I'm trying to decide on a push-pull amp to start building. I have most of the parts required, including transformers, and a couple of hand-me-down chassis I can use. I was thumbing through one of my old copies of the RCA Receiving Tubes manuals and found this in RC-19:
R18 and R19 (100k) take signal off the 6V6 plates to the 6AU6 driver cathodes for NFB.
R9 and R10 (330k) are a mystery to me. Simulation shows they increase the NFB a little. But how? I guess they feed a little output signal from the cathodes to the grids of the 6AU6s. That would create a little positive feedback, no?
Otherwise the design is straightforward.
I don't have a 9k or 10k OPT, but I do have a pair of 8k:VC OPTs.
The capacitor values introduce high pass filters in a couple of places that result in an infrasonic peak closed loop. I'd increase the values of C3, C4, C7, C8 to smooth that out. Otherwise the design looks well behaved, as the OPT is not within the feedback loop.
I would also improve the power supply; make it stiffer by using UF4007 diodes and bigger reservoir and first smoothing caps, and try to stabilize or regulate the 6V6 screen supply.
QUESTIONS:
???
R18 and R19 (100k) take signal off the 6V6 plates to the 6AU6 driver cathodes for NFB.
R9 and R10 (330k) are a mystery to me. Simulation shows they increase the NFB a little. But how? I guess they feed a little output signal from the cathodes to the grids of the 6AU6s. That would create a little positive feedback, no?
Otherwise the design is straightforward.
I don't have a 9k or 10k OPT, but I do have a pair of 8k:VC OPTs.
The capacitor values introduce high pass filters in a couple of places that result in an infrasonic peak closed loop. I'd increase the values of C3, C4, C7, C8 to smooth that out. Otherwise the design looks well behaved, as the OPT is not within the feedback loop.
I would also improve the power supply; make it stiffer by using UF4007 diodes and bigger reservoir and first smoothing caps, and try to stabilize or regulate the 6V6 screen supply.
QUESTIONS:
- Would it be better to use plate-grid FB around the 6V6 outputs, or is the 6V6-plate to 6AU6-cathode FB basically the same thing done a little differently?
- Is the addition of R9 and R10 a good idea? I've never seen that done in any other amp designs I've looked at.
- I assume using a 6FQ7 or 6CG7 in place of the 12AU7 would be an improvement and would not require much in the way of changes. Good idea?
- I might also try this with EL84 or even 6P15P outputs instead of 6V6s. Thoughts on that? I think it would be an improvement, and they'd be easier to drive as well.
???
Is it the same concept discussed here? https://www.diyaudio.com/community/threads/vacuum-tube-power-amplifier-2-x-50-w.308590/
I was thinking the same, but why not build it "as is" and see how it sounds and measures? Then, you could go nuts with different configs.
jeff
@nerdorama - I think the two are exactly the same.
@zintolo - It's an earlier and simpler design than the 50-watt amp in that other thread. Someday I'd like to make that 50W design. Many here (like Wavebourn and SY) have said that is an excellent design. But that would be a major undertaking, and I'd have to redesign the first two stages to use something other than the unobtainium 7199. Maybe just use the 12AU7 circuitry from this RC10 and accept the lower sensitivity. The voltage amp-split load inverter stages are not inside any feedback loops, so that could actually work. [EDIT -- OOPS! Wrong. The global NFB goes to the 7199 input stage cathode, so it would need something with at least close to similar gain. Oh well.]
@nicoch58 - Poindexter's Musical Machine is a PP triode amp, so it will make less power. Probably only 4 watts. I'm hoping to get at least 10W per channel. Eli Duttman's El Cheapo is similar but uses a 12AT7 LTP instead of the Musical Machine's 6GK5 triodes.
@vinylkid58 - Good points. I have some 12AU7s and 6AU6s, so I might as well give it a go 'as is'. I also have some 6CG7s and 6CF6s, which use the same sockets as the RCA-specified tubes, so the conversion wouldn't be difficult.
I simulated the 6AU6 LTP and 6V6 PP stages with a model of a Hammond 1608 OPT. The closed loop gain is just slightly higher than unity, so it looks like all the voltage gain comes from the first stage 12AU7 voltage amplifier. The second stage is the split-load phase inverter, which is also close to unity gain.
@zintolo - It's an earlier and simpler design than the 50-watt amp in that other thread. Someday I'd like to make that 50W design. Many here (like Wavebourn and SY) have said that is an excellent design. But that would be a major undertaking, and I'd have to redesign the first two stages to use something other than the unobtainium 7199. Maybe just use the 12AU7 circuitry from this RC10 and accept the lower sensitivity. The voltage amp-split load inverter stages are not inside any feedback loops, so that could actually work. [EDIT -- OOPS! Wrong. The global NFB goes to the 7199 input stage cathode, so it would need something with at least close to similar gain. Oh well.]
@nicoch58 - Poindexter's Musical Machine is a PP triode amp, so it will make less power. Probably only 4 watts. I'm hoping to get at least 10W per channel. Eli Duttman's El Cheapo is similar but uses a 12AT7 LTP instead of the Musical Machine's 6GK5 triodes.
@vinylkid58 - Good points. I have some 12AU7s and 6AU6s, so I might as well give it a go 'as is'. I also have some 6CG7s and 6CF6s, which use the same sockets as the RCA-specified tubes, so the conversion wouldn't be difficult.
I simulated the 6AU6 LTP and 6V6 PP stages with a model of a Hammond 1608 OPT. The closed loop gain is just slightly higher than unity, so it looks like all the voltage gain comes from the first stage 12AU7 voltage amplifier. The second stage is the split-load phase inverter, which is also close to unity gain.
Last edited:
It's a baby Williamson without the GNFB, probably to avoid the instability of the Williamson with the extra front stage. And uses pentode drivers. Better to just use a 6GH8A front end and stay in the Mullard camp and put the GNFB back in place.
6GH8A pentode DC-coupled to 6GH8A triode split-load inverter AC coupled to push-pull 6V6 (or EL34, or 6L6, or 6550, etc.) w/ gNFB from OPT secondary to input stage cathode = Dynaco ST70, MkIII, etc.
Ya, that kind of driver stage is good for EL84/6BQ5, too. Laurent published a 7189 amp with 7199's, so 6GH8A would sub for that with the pinout change.
Thanks. All quite true. But I want to try this local NFB thing without the OPT in the feedback loop. Will it sound different? Worse? Only one way to find out...
I had a Dyna ST35 (nice sounding little amp), a Dyna ST70, and I even had a pair of Dyna MkIII way back in the 1990s (I should've kept those). All of those use that same basic topology of a high gain voltage amplifier stage DC-coupled to a split-load phase inverter, AC-coupled to PP-UL outputs, with a global NFB loop from OPT secondary to voltage amp cathode. That's always a possibility, but I've wanted to try this 'anode-follower output stage' type of thing for a long time, and I still haven't gotten around to building one. (Embarrassing how long it's taken me.)
I had a Dyna ST35 (nice sounding little amp), a Dyna ST70, and I even had a pair of Dyna MkIII way back in the 1990s (I should've kept those). All of those use that same basic topology of a high gain voltage amplifier stage DC-coupled to a split-load phase inverter, AC-coupled to PP-UL outputs, with a global NFB loop from OPT secondary to voltage amp cathode. That's always a possibility, but I've wanted to try this 'anode-follower output stage' type of thing for a long time, and I still haven't gotten around to building one. (Embarrassing how long it's taken me.)
I haven't used this type of local feedback in a PP amp, but I have two SE amps with plate to plate feedback. They sound great to me. As well, they are parallel feed. They have -3dB frequencies of at or below 20Hz and over 60kHz at 1 watt. The most recent is still at these point at 10W and nearly this bandwidth at 15W. I think this sort of feedback works very well. I would follow the RCA schematic. Should work great.
First of all, the primary inductance of the output transformer Does affect the phase and the amount of negative feedback at low frequencies.
It is a true low frequency pole.
Then there is the loudspeaker impedance versus frequency, as it reflects back to the output transformer primary.
But that is a completely different consideration, and is discussed in multiple Tubes/Valves threads.
I built both Schade negative feedback, and output tube plate to driver cathode negative feedback stages.
I prefer the output tube plate to driver tube cathode negative feedback.
Schade negative feedback greatly increases the load the 6AU6 drivers would see.
Similarly, plate of the 6V6 to the grid of the 6V6, greatly increases the load on the 6AU6 drivers.
I like plates to drive high impedances, or at least moderate impedances; I do not like plates to drive low impedances (just my preference).
There is a difference in the effect of a Pentode versus a Triode that drives a low impedance.
Check out some of Gary Pimm's amplifiers that used pentode drivers, and output plate to output grid, or output plate to driver plate forms of negative feedback.
The 100k plate to cathode resistors pump DC current into the self bias resistors (that affects the self bias voltage). The ratio of the 100k . . . to the 1.8k self bias resistor that is in parallel with the cathode impedance sets the amount of the negative feedback.
R9 with R8; and R10 with R7, are part of what causes the total amount of the self-bias voltage.
Have Fun!
It is a true low frequency pole.
Then there is the loudspeaker impedance versus frequency, as it reflects back to the output transformer primary.
But that is a completely different consideration, and is discussed in multiple Tubes/Valves threads.
I built both Schade negative feedback, and output tube plate to driver cathode negative feedback stages.
I prefer the output tube plate to driver tube cathode negative feedback.
Schade negative feedback greatly increases the load the 6AU6 drivers would see.
Similarly, plate of the 6V6 to the grid of the 6V6, greatly increases the load on the 6AU6 drivers.
I like plates to drive high impedances, or at least moderate impedances; I do not like plates to drive low impedances (just my preference).
There is a difference in the effect of a Pentode versus a Triode that drives a low impedance.
Check out some of Gary Pimm's amplifiers that used pentode drivers, and output plate to output grid, or output plate to driver plate forms of negative feedback.
The 100k plate to cathode resistors pump DC current into the self bias resistors (that affects the self bias voltage). The ratio of the 100k . . . to the 1.8k self bias resistor that is in parallel with the cathode impedance sets the amount of the negative feedback.
R9 with R8; and R10 with R7, are part of what causes the total amount of the self-bias voltage.
Have Fun!
I like to think of most of my amplifiers as "Forward Facing".
(Forward Facing signals that do not have any "retreat" of the signal (no negative feedback loops).
Quad had a topology called Feed Forward. It had an amplifier, with its usual imperfections, just like all amplifiers.
They took the signal output of that amplifier, and sent it to a lower power "corrections amplifier".
The two amplifiers output's were combined.
That topology meant that the main power amplifier never had any "retreat" signals go through it again (negative feedback signal through the main amplifier).
Just 2 of very many ideas.
It is nice that we have so many amplifier topologies and philosophies to choose from.
(Forward Facing signals that do not have any "retreat" of the signal (no negative feedback loops).
Quad had a topology called Feed Forward. It had an amplifier, with its usual imperfections, just like all amplifiers.
They took the signal output of that amplifier, and sent it to a lower power "corrections amplifier".
The two amplifiers output's were combined.
That topology meant that the main power amplifier never had any "retreat" signals go through it again (negative feedback signal through the main amplifier).
Just 2 of very many ideas.
It is nice that we have so many amplifier topologies and philosophies to choose from.
I suspect that they're part of a plan to bias the 6AU6's with larger than normal values for R12/R13, with the intention of keeping the values of R19/R18 up to a tolerable value. They would otherwise become a significant wasted load on the output valves. Because the feedback path is DC coupled (yeah! done correctly!) getting all these values to play nice together requires some tricks and treats. The inadverdant small pos. feedback is not intentional, just the fog of war.
All good fortune,
Chris
I really like the RCA SP-10 design too. The "local" N Fdbks back to the driver cathodes don't load down the driver input grids. So the front end can be played with for different designs readily. Driver plates don't get loaded down. DC coupling of the N FDBK is a plus. Single pole loading of the "local" N FDBK loop. This Amp does everything right. I would have thought R9 and R10 would be the usual bootstrapping of the 6AU6 grid resistors, don't know what is happening there. For the 9K OT availability issue, you might try the "bigger 6V6", the 6JC5 with an 8K or 6K load. Since a little more current will be needed to drive the lower Z OT then, a little boost in driver gain could help, or go to the higher gm version of the 6JC5, the 6JA5. (or just put some JJ KT77 tubes in there, this Amp could use a power upgrade! And some 6197 tubes in for the 6AU6s, 2X the gm, can call it the RCA SP-35 )
Last edited:
Oooooh, nice clues!
Funny you should mention 6JC5. At one point I found some NOS ones for cheap and got a few.
I don't have any 6197. What else would serve as an upgrade from 6AU6? How about 6AC7? That has about twice the gm of 6AU6, and I have a lot of 'em. Or 6CF6? How about 6EJ7?
KT77 is very similar to EL34, correct?
EDIT TO ADD:
Question about the 6AU6 drivers. What's the idea with the 1M screen load resistor? That large a value brings the 6AU6 screen voltage down to about 65V, which puts the 6AU6 in a low gm area of operation. If higher gm is what is wanted here, why not readjust everything so that the driver pentodes are operating with a higher screen voltage and therefore higher gm?
Funny you should mention 6JC5. At one point I found some NOS ones for cheap and got a few.
I don't have any 6197. What else would serve as an upgrade from 6AU6? How about 6AC7? That has about twice the gm of 6AU6, and I have a lot of 'em. Or 6CF6? How about 6EJ7?
