The main thing I'd be concerned about is the totally foxxed up design of the power supply. 82uF across a 5Y3 is way, way, way too much. The Isurge= 440mA. When doing a design that used the beefier 5U4GB (Isurge= 1.0A) even 47uF was a spec buster. Your 5Y3's aren't gonna last very long. I've seen this done with commercial rigs, these excessive filter capacitors. It's cheaper than a proper PS with an RLC LPF, and they were counting on the average consumer's not being suspicious about the frequency they were replacing power diodes back in the day where nearly every pharmacy, convenience, and grocery store had those tube testers.
Get that reservior capacitor to a reasonable size that doesn't bust the Isurge spec, and replace that RC filter with an RLC LPF. You'll get better ripple suppression and improved voltage regulation.
Thanks Miles. I definitely intend to improve the PSU. I'd design it from scratch, using UF4007 rectifiers, etc. No 5Y3 and no crazy expensive 5AR4 or 5U4GB either. I might consider using a 5R4GY, but they're so large they wouldn't fit on the chassis. So it'll be solid state rectification for sure.
I have an Edcor power transformer, 500VCT 250mA. That should provide raw B+ of 300VDC with enough current delivery for a stereo pair of this circuit using 6P43P/EL86 outputs.
I have a Hammond filter choke rated for 300mA. I think its inductance is low, but I also have a stash of 800uF 330V electrolytic capacitors to make up for that. I also have a couple of 450uF 475V electrolytics. I even have some 75uF 400V film caps. 🙂
I think I will make a +190VDC (or thereabouts) feed coming off the main B+. for the 6P43P screens. So far I got there in simulations using a 15k screen dropping resistor and a 10uF cap. But I know a regulated screen supply should perform much better. Is use of a capacitance multiplier adequate for the screen supply, or does it really need to be fully voltage regulated? I'm thinking about where to put heatsinks...
1. EL86 Triode Wired Versus EL84 triode wired . . .
They have nearly the same Gm.
The EL86 has less than 1/2 the g2/g1 u (mu) of the EL84
rp = u/Gm
So . . .
The EL86 has slightly less than half the plate impedance, rp, versus the EL84 rp.
rp . . . 600-800 Ohms, versus 1700 Ohms.
2. Pentode operation, EL86 versus EL84 . . .
They both have high plate impedance, rp, it is Much larger than the impedance of the plate to center tap of the output transformer.
At AB1 operation, part of the signal swings, only one tube runs the output transformer, and the 1/2 primary is 1/4 of the primary p-p impedance.
Since the plate impedance is so large versus the primary impedances . . .
Schade, Cathode, or Global negative feedback is one of two that can "fix" that difficult plate to primary impedance ratio.
The only other help, is to reduce that ratio by confining the operation to Class A1 (makes the problem only 1/2 as bad).
That means that for the EL84, you need 6dB more negative feedback than for the EL86.
6 dB more negative feedback, with all the problems that it potentially brings.
3. I will leave the Ultra Linear analysis un-discussed, except to say that rp, is between the impedance of pentode versus triode wired numbers.
You can draw your own conclusions from that.
$0.03
They have nearly the same Gm.
The EL86 has less than 1/2 the g2/g1 u (mu) of the EL84
rp = u/Gm
So . . .
The EL86 has slightly less than half the plate impedance, rp, versus the EL84 rp.
rp . . . 600-800 Ohms, versus 1700 Ohms.
2. Pentode operation, EL86 versus EL84 . . .
They both have high plate impedance, rp, it is Much larger than the impedance of the plate to center tap of the output transformer.
At AB1 operation, part of the signal swings, only one tube runs the output transformer, and the 1/2 primary is 1/4 of the primary p-p impedance.
Since the plate impedance is so large versus the primary impedances . . .
Schade, Cathode, or Global negative feedback is one of two that can "fix" that difficult plate to primary impedance ratio.
The only other help, is to reduce that ratio by confining the operation to Class A1 (makes the problem only 1/2 as bad).
That means that for the EL84, you need 6dB more negative feedback than for the EL86.
