A little background:
If you remember, in the Counterpoint SA100 thread I mentioned I would look for a triode-pentode in one envelope (ECL82 or 86, or PCL or UCL equivalents) for a driver stage of a hybrid amp, quite similar to the one proposed by gianmaria, though, different in several key areas.
The idea was to use a common cathode amp feeding a follower, in order to be able to bootstrap the output of the follower onto the triode Rk in order to form a 'free' CCS.
The pentode part was originally intended to be triode connected, and the reason behind using such a tube is that the pentode can source substantial current, perfect for driving a solid state output stage, as well as simultaneously being a bootstrap source for the preceeding triode.
Since the triode connected pentode drives what is effectively a current amplifier with voltage gain of 1, the output of this amp can again be used to bootstrap the follower cathode, resulting in a CCS termination for the cathode. This last part, however, has not progressed further than a simulation, and it shows very good results.
Intuitively, though, using the pentode as a triode has one interesting advantage, and that is partial cancelation of the triode nonlinearity of the previous stage. Because the follower is not a perfect follower, but effectively a triode follower, the bootstrap does not result in a perfect CCS, but rather with one that has a nonlinear source impedance, in particular nonlinear in an oposite way to the nonlinearity of the triode stage it is terminating. It offers an interesting possibility.
On the other hand, using the pentode as a pentode, results in the bootstrap CCS looking more like a CCS which linearises the triode anyway, but also requires lower anode voltage for the same output swing - simply a practical consideration. I have not yet tried this, even in a simulator.
Since I am using a double bootstrap, DC coupling between tube and SS stage was seriously considered, hence my ideas for gianmaria's thread - hopefully, he can use some of what I wrote above for his own design.
If you remember, in the Counterpoint SA100 thread I mentioned I would look for a triode-pentode in one envelope (ECL82 or 86, or PCL or UCL equivalents) for a driver stage of a hybrid amp, quite similar to the one proposed by gianmaria, though, different in several key areas.
The idea was to use a common cathode amp feeding a follower, in order to be able to bootstrap the output of the follower onto the triode Rk in order to form a 'free' CCS.
The pentode part was originally intended to be triode connected, and the reason behind using such a tube is that the pentode can source substantial current, perfect for driving a solid state output stage, as well as simultaneously being a bootstrap source for the preceeding triode.
Since the triode connected pentode drives what is effectively a current amplifier with voltage gain of 1, the output of this amp can again be used to bootstrap the follower cathode, resulting in a CCS termination for the cathode. This last part, however, has not progressed further than a simulation, and it shows very good results.
Intuitively, though, using the pentode as a triode has one interesting advantage, and that is partial cancelation of the triode nonlinearity of the previous stage. Because the follower is not a perfect follower, but effectively a triode follower, the bootstrap does not result in a perfect CCS, but rather with one that has a nonlinear source impedance, in particular nonlinear in an oposite way to the nonlinearity of the triode stage it is terminating. It offers an interesting possibility.
On the other hand, using the pentode as a pentode, results in the bootstrap CCS looking more like a CCS which linearises the triode anyway, but also requires lower anode voltage for the same output swing - simply a practical consideration. I have not yet tried this, even in a simulator.
Since I am using a double bootstrap, DC coupling between tube and SS stage was seriously considered, hence my ideas for gianmaria's thread - hopefully, he can use some of what I wrote above for his own design.
Hi ilimzn,
Should post in that thread I guess. But now I see where you are going. Work on the input impedance of your SS output stage. Think about making it a consistant, high impedance. Then use a standard tube type coupling with a regular follower. Then you can concentrate on linearizing the tube voltage amplifier.
I found most of my distortion comes from the tube stage. Not nice sounding either. It is hard to deliver +- 50V or so with low distortion from a two stage tube front end, plus follower. This was measured at various points in the amp and with nothing being driven by the tube stage (no loading effects). Mind you, I was working with a modified version of the original circuit from Counterpoint. It's a pain to pull the board up. Looks like I'll completely redesign that as well.
-Chris
Should post in that thread I guess. But now I see where you are going. Work on the input impedance of your SS output stage. Think about making it a consistant, high impedance. Then use a standard tube type coupling with a regular follower. Then you can concentrate on linearizing the tube voltage amplifier.
I found most of my distortion comes from the tube stage. Not nice sounding either. It is hard to deliver +- 50V or so with low distortion from a two stage tube front end, plus follower. This was measured at various points in the amp and with nothing being driven by the tube stage (no loading effects). Mind you, I was working with a modified version of the original circuit from Counterpoint. It's a pain to pull the board up. Looks like I'll completely redesign that as well.
-Chris
anatech said:
Actually, it may be more relevant to this thread...
The output stage in mine is MOSFET, single pair. Gianmaria has decided for BJT tripple darlington, the good thing is he uses the firt two as class A by terminating them into oposing power supplies, this will present a much more consistent impedance to the previous stage compared to a regular EF darlington. A CFP pair may be an interesting alternative here (and easyer to thermally compensate) because it does not present a narrow 'notch' in input impedance around crossover, unlike an EF stage.
I would not consider driving a SS output stage (even MOSFET) without a follower, else any nonlinearity of the SS stage input impedance would present a nonlinear loading of the previous stage, which would compond the nonlinearity of a triode there.
This was one argument for a pentode follower. The high mu of a pentode means (in theory) less nonlinearity in follower mode.
As for large voltage swings, triodes do a MUCH better job with this if Ra is replaced by a CCS. Also, my design, like gianmaria's is targetted at ~~50W output power (with some headroom) which requires less than 60Vpp from the driver. This is possible from a single triode, assuming CCS on top, some 300V B+, and of course, no load
, but that's why there's a follower The input triode needs to be high-mu for reasonable sensitivity. Since these use low plate currents (~~1mA), a bootstrap CCS seemed like a good idea - FAR less complexity than a good high impedance SS CCS, since resistances are large, the bootstrap cap can have low enough capacitance to be good quality foil without braking the bank, and since there is a follower driving it, which normally drives an impedance some 20x lower, it does not present a problem to the follower either.
With a combined triode-pentode tube, the only remaining problem is that the heater needs to be some 75V up from ground to prevent heater to cathode insulation breakdown. With separate tubes the issue is a bit more relaxed but still present, or two heater supplies need to be used.
As for AC vs DC coupling of tube input and SS output, the only really serious problem I've found with the later is the fact that the input becomes referenced to the -V rail of the tube front end (DC coupling requires +-150V supplies for the front end plus DC servo and relais to clamp the SS input while tubes are cold (detected via output of DC servo). This requires a very well regulated, very low noise SS supply for the front end, so what you get is a relatively simple hybrid amp with quite complicated ancillaries...
Hi ilimzn,
Are you using lateral fets?
We both agree on the use of a follower, and a good regulator for the tube B+. We also agree on the current source for the amplification stages. I don't know how much difference there would be between a pentode or triode follower. Never tried a pentode.
You should have, in an ideal world, a separate heater supply for each follower since music is not always in phase. Heaven forbid you bridge the amp, it'll be 180° out by definition. I have seen many examples of the 6922 used with the heater at common potential. It is designed for this service. Pentodes may be a problem here as you don't normally see them in cascode. The heaters aren't designed for that abuse.
We will have to see how these turn out. I gave up on fets for a few reasons. I found the bipolars sounded much better, measured much better too. Easier to bias and more stable. I hope you have more luck than I did in the fet department.
I use around 280VDC for the front end, although I could bump it up to 300VDC. The regulator design is very important. The current sources I use are very simple. Red LED plus PNP transistor and a few other parts. Reliable and I don't run into capacitance problems. The current set is stable.
-Chris
Are you using lateral fets?
We both agree on the use of a follower, and a good regulator for the tube B+. We also agree on the current source for the amplification stages. I don't know how much difference there would be between a pentode or triode follower. Never tried a pentode.
You should have, in an ideal world, a separate heater supply for each follower since music is not always in phase. Heaven forbid you bridge the amp, it'll be 180° out by definition. I have seen many examples of the 6922 used with the heater at common potential. It is designed for this service. Pentodes may be a problem here as you don't normally see them in cascode. The heaters aren't designed for that abuse.
We will have to see how these turn out. I gave up on fets for a few reasons. I found the bipolars sounded much better, measured much better too. Easier to bias and more stable. I hope you have more luck than I did in the fet department.
I use around 280VDC for the front end, although I could bump it up to 300VDC. The regulator design is very important. The current sources I use are very simple. Red LED plus PNP transistor and a few other parts. Reliable and I don't run into capacitance problems. The current set is stable.
-Chris
Thank you to everybody!
The tubes stage as I said is a principle schematic. In the reality, there are:
- relay switching the output to ground since the DC = 0 and that works if any kind of problem (f.ex. absence of mains) occurs
- three different power for the tubes stage
- a servo circuit that help the output to be very close to 0 volt DC
So no problem for the tubes stage! It's save enough, and deliver only signal - NO DC - .
There is another delay circuit that swich on the SS stage after the tube stage is on (around 1 minute) and DC = 0.
So no problem.
There is a servo also on the SS stage for the DC voltage at output. But this is always slow (around 200mV) also without servo.
So my problem (I don't know if it could be solved, anyway) still remain.
Now I'm entring on point C on my ciruit, and everything works fine.
BUT, I would like to not pass throug diodes, and enter straigh to points A and B
(maybe using another or two BJT, with NO voltage amplification, seen that is gave exclusively by the tubes stage)
Otherwise the tube stage could be more simple and I could use two capacitor (around 2 uF) starting from the same point and entering on same time on points A and B. I know already...
Thank you, and if you could write a little schematic about...
Best wishes!
The tubes stage as I said is a principle schematic. In the reality, there are:
- relay switching the output to ground since the DC = 0 and that works if any kind of problem (f.ex. absence of mains) occurs
- three different power for the tubes stage
- a servo circuit that help the output to be very close to 0 volt DC
So no problem for the tubes stage! It's save enough, and deliver only signal - NO DC - .
There is another delay circuit that swich on the SS stage after the tube stage is on (around 1 minute) and DC = 0.
