square wave at the output of the amplifier.
--
What about the 6H30? I have a pair of those. With the CCS, there should be no issues dropping it in. The rp is low, so 18K should be fine.
6H30 Tube Data
Blair
Dropped a pair in there and have 160V plate voltage and 10.2-4V on the cathodes. I'm only using 2W plate resistors, so I'm not sure if they will handle the current.
6H30 Tube Data
Blair
Dropped a pair in there and have 160V plate voltage and 10.2-4V on the cathodes. I'm only using 2W plate resistors, so I'm not sure if they will handle the current.
Last edited:
OK,
10k looks like the value to go with.
I started at 22K because it was on the high side of what I have seen:
Results:
Load: 8.75 ohms
Pulled NFB off 8 ohm tap
39.7mV no NFB for 3.07V
57.0mV with 20K NFB resistor to 3.07V
74.8mV with 10K NFB resistor to 3.05V
It's a pretty sensitive amp.
Blair
10k looks like the value to go with.
I started at 22K because it was on the high side of what I have seen:
Results:
Load: 8.75 ohms
Pulled NFB off 8 ohm tap
39.7mV no NFB for 3.07V
57.0mV with 20K NFB resistor to 3.07V
74.8mV with 10K NFB resistor to 3.05V
It's a pretty sensitive amp.
Blair
6H30P is a horse of a different color. You could use 10k resistors in the plates, no problem. Its rp is around 2k or lower.
IMO, that's overkill. 6H30P is a very expensive tube, so you might want to save those for a more ambitious project, like an all-triode amp (big output triodes need a beefy driver to overcome their Miller capacitance).
I was thinking the 6CG7's you have in there now should work pretty well, and since this amp will have to have NFB around it, you might wind up liking the sound just fine. If you don't, then you can experiment with 12BH7, or maybe 6N6P (a Russian near-equivalent to ECC99 that is available for not too much money and can sound excellent). 5687 is another tube that could work.
Bear in mind that 6N6P or 5687 will need about 12 to 16mA plate current per triode. Does your power transformer have enough capacity to deliver 48mA to 64mA just for the LTPs?
The 12BH7A can get you there with about half the current draw. That's what I used in my Stereo 70 re-do.
--
IMO, that's overkill. 6H30P is a very expensive tube, so you might want to save those for a more ambitious project, like an all-triode amp (big output triodes need a beefy driver to overcome their Miller capacitance).
I was thinking the 6CG7's you have in there now should work pretty well, and since this amp will have to have NFB around it, you might wind up liking the sound just fine. If you don't, then you can experiment with 12BH7, or maybe 6N6P (a Russian near-equivalent to ECC99 that is available for not too much money and can sound excellent). 5687 is another tube that could work.
Bear in mind that 6N6P or 5687 will need about 12 to 16mA plate current per triode. Does your power transformer have enough capacity to deliver 48mA to 64mA just for the LTPs?
The 12BH7A can get you there with about half the current draw. That's what I used in my Stereo 70 re-do.
--
I just saw your post about the NFB resistor. 10k it is. Looks good. How does it sound?
Yes, it's bound to be a very sensitive amp. You certainly won't need to use a preamp with it. Just a volume control on the front of it and you'll be fine.
Next step -- What is the voltage out at the onset of clipping?
If it's about 12V peak, then that should be about 12.5W rms.
With 12V peak output, is the input voltage about 300mV?
--
Yes, it's bound to be a very sensitive amp. You certainly won't need to use a preamp with it. Just a volume control on the front of it and you'll be fine.
Next step -- What is the voltage out at the onset of clipping?
If it's about 12V peak, then that should be about 12.5W rms.
With 12V peak output, is the input voltage about 300mV?
--
Last edited:
Hi Ron,
I did not install it yet. Just gator clipped in.
With the open loop performance, I'm not too sure looking for a better tube than the 6CG7 is the right direction. If the NFB get any better results, it should be very nice!
I'll get it installed and let you know what happens.
Since I got around 12.75W into 10.25 ohms, you are correct that the 6L6 are biased at around 13.5W which is probably what I'd get into 8 ohms.
Blair
I did not install it yet. Just gator clipped in.
With the open loop performance, I'm not too sure looking for a better tube than the 6CG7 is the right direction. If the NFB get any better results, it should be very nice!
I'll get it installed and let you know what happens.
Since I got around 12.75W into 10.25 ohms, you are correct that the 6L6 are biased at around 13.5W which is probably what I'd get into 8 ohms.
Blair
Last edited:
Hi, friend
Today I bought a bogen rp40a, I was writing to you, to see if you would share the manual with me. I will appreciate it a lot
Atte
Patricio
From Chile