6 dB more negative feedback, with all the problems that it potentially brings.
3. I will leave the Ultra Linear analysis un-discussed, except to say that rp, is between the impedance of pentode versus triode wired numbers.
You can draw your own conclusions from that.
$0.03
In this particular circuit, the EL86 pentode (or EL84, or 6JC5, or whatever pentode we choose) has a local NFB loop from its plate to the cathode of the preceding driver stage. I think this combines the output pentode and driver pentode into a sort of composite device acting like a triode with low rp but higher gain than would be realized by simply shorting the pentode's screen grid to its plate. O.H. Schade showed this in his original 1930's paper on wrapping local feedback around a 6L6 from plate to grid (making that 6L6 into an anode follower, basically).
See Fig. 33(a) on pg. 360 of the paper - https://www.dos4ever.com/uTracer3/Schade.pdf
The resulting output waveforms are shown in Fig. 34 on pg. 362.
Interesting, right?
In Fig. 35 the curves of the beam pentode with local NFB applied look like what you'd get from a triode. The apparent rp of the 6L6 beam pentode with 10% NFB (n = 0.1) applied is reduced to 2k ohms and the mu is 10.
Again, I think that if you take the more commonly-used approach of running the pentodes open loop into the OPT primary, but then wrap NFB from the speaker output (OPT secondary) to the input stage voltage amplifier (global NFB), then the whole amplifier becomes a sort of composite device of a far more complex mix of ingredients.
Hopefully I'm not jumping to a bunch of erroneous conclusions due to my lack of understanding of physics and more advanced concepts of electronics.
Feel free to steer me into the light if I've veered off into the darkness.
See Fig. 33(a) on pg. 360 of the paper - https://www.dos4ever.com/uTracer3/Schade.pdf
The resulting output waveforms are shown in Fig. 34 on pg. 362.
Interesting, right?
In Fig. 35 the curves of the beam pentode with local NFB applied look like what you'd get from a triode. The apparent rp of the 6L6 beam pentode with 10% NFB (n = 0.1) applied is reduced to 2k ohms and the mu is 10.
Again, I think that if you take the more commonly-used approach of running the pentodes open loop into the OPT primary, but then wrap NFB from the speaker output (OPT secondary) to the input stage voltage amplifier (global NFB), then the whole amplifier becomes a sort of composite device of a far more complex mix of ingredients.
Hopefully I'm not jumping to a bunch of erroneous conclusions due to my lack of understanding of physics and more advanced concepts of electronics.
Feel free to steer me into the light if I've veered off into the darkness.
Last edited:
This FB circuit plate to K (preceding stage) isn't the same as ''Schade'' plate to plate >(output grid) FB. The question is, what is the polarity of the FB if it goes to the K instead of the plate of the preceding stage? Is not the K signal inverted vs. the plate? So if plate to plate FB is ''inverse'' FB, what is it at the K?
20to20,
If the pentode output tube was treated like a solid state inverting op amp, then negative feedback would be with a resistor from the plate to the grid (Rf), and the Impedance of the circuit that drove Rf, would be Ri. Ri is the driver tube plate rp, and that is in parallel with the driver plate load RL.
For easy, non DC coupled stages, we remove Rf from the output tube grid; and now we connect it to the plate of the driver tube (just 1 RC coupling away from the output tube grid. Schade negative feedback.
The problem with this, is that the negative feedback from Rf, puts a low impedance load on the driver tube plate.
An alternate method of negative feedback to the driver tube:
When you Drive a signal into the Cathode of a driver, then the Plate of the driver is IN PHASE with the Cathode (not inverted phase).
Driving the tube Cathode, and taking the output from the plate, is called a grounded grid / common grid stage.
Try the output tube plate to driver tube cathode, you might like it.
The easier way [lazier way] is to put a capacitor in series with Rf, so that the plate DC voltage does not affect the driver tube cathode bias (DC coupling requires more difficult and creative way to maintain the DC bias of the driver tube).
Start with a grid at 0V, the cathode at +5V, and the plate at 150V at the plate end of the load resistor RL, with the other end of RL at +300V
Now, apply a +1V signal to the cathode. The tube current goes down, and so the current in RL is also reduced, so the voltage across RL is reduced.
That means the plate voltage goes up, just like the cathode voltage going up from 5V to 6V..
The phase inversion of a triode stage is from grid to plate; It is not from cathode to plate.
If the pentode output tube was treated like a solid state inverting op amp, then negative feedback would be with a resistor from the plate to the grid (Rf), and the Impedance of the circuit that drove Rf, would be Ri. Ri is the driver tube plate rp, and that is in parallel with the driver plate load RL.
For easy, non DC coupled stages, we remove Rf from the output tube grid; and now we connect it to the plate of the driver tube (just 1 RC coupling away from the output tube grid. Schade negative feedback.
The problem with this, is that the negative feedback from Rf, puts a low impedance load on the driver tube plate.
An alternate method of negative feedback to the driver tube:
When you Drive a signal into the Cathode of a driver, then the Plate of the driver is IN PHASE with the Cathode (not inverted phase).
Driving the tube Cathode, and taking the output from the plate, is called a grounded grid / common grid stage.
Try the output tube plate to driver tube cathode, you might like it.
The easier way [lazier way] is to put a capacitor in series with Rf, so that the plate DC voltage does not affect the driver tube cathode bias (DC coupling requires more difficult and creative way to maintain the DC bias of the driver tube).
Start with a grid at 0V, the cathode at +5V, and the plate at 150V at the plate end of the load resistor RL, with the other end of RL at +300V
Now, apply a +1V signal to the cathode. The tube current goes down, and so the current in RL is also reduced, so the voltage across RL is reduced.
That means the plate voltage goes up, just like the cathode voltage going up from 5V to 6V..
The phase inversion of a triode stage is from grid to plate; It is not from cathode to plate.
Thanks for the discussion and clarifications!
I wasn't trying to state that output plate to driver cathode NFB is the same thing as output plate to grid NFB.
However, many have pointed out that the net result is exactly the same in practice, aside from how plate-grid NFB loads down the driver stage plate, So I'm running with that.
I've seen some posts where it's stated that cathode feedback to the output tube cathode from the OPT secondary works exactly like 'Schade' feedback. I don't have an opinion on that.
My point was that I think that in this RCA SP-10 output stage, the output tube and driver tube have NFB connecting them in such a way that the two become inseparable. That's similar to how 'Schade' (plate to grid) NFB works. The output pentode is transformed into a triode-like device. I've always been intrigued by that. A 'triode' with pentode power? Yup, that's what I want.
A nice thing is that the level of NFB is adjustable in these circuits by changing the value of Nfb in relation to the 'bottom leg' of its voltage divider.
Another nice thing is that the reactive OPT is not inside the NFB loop, so I don't have to figure out how to compensate for that. Or at least, I think that's right... I hope that's right....
[ADDED] -
When you run that feedback resistor from the plate of the output tube to the plate of the driver tube, you're still doing plate-to-grid feedback around the output tube with AC audio signal.
Some have experimented with simplifying down to making the feedback resistor the plate load resistor for the driver stage. Doug (bandersnatch) did that with his E-Linear Stereo 70 mod (6AU6 LTP into EL34 UL with screen-to-grid feedback around the EL34s). I can't find that schematic at the moment...
I wasn't trying to state that output plate to driver cathode NFB is the same thing as output plate to grid NFB.
However, many have pointed out that the net result is exactly the same in practice, aside from how plate-grid NFB loads down the driver stage plate, So I'm running with that.
I've seen some posts where it's stated that cathode feedback to the output tube cathode from the OPT secondary works exactly like 'Schade' feedback. I don't have an opinion on that.
My point was that I think that in this RCA SP-10 output stage, the output tube and driver tube have NFB connecting them in such a way that the two become inseparable. That's similar to how 'Schade' (plate to grid) NFB works. The output pentode is transformed into a triode-like device. I've always been intrigued by that. A 'triode' with pentode power? Yup, that's what I want.
