Actually yes, or the 6SN7 or 6FQ7 front end, in which case THD is lower.
The thing is, the THD is really really low from the SP-10, especially H3. Its THD is as low as the Engineer's Amp.
However, the SP-10 has more H2 than H3, whereas the Engineer's Amp has more H3 than H2.
So the question is, considering the current-starved driver stage, why does it look like it should be so clean? It should be all wrong, right?
The thing is, the THD is really really low from the SP-10, especially H3. Its THD is as low as the Engineer's Amp.
However, the SP-10 has more H2 than H3, whereas the Engineer's Amp has more H3 than H2.
So the question is, considering the current-starved driver stage, why does it look like it should be so clean? It should be all wrong, right?
When an output tube turns on, the N Fdbk increases the associated driver current, right? So how high does that driver stage current peak at, when the output tube saturates? Only when an output is on does the associated driver matter. So it's not really starved when it's functioning.
One way to think about it is that in a feedback amplifier if the driver is still capable of even trying to correct an overload, it will. Looking at the output of the driver, with feedback connected, when the output stage clips, the driver tries extra hard to push it beyond clipping, so makes a nipple. If the driver is incapable of any more correction, it flat tops.
All good fortune,
Chris
Yeah, the class AB driver thing is bound to be a red herring. But a strange and interesting thought.
All good fortune,
Chris
The class aB driver thing may still be causing the -product- of the driver gm and the output tube gm to be staying more or less constant. Effectively making the pair into a fairly linear channel with high N Fdbk besides. Seems like a good way to get low distortion. Two linearizng effects in combo. This seems to be a revelation in effective driver design to me. Most driver stages just have the current cranked up so driver gm stays high all the time.
This would be impractical to tune for widely different output tubes with the SP-10 biasing issue I suspect, (maybe an adjustable R divider ratio in the biasing could work? ) a tuneable CCS tail might also make this scheme practical? The CCS tail would be set to approximately null the DC Fdbk currents sum, with a pot. adjustment. Simulation needed I think.
I think figuring out how this SP-10/20 driver scheme is working may well be a new "breakthru" for driver designs. (well, a re-discovery of past genius )
This would be impractical to tune for widely different output tubes with the SP-10 biasing issue I suspect, (maybe an adjustable R divider ratio in the biasing could work? ) a tuneable CCS tail might also make this scheme practical? The CCS tail would be set to approximately null the DC Fdbk currents sum, with a pot. adjustment. Simulation needed I think.
I think figuring out how this SP-10/20 driver scheme is working may well be a new "breakthru" for driver designs. (well, a re-discovery of past genius )
Last edited:
Yeah, but I'd really argue that the output stage, working into a real speaker load is too different from the model to trust. Of course, God and the Devil dwell together in the details. Or possibly, I'm just too old-fashioned. Naw, can't be that.
All good fortune,
Chris
All good fortune,
Chris
OOPs, the DC Fdbk increases the driver gm as the output gm is increasing. So its NOT a constant product, but rather a tracking set of gm's. So the N Fdbk increases as the tube gets turned on. Hmmm, interesting, but does that help? Tube gm increases roughly as the Sqrt of current. With both tube types current tracking, the overall pair gm would then be increasing linearly with output current. I don't follow that being real useful, except maybe up near saturation.
Curious. Probably explains the exaggerated spikes near saturation anyway. Apparently RCA gave up on the scheme in later Amps.
So where does the driver current peak up to at the peaks? Probably should just use that current for the nominal idle.
Curious. Probably explains the exaggerated spikes near saturation anyway. Apparently RCA gave up on the scheme in later Amps.
So where does the driver current peak up to at the peaks? Probably should just use that current for the nominal idle.
Last edited:
This old relic is interesting to us both for its deeply correct topo and for its 1950s Strange New World zeitgeist, where Best Practice wasn't yet engraved so deep that operating driver valves at less than half a milliAmp was still Why Not? Today we "know better", and we really do, but they were making it all up as they went, and in real time.
The Paul Klipsch amplifier of 1945 was similarly driven and had the option of the same feedback, but that was not considered necessary with 2A3/6B4/6A5 output valves. First stage was just a 6J5 split-load inverter, drivers were 6SJ7.
