"but what is the advantage?"
The idea of anti-triode is for the output to remain true to the SE sound signature of the bottom device, but with the advantage of P-P power and efficiency. The top device ideally provides exactly the complementary current to the bottom device. The subtle distinction from being just an ordinary P-P output, is that the top device ideally mimics the accurate complemtary current of the bottom device, rather than just being driven in a complementary fashion on it's gate (which would then produce the odd harmonics of typical P-P, instead of the even harmonics of the bottom device acting alone).
Don
The idea of anti-triode is for the output to remain true to the SE sound signature of the bottom device, but with the advantage of P-P power and efficiency. The top device ideally provides exactly the complementary current to the bottom device. The subtle distinction from being just an ordinary P-P output, is that the top device ideally mimics the accurate complemtary current of the bottom device, rather than just being driven in a complementary fashion on it's gate (which would then produce the odd harmonics of typical P-P, instead of the even harmonics of the bottom device acting alone).
Don
Got it, in a nutshell, Class-A distortion signature and sound but with better efficiency and drive ability. I know the mu follower, have seen scraps on SEPP, and heard "anti-triode" thrown around before, but didn't know much on it. Thank you very much smoking-amp and kenpeter, cleared up a lot!
Cheers...I learned something new today!
James
Cheers...I learned something new today!
James
Well, it does operate in class A, and with better efficiency than just a CCS on top. Effectively its P-P class A as for power considerations. But the main attraction is to preserve SE sound (distortion signature) of the bottom device (presumably a triode) rather than ending up with the P-P sound from a similar SRPP arrangement.
To really accurately operate as SE emulation would require an Op. Amp controlling the top device, so its a compromise for simplicity as implemented here with mimimal parts.
To really accurately operate as SE emulation would require an Op. Amp controlling the top device, so its a compromise for simplicity as implemented here with mimimal parts.
So if you are not using a tube or depletion device on top can it be done somewhat simply with enhancement mode devices? Also is there a way to calculate the resistor spit, or just trial and error? Feel free too at anytime to just tell me to search 
, I can do that at the same and try to answer some things on my own.
This is still kind of on topic but I don't know if I am kind of thread jacking
, I can do that at the same and try to answer some things on my own.This is still kind of on topic but I don't know if I am kind of thread jacking
"So if you are not using a tube or depletion device on top can it be done somewhat simply with enhancement mode devices? Also is there a way to calculate the resistor spit, or just trial and error?"
Well, the diagram linked in post 19 above shows it with an enhancement Mosfet. Which required a bunch of LEDs to bias the gate voltage up. A depletion mode device could eliminate this DC offset issue.
Nominally, or ideally, the output (current sense) resistor(s) (bottom dev. plate to top device source) gets center tapped for output if the top device has much higher gm than the bottom device. So a Mosfet comes close to the 50/50 tapping for most tubes on the bottom. But keep in mind that the gm of a Mosfet varies approximately linearly with current, and they are spec'd at ampere currents typically. So in the tube circuit its gm is not so huge, but still usually 10x ballpark more than the tube.
To get 1st order improvement, assume some delta I current variation from the tube plate. Then calculate how much gate voltage variation is required on the Mosfet (at its reduced spec gm) to reduce it's current by the same amount. Now just add enough resistance to the bottom section of the split resistor so that it generates that additional voltage with the initial delta I variation.
The total resistance of the sense resistors and the Mosfet or depl. mode voltage offset set the idle current for the stage by setting the operating bias of the FET with some voltage across the sense resistors.
In general, the scheme relies on the top device having considerably more gm than the bottom device, in order to get close to SE signature. So a tube on top is not likely to cut it without an additional diffl. stage acting as an Op Amp up there.
Don
Well, the diagram linked in post 19 above shows it with an enhancement Mosfet. Which required a bunch of LEDs to bias the gate voltage up. A depletion mode device could eliminate this DC offset issue.
Nominally, or ideally, the output (current sense) resistor(s) (bottom dev. plate to top device source) gets center tapped for output if the top device has much higher gm than the bottom device. So a Mosfet comes close to the 50/50 tapping for most tubes on the bottom. But keep in mind that the gm of a Mosfet varies approximately linearly with current, and they are spec'd at ampere currents typically. So in the tube circuit its gm is not so huge, but still usually 10x ballpark more than the tube.
To get 1st order improvement, assume some delta I current variation from the tube plate. Then calculate how much gate voltage variation is required on the Mosfet (at its reduced spec gm) to reduce it's current by the same amount. Now just add enough resistance to the bottom section of the split resistor so that it generates that additional voltage with the initial delta I variation.
The total resistance of the sense resistors and the Mosfet or depl. mode voltage offset set the idle current for the stage by setting the operating bias of the FET with some voltage across the sense resistors.
In general, the scheme relies on the top device having considerably more gm than the bottom device, in order to get close to SE signature. So a tube on top is not likely to cut it without an additional diffl. stage acting as an Op Amp up there.
Don
Well, obviously there is considerable similarity between Mu follower, SRPP and Anti-triode. SRPP looks like a CCS loaded triode driving a cathode follower. Mu follower looks like a CCS loaded triode alone. Both of them typically operate with similar (ballpark) gm's for top and bottom devices.
Anti-triode wants a very high gm device on top, and tries to produce true complementary currents from the top and bottom devices. Ie, I(top) +I(bott) = constant. This relation does not hold well for Mu follower or SRPP.
So the old arguments about whether to take the output off the top device or the bottom device are settled here, the middle is best. (But SE signature requires the top device to have much higher gm)
Don
Anti-triode wants a very high gm device on top, and tries to produce true complementary currents from the top and bottom devices. Ie, I(top) +I(bott) = constant. This relation does not hold well for Mu follower or SRPP.
So the old arguments about whether to take the output off the top device or the bottom device are settled here, the middle is best. (But SE signature requires the top device to have much higher gm)
Don
The anti-triode uses the two resistors tapped to the output node to sense the currents in the top and bottom devices. A constant voltage is ideally maintained across the two resistors (for infinite gm on top) when I(bot) + I(top) = constant. The voltage drop across the top FET does not need to be half the B+. They can generally operate with less than that. But too low a voltage across the FET will get one into non-linear capacitance problems from the FET. Best to keep the FET with at least 50V across it mimimum (ie. at output signal positive peaks).
Don
Don
Well, the picture is tricky in respect to feedback here. 1st order operation is feedback free actually, since the top device just operates from the gate voltage developed by the current sense resistors. So one could call it feedback free I suppose in the usual sense. But in reality, the load generates voltage drops on the feedback resistors, so the device(s) operation is in effect monitored by the sense resistors as well.
Real physical reality is that NOTHING is feedback free, even a simple resistor at the end of a transmission line causes feedback. (Yes, I'm aware of some arguments over resistors not having feedback, they're flat wrong: The applied voltage wave propagates down the trans. line to the resistor, and a reflected wave, determined by the resistor impedance matching to the line Zo, propagates back from the resistor to the source, which then determines the current drawn from the source. If you put a sampling scope on the line, you will see the current initially rise to the line Zo determined value, then re-adjust to the resistor determined value after a nS or so for the signal to propagate down to the resistor and back. It's real. ). It all depends on whether you analyse the situation in nS time frames or mS time frames. In quantum mechanics, the waves even propagate into the future to make sure the present is compatible with the future as well as the past. Talk about feedback overload.
Don
Real physical reality is that NOTHING is feedback free, even a simple resistor at the end of a transmission line causes feedback. (Yes, I'm aware of some arguments over resistors not having feedback, they're flat wrong: The applied voltage wave propagates down the trans. line to the resistor, and a reflected wave, determined by the resistor impedance matching to the line Zo, propagates back from the resistor to the source, which then determines the current drawn from the source. If you put a sampling scope on the line, you will see the current initially rise to the line Zo determined value, then re-adjust to the resistor determined value after a nS or so for the signal to propagate down to the resistor and back. It's real. ). It all depends on whether you analyse the situation in nS time frames or mS time frames. In quantum mechanics, the waves even propagate into the future to make sure the present is compatible with the future as well as the past. Talk about feedback overload.
Don
OK, first I would like to say duh James, read smoking-amp's post AND click on the links...sorry about that
Second I have been slowly reading through the link you posted and I have to agree with G on page two, I have a headache! However some of it is sinking in. I am reading why not, but CCT1 still looks like a differential to me, I will get there (>>OK never mind, I have read on as I post<<). CCT2 obviously is what I am interested in since this is what we are talking about. However CCT3 really perks my interest, especially with the FET's canceling.
It was mentioned in that thread though just kind of touched on, how does the Murray do away with the resistors? I am assuming it is something to do with the circuit to the left of the output stage and that all will be reveled once I finish reading through the articles. Even if it is just a couple of resistors I do like the idea of getting a few more parts out of the way. What really has me curious though is if a blend of the two could be designed. The "anti-triode" CCT2 of yours does away with the DC blocking cap but has the resistors (big deal I guess), but the Murray loses them but regains the blocking cap.
I guess I am confused why the Murray is SEPP when it doesn't seem SE with DC compensation. Instead, to me at least, it looks like a SE amplifier with an elaborate yet normal CCS stuck on top. However this doesn't seem like the case since some are getting 30watts out of a triode strapped EL509, this doesn't seem like SE territory...something is going on here. Like I said though, the cap, it is stated as being 100uF! I would really like this amp if that could be done away with, 30watts of "SE" while not using GM70's or the like, that sounds like a plan to me.
OK, now the wheels are spinning, something is telling me to dig up some 6DN7's and shot for 300B power, wide bandwidth, and spend about $50!
Cheers
James
for those just entering
http://www.diyaudio.com/forums/showthread.php?postid=638005#post638005
for the Murray
http://www.wimdehaan.nl/tubeamps/murray/
edit: still wondering, should this be split?? Of course I guess the original thread is like 4 years old so...
Second I have been slowly reading through the link you posted and I have to agree with G on page two, I have a headache! However some of it is sinking in. I am reading why not, but CCT1 still looks like a differential to me, I will get there (>>OK never mind, I have read on as I post<<). CCT2 obviously is what I am interested in since this is what we are talking about. However CCT3 really perks my interest, especially with the FET's canceling.
It was mentioned in that thread though just kind of touched on, how does the Murray do away with the resistors? I am assuming it is something to do with the circuit to the left of the output stage and that all will be reveled once I finish reading through the articles. Even if it is just a couple of resistors I do like the idea of getting a few more parts out of the way. What really has me curious though is if a blend of the two could be designed. The "anti-triode" CCT2 of yours does away with the DC blocking cap but has the resistors (big deal I guess), but the Murray loses them but regains the blocking cap.
I guess I am confused why the Murray is SEPP when it doesn't seem SE with DC compensation. Instead, to me at least, it looks like a SE amplifier with an elaborate yet normal CCS stuck on top. However this doesn't seem like the case since some are getting 30watts out of a triode strapped EL509, this doesn't seem like SE territory...something is going on here. Like I said though, the cap, it is stated as being 100uF! I would really like this amp if that could be done away with, 30watts of "SE" while not using GM70's or the like, that sounds like a plan to me.
OK, now the wheels are spinning, something is telling me to dig up some 6DN7's and shot for 300B power, wide bandwidth, and spend about $50!
Cheers
James
for those just entering
http://www.diyaudio.com/forums/showthread.php?postid=638005#post638005
for the Murray
http://www.wimdehaan.nl/tubeamps/murray/
edit: still wondering, should this be split?? Of course I guess the original thread is like 4 years old so...
We are probably straying seriously from Gavin's original BOZ preamp thread, unless he doesn't mind. This "anti-triode" stuff seems to regularly ignite OT discussions whenever it comes up. Takes a while for most to understand the subtle ideas in it. And it can easily start firestorms over SE versus P-P. Sorta "anti-Christ" to SET true believers. But interesting anyway.
Ohh.. I should also mention that this idea has been invented a few times in similar forms, not just my idea. I think Ken and Michael K and Anatolyi (Wavebourn) have all discussed related ideas, differing in details. Not to mention Nelson P too. Sometimes an idea just has its time arrive. Some amps have been built using it too, I think they're linked at the end of the thread mentioned.
Have to look at the Murray stuff, I can't remember it at the moment.
Don
Ohh.. I should also mention that this idea has been invented a few times in similar forms, not just my idea. I think Ken and Michael K and Anatolyi (Wavebourn) have all discussed related ideas, differing in details. Not to mention Nelson P too. Sometimes an idea just has its time arrive. Some amps have been built using it too, I think they're linked at the end of the thread mentioned.
Have to look at the Murray stuff, I can't remember it at the moment.
Don
Yeah, I gotta hope that the SET fundamentalists don't come over and burn my "anti-triodes" at the stake.
On the Murray amp, it appears to compare the final output voltage with the drive voltage going to the bottom PL36. Then applies the amplified (by EF86) error to the top PL36. So it avoids the current sensing resistors (of "anti-triode") by using voltage feedback instead. I don't think it really tries to use complementary currents in the top and bottom PL36's directly at all. Rather, the top device is used as an error corrector for the bottom device. Can think of the EF86 as an inverter of the drive signal, plus injecting negated error detected between the drive sig. and the output.
This might still, in effect, come sorta close to output current complementarity in the final analysis with a load, would need to simulate it really to find out. But I think error residuals in the bottom device really are running the show here, not complementarity by design. So I would categorize the Murray amp as more of a conventional feedback design, but with an un-conventional assymetric output stage. No doubt there will be some even harmonic dist. remaining due to the assymetry, so it will likely be reminiscent of SE (but lower distortion than a non-feedback SET).
The "anti-triode" scheme does NOT try to fix any errors in the bottom triode's operation at all. Rather, it treats the triode's operation as gospel, just replicating its power by 2X. So if you put a bad triode in the bottom, you will get 2X bad results. (Of course, there could be feedbacks around the "anti-triode" or whole ampl. in some designs, but that's not really a feature of "anti-triode" per se.)
On the DC blocking caps, both schemes have the same issues here. Either have to block DC with the cap, or use + and - supplies and a DC servo circuit to null out DC from the output.
The figure 1 "anti-triode" schemes avoids the caps by using a CT'd output xfmr. There IS still an issue with DC balance in the fig. 1 scheme however, as pointed out by Michael K. Since the triode has some small non-linearity, the signal excursions + and - are not quite balanced, leading to average DC imbalance at large signal levels. This may be ignorable if the xfmr can handle a little DC unbalance (could be OK for E-I, but a problem for toroids), otherwise a DC servo or equiv. is needed. There were some discussions and suggested solutions on this in a later thread linked from the original thread.
Don
On the Murray amp, it appears to compare the final output voltage with the drive voltage going to the bottom PL36. Then applies the amplified (by EF86) error to the top PL36. So it avoids the current sensing resistors (of "anti-triode") by using voltage feedback instead. I don't think it really tries to use complementary currents in the top and bottom PL36's directly at all. Rather, the top device is used as an error corrector for the bottom device. Can think of the EF86 as an inverter of the drive signal, plus injecting negated error detected between the drive sig. and the output.
This might still, in effect, come sorta close to output current complementarity in the final analysis with a load, would need to simulate it really to find out. But I think error residuals in the bottom device really are running the show here, not complementarity by design. So I would categorize the Murray amp as more of a conventional feedback design, but with an un-conventional assymetric output stage. No doubt there will be some even harmonic dist. remaining due to the assymetry, so it will likely be reminiscent of SE (but lower distortion than a non-feedback SET).
The "anti-triode" scheme does NOT try to fix any errors in the bottom triode's operation at all. Rather, it treats the triode's operation as gospel, just replicating its power by 2X. So if you put a bad triode in the bottom, you will get 2X bad results. (Of course, there could be feedbacks around the "anti-triode" or whole ampl. in some designs, but that's not really a feature of "anti-triode" per se.)
On the DC blocking caps, both schemes have the same issues here. Either have to block DC with the cap, or use + and - supplies and a DC servo circuit to null out DC from the output.
The figure 1 "anti-triode" schemes avoids the caps by using a CT'd output xfmr. There IS still an issue with DC balance in the fig. 1 scheme however, as pointed out by Michael K. Since the triode has some small non-linearity, the signal excursions + and - are not quite balanced, leading to average DC imbalance at large signal levels. This may be ignorable if the xfmr can handle a little DC unbalance (could be OK for E-I, but a problem for toroids), otherwise a DC servo or equiv. is needed. There were some discussions and suggested solutions on this in a later thread linked from the original thread.
Don
Active current source on top assumes the device below has
either an intrinsic "Mu" or External Schading network. There
has to be some sort of voltage control. Else the circuit tends
toward unlimited gain, and becomes a latch.
You can't stick Aleph or Anti-Triode blindly on top the MOSFET
of post #1, as it is a current amplifier with no voltage control.
No control aside from the 10K load. You can't substitute any
CSS (Aleph or not) here for the 10K, unless you first implement
some other means of controlling the resulting voltage swing.
By Schading (since someone requested I speak in English)
I mean using an Anode follower, or other partial feedback
to transform Pentode (or MOSFET) curves into a Triode.
The original idea, as proposed by O.H. Schade long ago.
Obviously, before there were MOSFETs. Or Pass. Or Zen.
Or Schade may just have been the first to document it...
I wasn't there, I don't know...
either an intrinsic "Mu" or External Schading network. There
has to be some sort of voltage control. Else the circuit tends
toward unlimited gain, and becomes a latch.