A nice thing is that the level of NFB is adjustable in these circuits by changing the value of Nfb in relation to the 'bottom leg' of its voltage divider.
Another nice thing is that the reactive OPT is not inside the NFB loop, so I don't have to figure out how to compensate for that. Or at least, I think that's right... I hope that's right....
[ADDED] -
That's an anode follower if you do that with a tube (triode or pentode).
When you run that feedback resistor from the plate of the output tube to the plate of the driver tube, you're still doing plate-to-grid feedback around the output tube with AC audio signal.
Some have experimented with simplifying down to making the feedback resistor the plate load resistor for the driver stage. Doug (bandersnatch) did that with his E-Linear Stereo 70 mod (6AU6 LTP into EL34 UL with screen-to-grid feedback around the EL34s). I can't find that schematic at the moment...
Last edited:
One of the things that was mentioned often when Schade feedback became popular years ago was that the resulting low Rp helped to get surprisingly good results from experimental circuits with small/cheap OPTs, such as old tube radio transformers.
Though, no matter how low Rp we achieve through local feedback, the output impedance of the amp will never get below the Dcr of the OPT secondary winding unless additional global NFB is added.
Though, no matter how low Rp we achieve through local feedback, the output impedance of the amp will never get below the Dcr of the OPT secondary winding unless additional global NFB is added.
Good point about the the Zout not going lower than the DCR of the OPT secondary winding without global NFB including the OPT inside the loop.
But I think if I was worried about approaching technical perfection I'd be trying to build solid state amps. At this point in my life, my interest in tube amps is more about experimenting and trying to fashion a sound I like personally, not one that will please others. Of course I also need to use up some of the pile of parts I've collected over the years!!!
But I think if I was worried about approaching technical perfection I'd be trying to build solid state amps. At this point in my life, my interest in tube amps is more about experimenting and trying to fashion a sound I like personally, not one that will please others. Of course I also need to use up some of the pile of parts I've collected over the years!!!
Hear, hear!
If a very high damping factor is required by the speakers, then tube amps are generally not the best choice anyway.
If a very high damping factor is required by the speakers, then tube amps are generally not the best choice anyway.
Voltage sensing negative feedback can never reduce output impedance to (or below) zero, but current sensing positive feedback can. One method is to leave some portion of the output valve's cathode resistor unbypassed and connect that via resistor to the driver's cathode. smoking-amp famously notes that AP distortion analyzers, which have output transformers for galvanic isolation, use current sensing positive feedback for this purpose, which reduces distortion in the transformer.
Would be quite an undertaking when combined with negative voltage sensing feedback to the same cathode, but could probably be approached with LTspice.
All good fortune,
Chris
Would be quite an undertaking when combined with negative voltage sensing feedback to the same cathode, but could probably be approached with LTspice.
All good fortune,
Chris
The advantage of a Pentode in native mode is more output power.
But because the rp is high, damping factor suffers, and output transformer primary inductance can become a factor, especially where, for example, the woofer is high impedance in 2 frequency ranges that come from using a ported enclosure.
And at the woofer minimum impedance ranges, the high rp has trouble delivering power.
So, along comes negative feedback . . .
Output cathode negative feedback from a winding on the transformer may, or may not, suffer from the leakage inductance to the cathode winding.
With Schade negative feedback, leakage inductance does not have much effect on the feedback, the output transformer is not included in the feedback loop.
With Triode wiring, the negative feedback is inside the tube, and the leakage reactance is external to the tube.
But you get less power output, negating one of the features of a pentode.
You trade power for higher intrinsic damping factor; and often decide to build a no negative feedback circuit.
Tradeoffs, Tradeoffs, Tradeoffs.
Most topologies such as all the above named circuits, can be properly designed, with quality parts, and attention to other details . . .
and then, you end up with good "sounding" amplifiers.
The topology selection is up to you, but you need to carry out the implementation properly, so that you can . . .
"Enjoy the Music", I stole that phrase from a good audio equipment reviewer.
But because the rp is high, damping factor suffers, and output transformer primary inductance can become a factor, especially where, for example, the woofer is high impedance in 2 frequency ranges that come from using a ported enclosure.
