I agree with both of these statements for ordinary small signal work. I have been experimenting with cathode follower based output stages for about a year now. I even went so far as to develop a cathode follower with a DSP based SMPS to keep a constant voltage across the output tube. The results were superb. Over 20 watts at .5% distortion, .1 % at 5 watts. I submitted the whole thing to an embedded system design contest where a vacuum tube amplifier actually won a prize against several thousand microprocessor based gizmos. See the winners here, mine is the fourth one:
http://www.circuitcellar.com/microchip2007/
This thread made me aware of Macdonalds pure genius. I have downloaded all of his vacuum tube related patents and IEEE papers. He even described the method required to modulate the power supply to improve efficiency and power output 50 YEARS AGO! I don't think he ever built it. One of his IEEE papers outlines the most complicated vacuum tube amplifier design that I have ever seen. Multiple feedback loops like a Citation 2.
I have just finished a PC board design of an "Augmented Cathode Follower on steroids" and hope to fire it up this weekend. It is an all tube design except for SS rectification and a couple of CCS chips.
Ross MacDonald's web site:
http://www.jrossmacdonald.com/
Someone really needs to get him to register on this forum. Would be like having N. Crowhurst back.
George,
Any link or scan available for that complex multiple feedback amplifier by MacDonald?
Your variable switching power supply follower reminded me of this one:
http://www.diyaudio.com/forums/showthread.php?postid=871030#post871030
Would use a class D or class AB for the SS amp part, the tube amp stiill controls the final output.
There was also a SS design (Technics?) in the market at one time that used a class AB amp to control the common point of the + and - 5 V collector supplies for a class A amp. Gave good efficiency with class A output. Could be adapted for an efficient tube amp.
Don
http://www.jrossmacdonald.com/
Someone really needs to get him to register on this forum. Would be like having N. Crowhurst back.
George,
Any link or scan available for that complex multiple feedback amplifier by MacDonald?
Your variable switching power supply follower reminded me of this one:
http://www.diyaudio.com/forums/showthread.php?postid=871030#post871030
Would use a class D or class AB for the SS amp part, the tube amp stiill controls the final output.
There was also a SS design (Technics?) in the market at one time that used a class AB amp to control the common point of the + and - 5 V collector supplies for a class A amp. Gave good efficiency with class A output. Could be adapted for an efficient tube amp.
Don
George,
Any link or scan available for that complex multiple feedback amplifier by MacDonald?
It is an IEEE publication which are free to members, and costly to non members. Email me, I can send you a copy.
I plan to post the old patents on my web site as soon as I can figure out how to put PDF's on it.
Any link or scan available for that complex multiple feedback amplifier by MacDonald?
It is an IEEE publication which are free to members, and costly to non members. Email me, I can send you a copy.
I plan to post the old patents on my web site as soon as I can figure out how to put PDF's on it.
smoking-amp [/i] [B] The Gm is not constant for any tubes to a first approximation unless the current is held constant: Gm = k*Ip^(.33333) So thats the reason for the CCS in the cathode circuit. [/B][/QUOTE] Don said:
That is a good idea about compensating for the load current.
SY [/i] [B] The reality is said:
Don, That is an interesting point and I know that tubelab uses a source follower in his PowerDrive circuit. Would an N-channel MOSFET source follower benefit from constant current and its source-to-drain voltage being held constant?
Dave
The interelectrode capacitances of any solid state device vary with the voltage across the device. Some people claim that the time varying capacitances can cause phase intermodulation distortion which can be audible. This phenomenon is difficult to measure, but bootstrapping the source follower and CCS loading it should eliminate it. This is on my list of things to try.
Let us know what you find. 
tubelab.com said:
The problem is classical: the amp feeding the amp. There are patents exist about such approach, for example D class feeding A class, B class feeding A class. But while the output device theoretically now is working in a privileged conditions practically the amp that feeds it is not the ideal one so it's impact on a sound quality remains. Also, it has to be driven, i.s. it's non-linear input resistance still will impact.
Hey and interesting read - Some comments on stuff above.
Post #24 Fuling - High gm pentode cathode follower, pentodes definitely make better cathode followers than triodes
BUT
If you wire it as a triode then you loose the benefits, in fact if you fail to bootstrap the screen of the pentode, that is, hold the g2 to cathode voltage constant then it is the triode characteristics which apply (and you loose the benefits).
Post #28 This bootstrap of the anode load of the preceding stage is a common technique. It was used widely in SS amps to bootstrap the VAS load from the Amp Output. It has the benefit of puting the stage into almost constant current operation and hence maximises gain AND reduces distortion
BUT
Read the TUBCAD article about this - you loose some of the cathode follower benefits as a trade off.
POST # 37 Smoking Amp - Yep, the MOSFET Source Follower works really well - to minimise Drain Gate capacitance modulation the trick is to make sure that you maintain at least 25 to 30 volts across the MOSFET even at maximum positive output voltage swing at the Source - that is, its just a matter of gettting rail voltages correct. If you do that you will find its not necessary to bootstrap the DRAIN voltage or anything more sophisticated.
Cheers,
Ian
Post #24 Fuling - High gm pentode cathode follower, pentodes definitely make better cathode followers than triodes
BUT
If you wire it as a triode then you loose the benefits, in fact if you fail to bootstrap the screen of the pentode, that is, hold the g2 to cathode voltage constant then it is the triode characteristics which apply (and you loose the benefits).
Post #28 This bootstrap of the anode load of the preceding stage is a common technique. It was used widely in SS amps to bootstrap the VAS load from the Amp Output. It has the benefit of puting the stage into almost constant current operation and hence maximises gain AND reduces distortion
BUT
Read the TUBCAD article about this - you loose some of the cathode follower benefits as a trade off.
POST # 37 Smoking Amp - Yep, the MOSFET Source Follower works really well - to minimise Drain Gate capacitance modulation the trick is to make sure that you maintain at least 25 to 30 volts across the MOSFET even at maximum positive output voltage swing at the Source - that is, its just a matter of gettting rail voltages correct. If you do that you will find its not necessary to bootstrap the DRAIN voltage or anything more sophisticated.
Cheers,
Ian
I agree with Ian on the Mosfet CF drain voltage at 20V or more above max signal peak being sufficient to remove the need for a drain follower on top of the Mosfet. The Mosfet looks like a very high Zout pentode (so no plate fdbk to worry about) and the variable capacitance issue is largely gone with sufficient drain voltage. (The gate input capacitance is bootstrapped out besides in CF mode.)
I have attached a capacitance chart from the FQP1N50 datasheet (no affiliation, etc... just an excellent part for tube circuitry. I don't think Fairchild will mind me copying just the cap. chart from their copyrighted data sheet, probably sell some parts for them. )
Above 10 volts, the capacitance is pretty much minimal and stable.
As far as a CCS load for the Mosfet CF, the huge (1,040,000 uMHo) Gm makes this largely unnecessary (in low current tube circuitry) unless you are trying to get ppm distortion levels. And even with the tiny gate-source voltage variation to handle the varying current, most of this is the linear component with a small 2nd harmonic component. If you are trying to drive a low Z load like a speaker, then the CCS pulldown or a complementarily modulated current source is important.
One thing that does concern me however about the Mosfet CF is I just noticed that the FQP1N50 is on lifetime purchase status. This usually means it is going to be discontinued (It only came out in year 2000, lasted 8 years). The FQP1N60 is gone already. No idea whether they will replace it with something else, but these low current parts seem to be getting scarcer.
The TO-220 parts in general seem to be dissappearing in the industry, with surface mount or spring mount tabless parts replacing them. I see an FQU1N80 part with 800V rating, 45W, in the tabless I-PAK package that looks interesting, has 150pF input C compared to the 115pF of the FQP1N50. The P-channel FQP1P50 is on lifetime purchase status too.
Don
I have attached a capacitance chart from the FQP1N50 datasheet (no affiliation, etc... just an excellent part for tube circuitry. I don't think Fairchild will mind me copying just the cap. chart from their copyrighted data sheet, probably sell some parts for them. )
Above 10 volts, the capacitance is pretty much minimal and stable.
As far as a CCS load for the Mosfet CF, the huge (1,040,000 uMHo) Gm makes this largely unnecessary (in low current tube circuitry) unless you are trying to get ppm distortion levels. And even with the tiny gate-source voltage variation to handle the varying current, most of this is the linear component with a small 2nd harmonic component. If you are trying to drive a low Z load like a speaker, then the CCS pulldown or a complementarily modulated current source is important.
One thing that does concern me however about the Mosfet CF is I just noticed that the FQP1N50 is on lifetime purchase status. This usually means it is going to be discontinued (It only came out in year 2000, lasted 8 years). The FQP1N60 is gone already. No idea whether they will replace it with something else, but these low current parts seem to be getting scarcer.
The TO-220 parts in general seem to be dissappearing in the industry, with surface mount or spring mount tabless parts replacing them. I see an FQU1N80 part with 800V rating, 45W, in the tabless I-PAK package that looks interesting, has 150pF input C compared to the 115pF of the FQP1N50. The P-channel FQP1P50 is on lifetime purchase status too.
Don
Attachments
Oops, amend that
My previous statement above about using a CCS with a Mosfet CF needs amending.
The transconductance of a Mosfet/Fet varies roughly linearly with drain current. If we were to operate the FQP1N50 at, say 10 mA, we would get "only" about 15,000 uMHo Gm. This means that operating the Mosfet CF at just low currents per se will not buy us anything distortion-wise.
The real criteria is how much the signal causes the operating current to vary from the idle condition. To optimise this, we would use a higher value resistor from a negative supply, maybe -100V, which is just a means of approaching a CCS in effect. So for a Mosfet CF driving an output tube grid we would likely benefit from using a CCS if -100 volts or so is not avilable for the simpler resistor fix. On the other hand, all it does is get rid of a small amount of fairly harmless 2nd harmonic dist.
Another approach would be to try a class A, P-P, Mosfet buffer in place of the CF. Using two Mosfets, in class A P-P, causes the sum of their transconductances to be constant. But this gets more complicated (biasing, or gain matching for a modulated CCS version or White CF variants, etc) than just the CCS fix.
Don
My previous statement above about using a CCS with a Mosfet CF needs amending.
The transconductance of a Mosfet/Fet varies roughly linearly with drain current. If we were to operate the FQP1N50 at, say 10 mA, we would get "only" about 15,000 uMHo Gm. This means that operating the Mosfet CF at just low currents per se will not buy us anything distortion-wise.
The real criteria is how much the signal causes the operating current to vary from the idle condition. To optimise this, we would use a higher value resistor from a negative supply, maybe -100V, which is just a means of approaching a CCS in effect. So for a Mosfet CF driving an output tube grid we would likely benefit from using a CCS if -100 volts or so is not avilable for the simpler resistor fix. On the other hand, all it does is get rid of a small amount of fairly harmless 2nd harmonic dist.
Another approach would be to try a class A, P-P, Mosfet buffer in place of the CF. Using two Mosfets, in class A P-P, causes the sum of their transconductances to be constant. But this gets more complicated (biasing, or gain matching for a modulated CCS version or White CF variants, etc) than just the CCS fix.
Don
Re: Oops, amend that
Even better: degraded emitter follower in parallel with source follower loaded by voltage to current converter that supplies an idle current equal to peak current so current through MOSFET is always the same and input resistance is more linear and less frequency dependent. I tried such approach for an output stage in class A room heater ( constant 12A total from 120V outlet), it sounds better than everything I ever heard, though is an energy hungry.
smoking-amp said:
Even better: degraded emitter follower in parallel with source follower loaded by voltage to current converter that supplies an idle current equal to peak current so current through MOSFET is always the same and input resistance is more linear and less frequency dependent. I tried such approach for an output stage in class A room heater ( constant 12A total from 120V outlet), it sounds better than everything I ever heard, though is an energy hungry.
"Even better:
output stage in class A room heater ( constant 12A total from 120V outlet)"
Er, Umm...
Do I get any carbon credits for NOT building one of these?
Then again, we are getting some not too common snow outside today. Maybe I SHOULD throw one together. Anyone try using coal as a semiconductor? Lots of magic smoke inside!
output stage in class A room heater ( constant 12A total from 120V outlet)"
Er, Umm...
Do I get any carbon credits for NOT building one of these?
Then again, we are getting some not too common snow outside today. Maybe I SHOULD throw one together. Anyone try using coal as a semiconductor? Lots of magic smoke inside!
Attachments
"Was it your breakfast? "
Nah.., I found it while kayaking the Catawba river here near a R-R bridge on the banks, probably from the steam-engine railroad days. Measures over 2 megohms across 1/16 inch. Obviously needs some N or P type doping. I need some naturally occuring N and P type coal with lower resistance.
Probably easier to make a coal FET, just need N type coal and then plate some copper on it for a gate.
Lots of Mica around here for making tubes:
Don
Nah.., I found it while kayaking the Catawba river here near a R-R bridge on the banks, probably from the steam-engine railroad days. Measures over 2 megohms across 1/16 inch. Obviously needs some N or P type doping. I need some naturally occuring N and P type coal with lower resistance.
Probably easier to make a coal FET, just need N type coal and then plate some copper on it for a gate.
Lots of Mica around here for making tubes:
Don
Attachments
"What are you thinking of "
Well, I wasn't thinking of anything in particular, since there are quite a few buffer options. For a complementary source follower, I might use FQP1N50 and FQP1P50 (if you can still buy them next week) or there are some Zetex parts too, but very watt limited.
Getting similar gm and cap. specs between N and P channel parts is a bit challenging. FQP2N50 and FQP2P25 are a better match if 250V is sufficient. In class A mode, the two transconductances sum to a near constant, which should give low distortion. (NPN and PNP don't sum to constant gm, maybe what Anatoliy is concerned about) Bunch of extra parts required to do the gate biasing though.
Beside the source followers there are all the SRPP, Mu follower, White CF, unity gain or augmented buffers (usually LTP and CF), and the compensated current source load can be incorporated into several as well to keep device current constant. One can also put the dreaded N-FDBK around them too, if they don't already use it.
Don
Well, I wasn't thinking of anything in particular, since there are quite a few buffer options. For a complementary source follower, I might use FQP1N50 and FQP1P50 (if you can still buy them next week) or there are some Zetex parts too, but very watt limited.
Getting similar gm and cap. specs between N and P channel parts is a bit challenging. FQP2N50 and FQP2P25 are a better match if 250V is sufficient. In class A mode, the two transconductances sum to a near constant, which should give low distortion. (NPN and PNP don't sum to constant gm, maybe what Anatoliy is concerned about) Bunch of extra parts required to do the gate biasing though.
Beside the source followers there are all the SRPP, Mu follower, White CF, unity gain or augmented buffers (usually LTP and CF), and the compensated current source load can be incorporated into several as well to keep device current constant. One can also put the dreaded N-FDBK around them too, if they don't already use it.
Don
Hybrid component circuits
Another design variation is to use both tubes and Mosfets in the same "circuit". This might take the form of a Complex Feedback Pair (CFP or Sziklai pair), Baxandall pair, Darlington, or even triples.
These look like three terminal devices and can be dropped into any standard circuit, like the CF for example. The Mosfet embeded in them makes the tube appear to have a huge Gm, yet still operates like a tube voltage amplifier.
(There is a whole subject area on composite amplifiers that mix a voltage and a current amplifier, or two V or I amps together (series or parallel). One can make one of these a tube and the other SS, usually the tube as the V amp and SS as the I amp.)
Another unusual scheme is putting a thermionic diode in series with the base of a bipolar transistor. This gives it tube like transconductance, useful for P-P class AB complementary totem pole ( smooth crossover) too.
I use a pentode circuit with a Mosfet follower between the plate and screen grid (zener or neon voltage dropper in series for the screen) with a CCS pull-up on the plate. This gives a gain of Mu (screen to g1) exactly, with large current buffering capability. Think of pentodes as a triode with separate output and feedback connections.
Another, variation is the use of unbalanced balun coils between tube and SS devices to enforce a current sharing ratio (by turns ratio) between them. I call them Un-Transformer circuits. I usually use a ferrite core xfmr since they only have to handle the Mosfet turn on voltage and operate with near zero flux, due to the current balance nulling the core flux.
The hybrid domain has so many possibilities. It is surprising that the usual published "hybrid amps" seem to just splice tubes and SS together at the output.
Don
Another design variation is to use both tubes and Mosfets in the same "circuit". This might take the form of a Complex Feedback Pair (CFP or Sziklai pair), Baxandall pair, Darlington, or even triples.
These look like three terminal devices and can be dropped into any standard circuit, like the CF for example. The Mosfet embeded in them makes the tube appear to have a huge Gm, yet still operates like a tube voltage amplifier.
(There is a whole subject area on composite amplifiers that mix a voltage and a current amplifier, or two V or I amps together (series or parallel). One can make one of these a tube and the other SS, usually the tube as the V amp and SS as the I amp.)
Another unusual scheme is putting a thermionic diode in series with the base of a bipolar transistor. This gives it tube like transconductance, useful for P-P class AB complementary totem pole ( smooth crossover) too.
I use a pentode circuit with a Mosfet follower between the plate and screen grid (zener or neon voltage dropper in series for the screen) with a CCS pull-up on the plate. This gives a gain of Mu (screen to g1) exactly, with large current buffering capability. Think of pentodes as a triode with separate output and feedback connections.
Another, variation is the use of unbalanced balun coils between tube and SS devices to enforce a current sharing ratio (by turns ratio) between them. I call them Un-Transformer circuits. I usually use a ferrite core xfmr since they only have to handle the Mosfet turn on voltage and operate with near zero flux, due to the current balance nulling the core flux.
The hybrid domain has so many possibilities. It is surprising that the usual published "hybrid amps" seem to just splice tubes and SS together at the output.
Don
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.