You can't stick Aleph or Anti-Triode blindly on top the MOSFET
of post #1, as it is a current amplifier with no voltage control.
No control aside from the 10K load. You can't substitute any
CSS (Aleph or not) here for the 10K, unless you first implement
some other means of controlling the resulting voltage swing.
By Schading (since someone requested I speak in English)
I mean using an Anode follower, or other partial feedback
to transform Pentode (or MOSFET) curves into a Triode.
The original idea, as proposed by O.H. Schade long ago.
Obviously, before there were MOSFETs. Or Pass. Or Zen.
Or Schade may just have been the first to document it...
I wasn't there, I don't know...
Attachments
And in response to Smoking's assertion that Gm of the Anti-
device must always be higher. Yes and No.... Depending.
If reference device gain is cut down by a feedback loop,
such as the case with anode followers. You could still
probably chisel a decent Anti out of an identical block
of marble (or sand or whatever...) Since both device
and anti-device are smaller than the starting material.
Its only where you run the main device balls to the wall,
at full Mu or Gm. The virtual Anti-device best be carved
out of something bigger. That doesn't have the wrong
corners already rounded off.
Paraphase is a good example where this usually fails:
Trying to carve a mirror opposite Anti-Triode out of a
Triode of the same starting size... You can take away
material, but you can't put a nose back on the statue's
butt once you've broken it off to carve a rear end on its
face. You gotta start with something big enough to cut
all surfaces down to the proper anti-shape.
And any unexpected behavior of the anti-device is seen
by the reference as-if it were something odd in the load,
creating an even greater confusion. Not that I ever do.
device must always be higher. Yes and No.... Depending.
If reference device gain is cut down by a feedback loop,
such as the case with anode followers. You could still
probably chisel a decent Anti out of an identical block
of marble (or sand or whatever...) Since both device
and anti-device are smaller than the starting material.
Its only where you run the main device balls to the wall,
at full Mu or Gm. The virtual Anti-device best be carved
out of something bigger. That doesn't have the wrong
corners already rounded off.
Paraphase is a good example where this usually fails:
Trying to carve a mirror opposite Anti-Triode out of a
Triode of the same starting size... You can take away
material, but you can't put a nose back on the statue's
butt once you've broken it off to carve a rear end on its
face. You gotta start with something big enough to cut
all surfaces down to the proper anti-shape.
And any unexpected behavior of the anti-device is seen
by the reference as-if it were something odd in the load,
creating an even greater confusion. Not that I ever do.
"You can take away
material, but you can't put a nose back on the statue's
butt once you've broken it off to carve a rear end on its
face."
Clearly spoken from experience...
What I think you're getting at may reduce down to this:
In the anti-triode current mirror circuits there must be
enough reserve idle current to prevent clipping of the
high-gm/gfs device at either extreme of signal. In other
words, the idle current must be substantially more than
the peak output current swing, either direction. If not,
the "soft limiting into 2nd harmonic" characteristic of
SET is lost.
In practical terms, this means (to my disappointment)
you can't get 2X efficiency while using load lines that
deliver high anode efficiency, i.e. where peak signal
current is > 2X idle current. It turns out to be another
application of the free lunch principle.
You can get 2X the power but you need to raise idle
current a bit. To offset this, you don't need 2X the B+
because you can use all but 20 volts across the FET
for signal.
I'm currently exploring class A2 and the totem-pole
variants. With A2 you use more of the tube voltage for
signal also, which allows approaching a 50% output
power to idle power ratio.
Cheers,
Michael
material, but you can't put a nose back on the statue's
butt once you've broken it off to carve a rear end on its
face."
Clearly spoken from experience...
What I think you're getting at may reduce down to this:
In the anti-triode current mirror circuits there must be
enough reserve idle current to prevent clipping of the
high-gm/gfs device at either extreme of signal. In other
words, the idle current must be substantially more than
the peak output current swing, either direction. If not,
the "soft limiting into 2nd harmonic" characteristic of
SET is lost.
In practical terms, this means (to my disappointment)
you can't get 2X efficiency while using load lines that
deliver high anode efficiency, i.e. where peak signal
current is > 2X idle current. It turns out to be another
application of the free lunch principle.
You can get 2X the power but you need to raise idle
current a bit. To offset this, you don't need 2X the B+
because you can use all but 20 volts across the FET
for signal.
I'm currently exploring class A2 and the totem-pole
variants. With A2 you use more of the tube voltage for
signal also, which allows approaching a 50% output
power to idle power ratio.
Cheers,
Michael
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.