And at the woofer minimum impedance ranges, the high rp has trouble delivering power.
So, along comes negative feedback . . .
Output cathode negative feedback from a winding on the transformer may, or may not, suffer from the leakage inductance to the cathode winding.
With Schade negative feedback, leakage inductance does not have much effect on the feedback, the output transformer is not included in the feedback loop.
With Triode wiring, the negative feedback is inside the tube, and the leakage reactance is external to the tube.
But you get less power output, negating one of the features of a pentode.
You trade power for higher intrinsic damping factor; and often decide to build a no negative feedback circuit.
Tradeoffs, Tradeoffs, Tradeoffs.
Most topologies such as all the above named circuits, can be properly designed, with quality parts, and attention to other details . . .
and then, you end up with good "sounding" amplifiers.
The topology selection is up to you, but you need to carry out the implementation properly, so that you can . . .
"Enjoy the Music", I stole that phrase from a good audio equipment reviewer.
Sounds familiar, I think there is a thread about this concept somewhere here.
You could add a small amount of global N Fdbk to fix up the damping factor further.
But you can also do similar with an additional local positive current Fdbk loop. A small value R in the output tubes cathodes samples current draw thru the OT.
Knowing the OT primary resistance, you can compensate it out by a weak resistive Pos Fdbk from the output tube cathodes back to the driver cathodes. This is adjusted so it puts in just enough additional drive signal to the driver to overcome the resistive V loss in the OT primary. So it's a negative resistance effectively. Since OT magnetizing current also shows up there, that gets compensated out too, so its drop across the OT primary R is eliminated. So transformer distortion from magnetizing current is eliminated too. Audio Precision has a patent on this, but expired now.
Audiophiles using gapped SE OTs need to learn about this technique. P-P can be cleaned up better than traditional SE. But all that 3rd Harmonic dist. from gapped SE OTs will be gone with this compensation.
The local N Fdbk, like in the SP-10/20, cleans up the tube distortion already, and greatly lowers output Z from the tubes. Only the OT secondary resistance remains after the current Fdbk is added. Generally pretty low.
But you can also do similar with an additional local positive current Fdbk loop. A small value R in the output tubes cathodes samples current draw thru the OT.
Knowing the OT primary resistance, you can compensate it out by a weak resistive Pos Fdbk from the output tube cathodes back to the driver cathodes. This is adjusted so it puts in just enough additional drive signal to the driver to overcome the resistive V loss in the OT primary. So it's a negative resistance effectively. Since OT magnetizing current also shows up there, that gets compensated out too, so its drop across the OT primary R is eliminated. So transformer distortion from magnetizing current is eliminated too. Audio Precision has a patent on this, but expired now.
Audiophiles using gapped SE OTs need to learn about this technique. P-P can be cleaned up better than traditional SE. But all that 3rd Harmonic dist. from gapped SE OTs will be gone with this compensation.
The local N Fdbk, like in the SP-10/20, cleans up the tube distortion already, and greatly lowers output Z from the tubes. Only the OT secondary resistance remains after the current Fdbk is added. Generally pretty low.
Last edited:
There were posts in Tubes / Valves about single ended amplifiers that used negative resistance output stages, to control the woofer.
1977 Klipsch Heresy - 7.8, 7.7
1980s Klipsch Heresy - 7.3, 7.4
1973 Kenwood KL-5050 - 5.7, 5.6
Interesting to note, the KL-5050 claim pretty much the same performance as the Heresy (efficency, FR, etc).
You can use a special topology.
It is different than single ended, and it is different than push pull.
I thought I was the first to do it, but then I found out the French had already done that.
Start with a push pull amplifier.
Remove the output transformer.
Then,
Connect one output plate to the primary of an SE air gapped transformer, and the other end of the primary to B+
And Connect the other output plate to the primary of a second SE air gapped transformer, and the other end of the primary to B+,
IMPORTANT NOTE: because the 2 plate currents are in opposite phase (one increasing, and the other reducing) you have to swap the output transformers primary plate lead and primary B+ lead.
Now,
Connect the like taps of the secondary together . . . 2-Common, 2-4 Ohm, 2-8 Ohm, and 2-16 Ohm taps.
As long as the newly created special case push pull amplifier is operated in Class A, there will always be current in each of the primary windings.
There will never be a collapse to zero, of either primary current.
Think about the results of the magnetic effects on the two sets of laminations:
Single ended output transformers, current always in one direction on one primary, and current always in the other direction on the other primary.
The effective special case Push Pull output will have symmetrical damping factor, just like real push pull.
Have Fun!
It is different than single ended, and it is different than push pull.
I thought I was the first to do it, but then I found out the French had already done that.
Start with a push pull amplifier.
Remove the output transformer.
Then,
Connect one output plate to the primary of an SE air gapped transformer, and the other end of the primary to B+
And Connect the other output plate to the primary of a second SE air gapped transformer, and the other end of the primary to B+,
IMPORTANT NOTE: because the 2 plate currents are in opposite phase (one increasing, and the other reducing) you have to swap the output transformers primary plate lead and primary B+ lead.
Now,
Connect the like taps of the secondary together . . . 2-Common, 2-4 Ohm, 2-8 Ohm, and 2-16 Ohm taps.
As long as the newly created special case push pull amplifier is operated in Class A, there will always be current in each of the primary windings.
There will never be a collapse to zero, of either primary current.
Think about the results of the magnetic effects on the two sets of laminations:
Single ended output transformers, current always in one direction on one primary, and current always in the other direction on the other primary.
The effective special case Push Pull output will have symmetrical damping factor, just like real push pull.
Have Fun!
Last edited:
I suspect that will give you pretty close to a P-P OT, except for 6X the $$$, but this is DIY after all. Would be interesting to see what happens to the sum of the OT magnetizing currents. Maybe a simulation would tell.
3rd Harmonic is either expansive or compressive, so summing similars would be expected to give similars. Except the inductive magnetizing current is 90 degrees out of phase for each 180 deg. OT, Cancellation?
Last edited:
smoking-amp,
You are mostly correct, very close, but "No Cigar".
Consider:
1. Many of the push pull transformer's lamination's magnetic domains line up alternately in two directions, according to the push winding and pull winding reversing the total magnetic flux direction from those two coils.
One coil has more current than the other coil; then they have equal current; then the other coil has more current than the first coil.
The sum total of the two coils magnetic flux Reverses Direction.
2. With the 2 SE transformers, more of the lamination's magnetic domains do not alternately line up in two directions, more of their domains always line up in one direction. Each coil and its associated laminations only have one flux direction, during the complete signal polarity variations.
3. Can you see the possibility of a different affect on the push pull laminations, versus the effect on the SE laminations?
4. Note, In normal operation of the push pull transformer, and in normal operation of the 2 single ended transformers . . .
many of the magnetic domains do their own thing, and do not line up in one direction, or in the other direction, but instead are in random angles of 360 degrees, in all three dimensions of the laminations.
5. I went back and edited my post # 198 after you read it.
You are mostly correct, very close, but "No Cigar".
Consider:
1. Many of the push pull transformer's lamination's magnetic domains line up alternately in two directions, according to the push winding and pull winding reversing the total magnetic flux direction from those two coils.
One coil has more current than the other coil; then they have equal current; then the other coil has more current than the first coil.
The sum total of the two coils magnetic flux Reverses Direction.
2. With the 2 SE transformers, more of the lamination's magnetic domains do not alternately line up in two directions, more of their domains always line up in one direction. Each coil and its associated laminations only have one flux direction, during the complete signal polarity variations.
3. Can you see the possibility of a different affect on the push pull laminations, versus the effect on the SE laminations?
4. Note, In normal operation of the push pull transformer, and in normal operation of the 2 single ended transformers . . .
many of the magnetic domains do their own thing, and do not line up in one direction, or in the other direction, but instead are in random angles of 360 degrees, in all three dimensions of the laminations.
5. I went back and edited my post # 198 after you read it.
Last edited: