Smoking-amp, I find the schematic that you posted quite interesting. I was chasing a "universal driver" board design several years ago. It resulted in a couple of amps that were seen in this thread:
http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp.html?highlight=6l6GC+AB2
Fast forward maybe 6 years and I dug out the old breadboard and revisited the design. I found that connecting the "Schade" feedback from the output tube plates to the opposite predriver plates was the best way to clean up some crummy OPT's. The resulting schematic is pretty much identical to the one you posted except the CCS in the tail of the first stage is solid state, and there IS a CCS in the tail of the second stage. There are mosfet source followers on the board outputs for driving the grids of any tube to any positive or negative voltage.
The original breadboard indeed used a 6SL7 and a 6SN7. I got a small batch of PC boards made, but they used 9 pin tubes because there is a much wider selection of tubes that will work.
Right before I boxed everything up to move, I had an amp running with a 5751 in the input stage, a 12BH7 for the second stage. The output tubes were 13GB5's in dual drive. There was far too much gain, even with feedback.
There was another amp with un-named output tubes in dual drive and 6CG7 in both driver stages that worked much better, bur it was ONLY a 50 watt design.
I have noticed some instability with large amounts of partial feedback, and it is OPT dependent. I am under the assumption that the $16 OPT's that I am experimenting with cause a phase shift somewhere in the 50 to 100 KHz region that makes negative feedback positive!
I have spent much of today building myself a new workbench with materials rescued from the construction dumpster while my house was being built. It will still be a long time before I have a fully operational lab for Tubelab.
http://www.diyaudio.com/forums/tubes-valves/133034-6l6gc-ab2-amp.html?highlight=6l6GC+AB2
Fast forward maybe 6 years and I dug out the old breadboard and revisited the design. I found that connecting the "Schade" feedback from the output tube plates to the opposite predriver plates was the best way to clean up some crummy OPT's. The resulting schematic is pretty much identical to the one you posted except the CCS in the tail of the first stage is solid state, and there IS a CCS in the tail of the second stage. There are mosfet source followers on the board outputs for driving the grids of any tube to any positive or negative voltage.
The original breadboard indeed used a 6SL7 and a 6SN7. I got a small batch of PC boards made, but they used 9 pin tubes because there is a much wider selection of tubes that will work.
Right before I boxed everything up to move, I had an amp running with a 5751 in the input stage, a 12BH7 for the second stage. The output tubes were 13GB5's in dual drive. There was far too much gain, even with feedback.
There was another amp with un-named output tubes in dual drive and 6CG7 in both driver stages that worked much better, bur it was ONLY a 50 watt design.
I have noticed some instability with large amounts of partial feedback, and it is OPT dependent. I am under the assumption that the $16 OPT's that I am experimenting with cause a phase shift somewhere in the 50 to 100 KHz region that makes negative feedback positive!
I have spent much of today building myself a new workbench with materials rescued from the construction dumpster while my house was being built. It will still be a long time before I have a fully operational lab for Tubelab.
Hi George,
I suspect that basic design goes back a long ways. Then CCSs came along to make it better. The partial feedback arrangement makes it possible to add more gain in the loop than simple Schade, so its more effective. But the OT is still getting in the way some when we take feedback from both sides, while only one side is driven in class AB. Maybe that's why RCA did the combined Schade + partial Fdbk, to reduce the gain some in the larger loop.
The other (inactive tube side) Fdbk is having to pass through the OT from P to P, and that is usually poorly coupled in OTs. The one saving grace is the lack of any loading on it then. Some kind of HF low pass filter on the feedbacks might help, but it would have to cut off below the OT bad spot, so that is not real attractive.
A Circlotron type OT should fix that P to P coupling issue though. Or taking the Fdbks from a smaller % CFB winding. Using the %UL taps may work too.
If one has an abundance of identical OTs, the following would work good:
A Crowhurst Twin (a reduced CFB version of that).
If one takes the partial Fdbk back to the driver cathodes, then the Fdbks don't need to be crossed. So the driver could be split into separate halves, each with a Fdbk from just the tube being controlled. That avoids the issue of Fdbk coupled P to P through the OT, not sure how well that will work for crossover control though. Maybe provide a low R value tail, LTP, driver for that setup, so only modest cross effects occur. Could then adjust the tail R upwards to just below the setting that causes instability.
Don
I suspect that basic design goes back a long ways. Then CCSs came along to make it better. The partial feedback arrangement makes it possible to add more gain in the loop than simple Schade, so its more effective. But the OT is still getting in the way some when we take feedback from both sides, while only one side is driven in class AB. Maybe that's why RCA did the combined Schade + partial Fdbk, to reduce the gain some in the larger loop.
The other (inactive tube side) Fdbk is having to pass through the OT from P to P, and that is usually poorly coupled in OTs. The one saving grace is the lack of any loading on it then. Some kind of HF low pass filter on the feedbacks might help, but it would have to cut off below the OT bad spot, so that is not real attractive.
A Circlotron type OT should fix that P to P coupling issue though. Or taking the Fdbks from a smaller % CFB winding. Using the %UL taps may work too.
If one has an abundance of identical OTs, the following would work good:
A Crowhurst Twin (a reduced CFB version of that).
If one takes the partial Fdbk back to the driver cathodes, then the Fdbks don't need to be crossed. So the driver could be split into separate halves, each with a Fdbk from just the tube being controlled. That avoids the issue of Fdbk coupled P to P through the OT, not sure how well that will work for crossover control though. Maybe provide a low R value tail, LTP, driver for that setup, so only modest cross effects occur. Could then adjust the tail R upwards to just below the setting that causes instability.
Don
Last edited:
Another thought on the partial feedback to the driver grids approach. This approach is the easiest to design (just set the forward gain in the attenuator ratio like for Schade), so it would be nice to eliminate the issue of crossed feedbacks with a class aB output stage.
By using a small CFB winding, the partial feedbacks can be taken from the cathodes rather than the plates (or UL taps). This inverts the phases, so no feedback crossover (for phase) is required. Then the driver stage does not necessarily need to be differential. An R tail should be fine. A simple approach would be to just use the secondary 0,4,16 Ohm taps for P-P CFB, but this introduces questions of turns matching in the secondary, and makes most of the Fdbk signal (like 90%) have to cross primary to secondary (a stability issue maybe).
The Crowhurst Twin like arrangement or similar (below) would work perfectly for CFB derived partial Fdbk taps.
Another reason I like the Crowhurst Twin like arrangements is that each OT used in it can be thought of as one bobbin section of a split bobbin balanced OT. By flipping around the phase of one OT ( phase flip both primary and secondary leads), one cancels out any inherent P-P winding assymetry in a cheap OT. (the cap OT to OT couplings cross link opposite dis-symmetry sides to balance it out, for leakage L and winding resistance of the P-P sides) So one has a simple way of making the equivalent of a very expensive OT, using two low primary Z Edcors for example. Not to mention the doubling of possible output power. A 200 Watt, high quality, off the shelf OT is no sweat at all. And its still much cheaper ($100+100 versus $335 1650WA at Mouser) than Hammond's 280 Watter.
By using a small CFB winding, the partial feedbacks can be taken from the cathodes rather than the plates (or UL taps). This inverts the phases, so no feedback crossover (for phase) is required. Then the driver stage does not necessarily need to be differential. An R tail should be fine. A simple approach would be to just use the secondary 0,4,16 Ohm taps for P-P CFB, but this introduces questions of turns matching in the secondary, and makes most of the Fdbk signal (like 90%) have to cross primary to secondary (a stability issue maybe).
The Crowhurst Twin like arrangement or similar (below) would work perfectly for CFB derived partial Fdbk taps.
Another reason I like the Crowhurst Twin like arrangements is that each OT used in it can be thought of as one bobbin section of a split bobbin balanced OT. By flipping around the phase of one OT ( phase flip both primary and secondary leads), one cancels out any inherent P-P winding assymetry in a cheap OT. (the cap OT to OT couplings cross link opposite dis-symmetry sides to balance it out, for leakage L and winding resistance of the P-P sides) So one has a simple way of making the equivalent of a very expensive OT, using two low primary Z Edcors for example. Not to mention the doubling of possible output power. A 200 Watt, high quality, off the shelf OT is no sweat at all. And its still much cheaper ($100+100 versus $335 1650WA at Mouser) than Hammond's 280 Watter.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.