I'd bet that the shape of the nipple shown at driver output / output valve G1, at output valve overload, could be pretty well predicted by a modeling program that included output valve grid current and the RC time constants around G1. This could be tested by varying frequency and by observing G1 voltage in comparison to driver plate voltage (voltage across the coupling cap). But whatever, an amplifier's output shouldn't ever be limited by the drivers. Low mu output stages, great in their own ways, make great drive voltage demands on drivers, if we demand that driver B+ be derived from output valves' B+. Somewhere around twice as much B+ for drivers sounds about right for a mu of 4 or 5 output stage. We never do this because we're stuck in a 1950s zeitgeist.
The RCA takes advantage of triode-mu-of-8(or something) output valves, is certainly brilliant, but looks twitchy to us now, relying on special, exotic (to our eyes) valve characteristics from off the charts, that we know to be in very poor performance regions, highly dependent on the particular characteristics of individual valves. Not a modern engineering viewpoint. Still, way exotic and cool, and the topo is sadly neglected these days. Just needs the JC touch, or even a touch of HL.
All good fortune,
Chris
The Paul Klipsch amplifier of 1945 was similarly driven and had the option of the same feedback, but that was not considered necessary with 2A3/6B4/6A5 output valves. First stage was just a 6J5 split-load inverter, drivers were 6SJ7.
I'd bet that the shape of the nipple shown at driver output / output valve G1, at output valve overload, could be pretty well predicted by a modeling program that included output valve grid current and the RC time constants around G1. This could be tested by varying frequency and by observing G1 voltage in comparison to driver plate voltage (voltage across the coupling cap). But whatever, an amplifier's output shouldn't ever be limited by the drivers. Low mu output stages, great in their own ways, make great drive voltage demands on drivers, if we demand that driver B+ be derived from output valves' B+. Somewhere around twice as much B+ for drivers sounds about right for a mu of 4 or 5 output stage. We never do this because we're stuck in a 1950s zeitgeist.
The RCA takes advantage of triode-mu-of-8(or something) output valves, is certainly brilliant, but looks twitchy to us now, relying on special, exotic (to our eyes) valve characteristics from off the charts, that we know to be in very poor performance regions, highly dependent on the particular characteristics of individual valves. Not a modern engineering viewpoint. Still, way exotic and cool, and the topo is sadly neglected these days. Just needs the JC touch, or even a touch of HL.
All good fortune,
Chris
How about trying a reduction in g2 voltage whilst leaving plate current and plate loading alone? This will of course limit its voltage output eventually, and will likely require an adjustment of the voltage divider( 1M and 330k ) that is setting the operating point with the cathodes elevated by the DC coming through the nFBK network... 🙂
Douglas
Is it just possible that the FB is positive FB? If everyone's favorite local type ''Shade'' is NFB going from plate back to plate, then isn't the signal going into the cathode instead going to be (+)?
I have a chassis ready to go, and a pair of Stancour outputs from a Heath kit AA 100….. 8k to 0-4-8-16 ohm…
I think I’m gonna build one of these just like the schematic, and see how it does. I have some new old stock 6v6gt and about 200 6au6😎
I think I’m gonna build one of these just like the schematic, and see how it does. I have some new old stock 6v6gt and about 200 6au6😎
A rising signal on the 6AU6 grid causes a rising plate voltage on its driven 6V6 plate. The rising 6V6 plate/6AU6 cathode works to shrink the g1-k voltage of the 6AU6. IOW, it is indeed nfb... 🙂
Douglas
Let's look at it by comparing the actual phase of the signal. If you reduce the FB resistor to increase the FB signal to a much higher level so that it swamps the grid and goes above its level, then what do you call it?
What I'm trying to figure out is how there can be a complete sine waave at the grid of 6V6 if the driver is indeed running near class B or deep AB. The sine would be truncated, but instead it's shape is more like a complete Class A wave. I'm thinking some (+) FB is taking over when the driver would be in cutoff, and keeps it humming along. Or else the driver is indeed always biased for class A, and the biasing theory of AB,B is wrong.
You could bypass the driver's K and see what it sends to the outputs. That would keep the biasing in place. You could also untie it from the plate and bootstap it over to the B+. ...yes?
The driver sure looks like Class A to me. For sure it can swing +/- 100V of the 125V that are shown dropped across its plate loads( and also used for the .37 mA idle current calc).
Douglas
Was getting too late to think correctly last night in post 107.
The N Fdbk does TRY to turn on the 6AU6 when the output 6V6 is turning on, but the neg. input to the 6AU6 grid is overcoming that in order to actually turn on the 6V6. So the tubes ARE operating in complementary fashion. gm product could be staying more or less stable. Gain for the combo could be stabilized.
But why doesn't the 6AU6 run out of gas when the output 6V6 is fully turned on?
I guess the 6AU6 plate load resistor can always turn on the 6V6 as long as it's a low enough R to B+ to pull up the 6V6 grid resistor. So the 6AU6 is running in a starved class A.
This could easily run into slew rate problems near output tube full turn on, especially for bigger tubes or UL or Triode outputs.
Those 6AU6 load resistors would need lowering, and the driver current increased for more difficult output tubes.
Seems to me that the gm product stabilization can still work with higher current in the driver, just a bigger number for the gm product. Might as well design the driver for more current to run bigger outputs. Still working for small outputs. ( DC bias issue handled by a CCS tail on the drivers. That does sacrifice the independence of the two driver tubes however. )
Maybe the special sound comes from slew rate limiting of high level HF peaks? Wouldn't work well for drums.
Hmm, the CCS tail may not regulate well against Fdbk resistors coming all the way from B+. (which would be higher for bigger tubes possibly ) Maybe not working for bias control.
The N Fdbk does TRY to turn on the 6AU6 when the output 6V6 is turning on, but the neg. input to the 6AU6 grid is overcoming that in order to actually turn on the 6V6. So the tubes ARE operating in complementary fashion. gm product could be staying more or less stable. Gain for the combo could be stabilized.
But why doesn't the 6AU6 run out of gas when the output 6V6 is fully turned on?
I guess the 6AU6 plate load resistor can always turn on the 6V6 as long as it's a low enough R to B+ to pull up the 6V6 grid resistor. So the 6AU6 is running in a starved class A.
This could easily run into slew rate problems near output tube full turn on, especially for bigger tubes or UL or Triode outputs.
Those 6AU6 load resistors would need lowering, and the driver current increased for more difficult output tubes.
Seems to me that the gm product stabilization can still work with higher current in the driver, just a bigger number for the gm product. Might as well design the driver for more current to run bigger outputs. Still working for small outputs. ( DC bias issue handled by a CCS tail on the drivers. That does sacrifice the independence of the two driver tubes however. )
Maybe the special sound comes from slew rate limiting of high level HF peaks? Wouldn't work well for drums.
Hmm, the CCS tail may not regulate well against Fdbk resistors coming all the way from B+. (which would be higher for bigger tubes possibly ) Maybe not working for bias control.
Last edited:
Needs a gyrator version of a CCS for the driver tail. Maintains a voltage but still high Z. by slow adjusting the maintained current level.
I guess there is enough room down at the bottom of the curves for that if the effective bias is 2v. But my sheet shows that OP as 2mA @ 200v, but when I calculate the the currect through the Rk at 3.6mA and then look at the series current coming down the FB line to it calculated to 3.2mA, that leaves only .4mA from the tube idle, not 2mA. So where is this figuring going wrong or are the voltages on the schematic not true? If the real idle current is .4mA it's got no room to swing. And this is for a tube that the sheet says is typically set up to idle at 8mA - 11mA. .... to me this is pointing to being just another ''esoteric'' implementation, you play with for the sake of killing time.
Last edited:
The driver plate resistors can still pull up the output tube grid V even though the driver is shutting off. The driver just loses control with no gm left. Slew limiting at peaks.
Kind of like letting your car's steering run off the rack gear at max turning rate. Keeps the output tubes from slamming into grid2 current excess anyway.
I guess this prevents hard slamming into saturation, even though the output tube's gm is maximized there. So they have a soft landing technique working for high N Fdbk.
Kind of like letting your car's steering run off the rack gear at max turning rate. Keeps the output tubes from slamming into grid2 current excess anyway.
I guess this prevents hard slamming into saturation, even though the output tube's gm is maximized there. So they have a soft landing technique working for high N Fdbk.
Last edited: