Hi RunEight,
Thank you muchly; this looks pretty good!
Tim,
If fed to the more sensitive 6SL7, probably running around -1V of bias, I'd be crossing three tubes, not two, and the poles would start to get tricky to manage.
I'd probably use something like a 6CG7 (nine pin 6SN7) with a mu around 20. Two of these, each with Av of 10, would give the OLG of 100 I'm after, but the question is, would feedback to the cathode of the first of these two triodes permit 80Vpp into that triode's grid?
If yes, then I'd precede the two triodes with a tube input stage to add the sonic overlay. Of course, any sonic overlay of triodes two and three would be utterly lost in the NFB; alternatively we could give triodes two and three a gain of say 5, and the input triode a gain, with no feedback, of say 5 again, giving us our gain of 26dB or so.
I'm sorry I've not been quick on this; it's a tricky design brief.
Cheers,
Hugh
Thank you muchly; this looks pretty good!
Tim,
If fed to the more sensitive 6SL7, probably running around -1V of bias, I'd be crossing three tubes, not two, and the poles would start to get tricky to manage.
I'd probably use something like a 6CG7 (nine pin 6SN7) with a mu around 20. Two of these, each with Av of 10, would give the OLG of 100 I'm after, but the question is, would feedback to the cathode of the first of these two triodes permit 80Vpp into that triode's grid?
If yes, then I'd precede the two triodes with a tube input stage to add the sonic overlay. Of course, any sonic overlay of triodes two and three would be utterly lost in the NFB; alternatively we could give triodes two and three a gain of say 5, and the input triode a gain, with no feedback, of say 5 again, giving us our gain of 26dB or so.
I'm sorry I've not been quick on this; it's a tricky design brief.
Cheers,
Hugh
How's this cheezily-drawn (note excess or absence of coupling caps, you can figure out the biasing and stuff) idear. The sand hanging off the plate looks weird but think of that tube as a CF with the loads in the plate. NFB is wrapped around both tubes and the SSatanic part as needed. No signal loss between 1st K and output so it will indeed be a follower, voltage being limited to +V capability (uh, headroom) and wattage of the cathode resistor that output V appears across. 
Tim
Tim
Attachments
Well, I guess I am confused!
Tim, if the BJT here represents a more complex output stage (or even if not), then isn't this a fairly conventional global feedback loop into the cathode of the first stage?
Or if preceeded by a gain stage, then a hybrid unity gain power buffer with local NFB?
In which case, the way to do it is very well understood and the cascaded stage can be put together a number of different ways depending, as you say, on fiddling with coupling caps, etc.
OTOH, I really did think we were talking about an isolated, local NFB unity gain buffer that could swing 100vpp independently of what preceeds or follow it.
Hugh, which problem are we trying to solve?
Tim, if the BJT here represents a more complex output stage (or even if not), then isn't this a fairly conventional global feedback loop into the cathode of the first stage?
Or if preceeded by a gain stage, then a hybrid unity gain power buffer with local NFB?
In which case, the way to do it is very well understood and the cascaded stage can be put together a number of different ways depending, as you say, on fiddling with coupling caps, etc.
OTOH, I really did think we were talking about an isolated, local NFB unity gain buffer that could swing 100vpp independently of what preceeds or follow it.
Hugh, which problem are we trying to solve?
Look closely at the NFB. It *is* a follower! The 1st cathode must track the output voltage, thus signal input must be Vin + Vo, where Vin is the actual drive the 1st tube sees.
Indeed it is a common NFB arrangement. But most such arrangements leave some residual gain for line-level inputs; that's what your 6SL7 is for.
Tim
Indeed it is a common NFB arrangement. But most such arrangements leave some residual gain for line-level inputs; that's what your 6SL7 is for.
Tim
Hi Runeight,
Yes, that's precisely it!
Goal: To create a tube buffer to drive a SS PP Class AB output stage with sufficient NFB to smooth the SS non-linearities and crossover disjunction.
Why unity gain? Because I'm trying to:
1. Maximize OLG from just two medium mu triodes,
2. Absorb all this gain in the Triode1/Triode2/SS OP stage so as to reduce distortion to virtually unmeasurable levels.
3. Add on an SRPP front end (high detail, good drive, excellent overload), OUTSIDE the NFB loop, so as to add the sonic overlay of a triode.
The reason I wanted to use tubes in this buffer is that every attempt to use SS as voltage amplification is flawed sonically because of the very high, poor quality parasitics of nearly every transistor I've ever used in common emitter. This always mandates lag compensation, which is not required with a tube buffer. It is arguable that tubes might be best, in fact, because it should be possible to do this with mosfets/fets very neatly too.
I like your 6pi45 circuit, Runeight, looks very clever. I suspect it could be as easily done with a bipolar or mosfet diff pair, and this would bring voltages down considerably, which could be an advantage as it will eliminate the current sink.
Tim, your circuit is a conventional, cascaded circuit block and I'm well familiar with it. Your addition of an emitter follower buffer at the end is probably not required in this application as the drive to a double emitter follower is high impedance and can be effected capacitively straight from the plate. My doubts revolve around its overload capacity; I'm sure it may not take 80Vpp at the input.
Since the circuit is hybrid, there is no reason not to use the whole gamut of devices; triodes, pentodes, jfets, mosfets, bipolars. It could be quite an exotic mix!
Thank again guys. This is very interesting!
Cheers,
Hugh
Yes, that's precisely it!
Goal: To create a tube buffer to drive a SS PP Class AB output stage with sufficient NFB to smooth the SS non-linearities and crossover disjunction.
Why unity gain? Because I'm trying to:
1. Maximize OLG from just two medium mu triodes,
2. Absorb all this gain in the Triode1/Triode2/SS OP stage so as to reduce distortion to virtually unmeasurable levels.
3. Add on an SRPP front end (high detail, good drive, excellent overload), OUTSIDE the NFB loop, so as to add the sonic overlay of a triode.
The reason I wanted to use tubes in this buffer is that every attempt to use SS as voltage amplification is flawed sonically because of the very high, poor quality parasitics of nearly every transistor I've ever used in common emitter. This always mandates lag compensation, which is not required with a tube buffer. It is arguable that tubes might be best, in fact, because it should be possible to do this with mosfets/fets very neatly too.
I like your 6pi45 circuit, Runeight, looks very clever. I suspect it could be as easily done with a bipolar or mosfet diff pair, and this would bring voltages down considerably, which could be an advantage as it will eliminate the current sink.
Tim, your circuit is a conventional, cascaded circuit block and I'm well familiar with it. Your addition of an emitter follower buffer at the end is probably not required in this application as the drive to a double emitter follower is high impedance and can be effected capacitively straight from the plate. My doubts revolve around its overload capacity; I'm sure it may not take 80Vpp at the input.
Since the circuit is hybrid, there is no reason not to use the whole gamut of devices; triodes, pentodes, jfets, mosfets, bipolars. It could be quite an exotic mix!
Thank again guys. This is very interesting!
Cheers,
Hugh
Yes, SY, absolutely.
This to eliminate the non-linearities, particularly crossover, but also to compensate the wide variation in Vbe as the transistors swing into heavy conduction.
There will be phase shift here; inevitable. However, with fast transistors, the phase shift in emitter followers is very small, and it's possible, given the use of lower frequency tubes, that lag compensation need not be used.
Cheers,
Hugh
This to eliminate the non-linearities, particularly crossover, but also to compensate the wide variation in Vbe as the transistors swing into heavy conduction.
There will be phase shift here; inevitable. However, with fast transistors, the phase shift in emitter followers is very small, and it's possible, given the use of lower frequency tubes, that lag compensation need not be used.
Cheers,
Hugh
Hi Frank,
Thank you for your input! You wrote:
Not quite. The feedback point must be at the cathode of the first of the two stage buffer for correct phase. At the grid of the 1st tube of the buffer, assume inphase; second grid antiphase, input and output from output stage, inphase again. The feedback is taken from the output, inphase, and must be plumbed into the cathode of the first tube; not any stage BEFORE that tube.
Actually, I want to keep the very first stage, an SRPP, out of the feedback loop entirely. Why? I do not want the tube sonics submerged in high levels of feedback - the sonic kiss of death.....!
Cheers,
Hugh
Thank you for your input! You wrote:
Not quite. The feedback point must be at the cathode of the first of the two stage buffer for correct phase. At the grid of the 1st tube of the buffer, assume inphase; second grid antiphase, input and output from output stage, inphase again. The feedback is taken from the output, inphase, and must be plumbed into the cathode of the first tube; not any stage BEFORE that tube.
Actually, I want to keep the very first stage, an SRPP, out of the feedback loop entirely. Why? I do not want the tube sonics submerged in high levels of feedback - the sonic kiss of death.....!
Cheers,
Hugh
Just for completeness, the circuit I drew out can be done with 6922s through direct substitution for the 6C45PIs. This will save on bottles.
Also, I mistyped. The stage can do over 50vp (not 50vpp). In fact in can take 70vp (140pp) in and output the same.
In a way, this circuit doesn't really solve the problem of no CFs since the first half of the CK pair is a CF and with the direct coupling of output to input on the grounded grid cascoded second stage, we basically have a plate follower.
Oh well . . . .
Also, I mistyped. The stage can do over 50vp (not 50vpp). In fact in can take 70vp (140pp) in and output the same.
In a way, this circuit doesn't really solve the problem of no CFs since the first half of the CK pair is a CF and with the direct coupling of output to input on the grounded grid cascoded second stage, we basically have a plate follower.
Oh well . . . .
Say, Hugh, strangely enough, your answer may have appeared on the Headwize forum. Here's a schematic showing, I think, exactly what you want, except the output stage is tubes.
I've simulated this arrangement using a pair of 5687s with a simple load resistor for the power stage and it does give almost unity gain. But, it does not have vanishingly low distortion.
Web page is in German, but the schematic is easy to read.
Makovski Amp
I've simulated this arrangement using a pair of 5687s with a simple load resistor for the power stage and it does give almost unity gain. But, it does not have vanishingly low distortion.
Web page is in German, but the schematic is easy to read.
Makovski Amp
Hi Runeight,
Thank you sincerely for the Pickie-Makovski schemat.
It is an excellent start; there really is no point re-inventing the wheel as most of it has been done before.
I note two CF are used here in antiphase. Very interesting circuit, shows imagination.
It should be possible to direct couple a pair of 5687s, thus keeping it all in the one bottle. I'm not sure if feedback directly to the first cathode is possible; if so, it would appear a particularly neat trick, something like Futterman's trick OTL!
I think at this point I should start building and testing. This is sufficiently obscure not to be easily modelled in PSpice.
Once again, thanks.....!
Cheers,
Hugh
Thank you sincerely for the Pickie-Makovski schemat.
It is an excellent start; there really is no point re-inventing the wheel as most of it has been done before.
I note two CF are used here in antiphase. Very interesting circuit, shows imagination.
It should be possible to direct couple a pair of 5687s, thus keeping it all in the one bottle. I'm not sure if feedback directly to the first cathode is possible; if so, it would appear a particularly neat trick, something like Futterman's trick OTL!
I think at this point I should start building and testing. This is sufficiently obscure not to be easily modelled in PSpice.
Once again, thanks.....!
Cheers,
Hugh
I think the trick is that cathode is a the output DC zero at idle and the grid is biased slightly negative (as you can easily see). So the cathode simply floats along with the output.
I was able to simulate this very easily with one 5687 (two triodes) by feeding the output of the second plate directly back into the cathode of the first tube through a capacitor. I placed a load resistor on the output attached to a negtive supply so that I could pull the output DC down to floating zero. So, the second tube saw a big load (100K) while driving the first tube's cathode directly. It worked perfectly in terms of gain and it swung 50vp easily.
You right, though, simulations must be used with one eye always on the real world.
If you build this, I'd really like to know what you find.
I was able to simulate this very easily with one 5687 (two triodes) by feeding the output of the second plate directly back into the cathode of the first tube through a capacitor. I placed a load resistor on the output attached to a negtive supply so that I could pull the output DC down to floating zero. So, the second tube saw a big load (100K) while driving the first tube's cathode directly. It worked perfectly in terms of gain and it swung 50vp easily.
You right, though, simulations must be used with one eye always on the real world.
If you build this, I'd really like to know what you find.
Hi Runeight,
I will most certainly let you know very quickly; but I'm also attracted to the Makovskie circuit, which looks like it might not require such a complex power supply.
Thank you again!
The first net Search Engine, huh? Now that was bloody clever! Do you often muse on what you might have spawned??
Cheers,
Hugh
I will most certainly let you know very quickly; but I'm also attracted to the Makovskie circuit, which looks like it might not require such a complex power supply.
Thank you again!
The first net Search Engine, huh? Now that was bloody clever! Do you often muse on what you might have spawned??
Cheers,
Hugh
It is the absolute truth that the group I led at a research consortium in Austin, TX developed the first search engine which was copied by all the others. The only "muse" is that the folks we worked for didn't understand the commercial value of what we had and wouldn't let us take out the intellectual property. It was a very difficult time. I will tell you this though, from the very day we began to track search statistics, porno terms were always the top ten most searched expressions. So many things changing while still remaining the same . . . . . .
However, regarding the problem at hand. The more I look at the circuit (and there is a great explanation by PRR on the headwize thread about 6c33 amps), the more conventional it looks, mostly.
Setting aside the NFB for the moment, the output section is a conventional grounded cathode feeding a split load feeding a pair of CFs buffering the output grids. There are some important things that I missed on first look that PRR graciously pointed out to me. One of those was the feedback loop into the cathode of the first tube.
I think in a normal futterman, the output is fed back into the bottom of the split load where it balances the drive signals into the upper/lower output tubes. There are problems with this technique, but it mostly works.
However, in this Makovski design, the output is 100% fed back into the cathode of the grounded cathode creating the unity gain buffer over the entire output section. Drive signals are not balanced, except for a specific load.
If you replaced the srpp topology at the output with a bipolar SS current buffer, my guess is that this would work smooth as silk. And since the SS stage won't suffer the unbalanced drive signal problem, much complexity will go away.
I will tell you this though, from the very day we began to track search statistics, porno terms were always the top ten most searched expressions. So many things changing while still remaining the same . . . . . .However, regarding the problem at hand. The more I look at the circuit (and there is a great explanation by PRR on the headwize thread about 6c33 amps), the more conventional it looks, mostly.
Setting aside the NFB for the moment, the output section is a conventional grounded cathode feeding a split load feeding a pair of CFs buffering the output grids. There are some important things that I missed on first look that PRR graciously pointed out to me. One of those was the feedback loop into the cathode of the first tube.
I think in a normal futterman, the output is fed back into the bottom of the split load where it balances the drive signals into the upper/lower output tubes. There are problems with this technique, but it mostly works.
However, in this Makovski design, the output is 100% fed back into the cathode of the grounded cathode creating the unity gain buffer over the entire output section. Drive signals are not balanced, except for a specific load.
If you replaced the srpp topology at the output with a bipolar SS current buffer, my guess is that this would work smooth as silk. And since the SS stage won't suffer the unbalanced drive signal problem, much complexity will go away.
Runeight, Tim, SY,
AFter combing the tube datasheets I've come up with a simpler solution.
There are two options. Use a pentode, or a triode, in anode follower mode. Since there are pentodes with excellent harmonic profiles, such as the 12GN7, this ain't so silly.
However, there is an interesting hi mu triode which appears to be well regarded sonically. This is the 6GK5. Plate current of 11.5mA at Vak 150V, mu of 78, plate impedance of 5K7. In anode follower, it could be configured to unity gain by feeding back to the grid. I'll need to build it up and check it will work, but John Broskie's math looks quite doable. This should give an OLG around 55, plenty for government work, and thus a highish feedback factor to smooth the output stage non-linearities.
For unity gain, the input and the feedback resistors would typically be 47K, this then is the input impedance.
It is very simple, and should be effective.
I shall report back...
Happy New Year to all!
Cheers,
Hugh
AFter combing the tube datasheets I've come up with a simpler solution.
There are two options. Use a pentode, or a triode, in anode follower mode. Since there are pentodes with excellent harmonic profiles, such as the 12GN7, this ain't so silly.
However, there is an interesting hi mu triode which appears to be well regarded sonically. This is the 6GK5. Plate current of 11.5mA at Vak 150V, mu of 78, plate impedance of 5K7. In anode follower, it could be configured to unity gain by feeding back to the grid. I'll need to build it up and check it will work, but John Broskie's math looks quite doable. This should give an OLG around 55, plenty for government work, and thus a highish feedback factor to smooth the output stage non-linearities.
For unity gain, the input and the feedback resistors would typically be 47K, this then is the input impedance.
It is very simple, and should be effective.
I shall report back...
Happy New Year to all!
Cheers,
Hugh
I've used the 6GK5 in phono amps. It's pretty quiet, but tends toward microphonics- probably not an issue for you in a power amp. But it will be a challenge to get sufficient swing.
Why not keep it simple and reduce the open loop gain requirement a bit? That will let you use the simple diff amp with feedback from the emitter follower, which has the great virtue of simplicity.
Why not keep it simple and reduce the open loop gain requirement a bit? That will let you use the simple diff amp with feedback from the emitter follower, which has the great virtue of simplicity.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.