KT77 is very similar to EL34, correct?
EDIT TO ADD:
Question about the 6AU6 drivers. What's the idea with the 1M screen load resistor? That large a value brings the 6AU6 screen voltage down to about 65V, which puts the 6AU6 in a low gm area of operation. If higher gm is what is wanted here, why not readjust everything so that the driver pentodes are operating with a higher screen voltage and therefore higher gm?
Last edited:
Plate current through the 6AU6's is about half a milliAmp (call it 150VDC / 330K Ohm) so cathode voltage is 1K8 x 0.5mA = 0.9 VDC (maybe 1.0 VDC including G2 current). 3/4 of that DCV appears at the G1, so actual bias voltage is only about 0.25 VDC, kinda small. To take advantage of the bigger vertical pentodes might likely require quite a bit more current. This would be a generally good thing anyway, for driving the capacitances appearing at the output valves' G1 and allowing smaller output valve grid leak resistors for best valve lifetime.
If the R9/R10 solution is too sketchy for you (can't blame ya) a small positive DCV applied to the bottom of the drivers' grid leak resistors, a few volts, maybe even from a battery, would do the same job, and allow the same design freedom in choosing R12/R13 and R19/R18. Could even keep them the same values, seem well chosen. Use nice resistors for R19/R18, maybe several in series, and maybe wirewound or something else with plenty of dissipation (about a Watt of heat) and best voltage linearity (big signal voltages here).
All good fortune,
Chris
If the R9/R10 solution is too sketchy for you (can't blame ya) a small positive DCV applied to the bottom of the drivers' grid leak resistors, a few volts, maybe even from a battery, would do the same job, and allow the same design freedom in choosing R12/R13 and R19/R18. Could even keep them the same values, seem well chosen. Use nice resistors for R19/R18, maybe several in series, and maybe wirewound or something else with plenty of dissipation (about a Watt of heat) and best voltage linearity (big signal voltages here).
All good fortune,
Chris
6CB6A, 6EW6, 6BN11, 6J11 to name few I've seen, or what you mentioned. The bigger 6197 (an optimized 6CL6 with beautiful curves) is cheap, like $3. I got a big sealed box of them, Mil rated, new a few years ago for $ 0.35 each, so I have 6197 on the brain now, everything needs a 6197 ! The Mil GE 6197s I got are virtually matched right out of the sealed box (200 of them!)
6HZ8 is another upgraded 6CL6 with a mu 70 triode included. Might be useful for the front end.
If the 6AU6 is running at such low current, then just increasing current as you said. RCA'S later 50 Watt Amp has 120V on the 6CB6A screen grids.
I wonder if the SP-10 was using such low 6AU6 current to get some Harmonic Cancellation with the 6V6s.
Yeah, KT77 is plug-in replaceable with EL34. The KT77 was supposedly designed for UL if you want to try UL, but works just fine as a Beam Pent. KT77 has better triode curves too. The KT77 B. pentode curves look just like most pentodes with rounded knees.
6HZ8 is another upgraded 6CL6 with a mu 70 triode included. Might be useful for the front end.
If the 6AU6 is running at such low current, then just increasing current as you said. RCA'S later 50 Watt Amp has 120V on the 6CB6A screen grids.
I wonder if the SP-10 was using such low 6AU6 current to get some Harmonic Cancellation with the 6V6s.
Yeah, KT77 is plug-in replaceable with EL34. The KT77 was supposedly designed for UL if you want to try UL, but works just fine as a Beam Pent. KT77 has better triode curves too. The KT77 B. pentode curves look just like most pentodes with rounded knees.
Last edited:
We're (my build partner itishifi.com does all the work; I just pester) currently building an amp with 6J11 driver (a compactron 12 pin with, supposedly, a pair of 6EW6 inside) but we're going to have to cut-and-try the drivers' cathode resistor because that smoking-amp guy hasn't, that I've seen, published his magical triode curves for the 6EW6. Not the end of civilization, but Hint Hint. Nudge, nudge, wink, wink.
Also, in your honor, they'll be driving trioded 6HJ5's. And using trobbins' version of the "yellow sheet mod". I'm constantly surprised by the wealth of real knowledge that bubbles up on diyAudio.
Much thanks, as always, to all,
Chris
Also, in your honor, they'll be driving trioded 6HJ5's. And using trobbins' version of the "yellow sheet mod". I'm constantly surprised by the wealth of real knowledge that bubbles up on diyAudio.
Much thanks, as always, to all,
Chris