So no problem.
There is a servo also on the SS stage for the DC voltage at output. But this is always slow (around 200mV) also without servo.
So my problem (I don't know if it could be solved, anyway) still remain.
Now I'm entring on point C on my ciruit, and everything works fine.
BUT, I would like to not pass throug diodes, and enter straigh to points A and B
(maybe using another or two BJT, with NO voltage amplification, seen that is gave exclusively by the tubes stage)
Otherwise the tube stage could be more simple and I could use two capacitor (around 2 uF) starting from the same point and entering on same time on points A and B. I know already...
Thank you, and if you could write a little schematic about...
Best wishes!
gianmaria said:
Gianmaria, you will need to disable the BJT stage servo as this will no longer be needed for a DC connection between the stages, but the input of the tube side servo needs to be connected to the output of the BJT stage, since that is where you really want 0V DC.
Regarding avoiding the diodes - you have to provide an AC path that bypasses them. There is really no simpler way to do this except to put two capacitors, one between points A and C, and one between B and C.
You could also, assuming you reconnect the tube end DC servo input from the BJT stage output, forget anbout point C and put a single capacitor across A and B, and input the signal from the tube stage at either A or B. The DC servo will take care that the output of the tube stage goes 3x Vbe + or - (depending on wether you use point A or B as inputs), in order for the BJT stage output to be at 0V DC. Even if the bias chain (between A and B) was put in circuit with the cathode current of you pentode (then you do not need extra resistors to = and - rail of the BJT stage), you would still need a capacitor between A and B. You could replace the diodes with a VBE multiplier - the standard circuit used to generate bias on many amps via a small transistor mounted on the heatsink with them - but even that would need a capacitor across the equivalent of points A and B as it is also a 'sort of' diode.
Finally, you could use something like the circuit I am attaching - but this circuit is really an amplifier strapped for unity gain, so you again either need caps across the 6.8k resistors (and even that is not a very good solution), or across the C/E of both the bias adjust transistors, or you risk HF instability. The lefthand side BJTs are small signal and in thermal contact with the heatsink. To be honest I would rather use a regular Vbe multiplier bypassed by a cap than this circuit.
Attachments
anatech said:
No, vertical FETs and one of the reasons I do not need over 60Vpp or so from the tubes is that it is very difficult to find good complements amongst VMOS, but I did manage it for rail voltages up to some +-45V, which is my maximum unloaded rail voltage.
In case I do decide for DC coupling, I need to split B+ into two halves, positive and negative. Both need to be well regulated, particulairly B- as that is the tube input reference - which also implies the use of prime quality bypass on the B- rail. This is the biggest argument against DC coupling - what would be the AC coupling caps for the output, become a single, vastly larger, more expensive and lower quality cap, or, if you ant a quality cap, you are looking at 20uF 200VDC foil caps - not cheap at all. The cost issue, however, becomes academic if I find a good surplus cap...
Regarding pentode followers, they should be, as far as follower action goes, better than triodes. Still, the point is fairly academic as the triodes for currents in question (10-20mA) still have a mu of 10 or more, so even as a follower their gain is sufficiently close to 1 (>0.9). The use of a pentode follower was more of an idea, since a pentode was already available inside an ECL86.
ECL86 alowes +-100V heater to cathode differential both for triode and pentode, and in my circuit this is just BARELY enough. Since the tube uses the same heater supply for both the triode and the pentode section, and the DC coupled version uses +-150V rails for the tubes and +-40V for the MOSFETs, the cathode of the follower is clamped to the +40V supply via diode, and the filament is biassed halfway between the -150 and +40V supplies. At full output there is only 5V to spare off of the 100V maximum
The PCL86 is more forgiving with 150V max differential and is the preferred tube for this design.An alternative would be to use ECC83 as voltage amp and ECC82 or 6922 as follower, in this case they would have separate heater supplies and no problems.
In a DC coupled version, FETs are very easy to bias and stabilise. OTOH, my design has only a single pair, biassed quite rich (250mA) and a small amount of global feedback (5dB) - more than that and the harmonic distribtion falls to pieces.
The extra 20V should be of no great consequence, but in this case, assuming you are not violating the maximum ratings, more is better. Current sources with a single comatible BJT tend to work better for currents that are a bit more than the .9mA I'm using for the ECC83/triode from the PCL86. The low current tends to lowr the gm of the BJT resulting in fairly lowish output impedances, so no advantage against a simple bootstrap - but that's for my chosen circuit. Getting into the 5-10mA region would present a completely different picture, so your solution is probably better for your design.
Dear ilimzn, all,
thank you for your detailed explanation and for your schematic!
The point is clearer to me now.
Everything was born from some experiences:
I noticed very clearly and without any doubt, that the sound has got an improvement (is more neutral, clear, controlled) if (with exactly the same schematic BJT stage and Tubes stage), I entered to point A and B (of course with two capacitors), instead of point C (without capacitor or with one capacitor, exactly the same type I used when entered to points A and B).
So if I have to use the two capacitors on poits A and B, that is seems to be - at the end - the best solution, I would go to change the tubes stage to a simplest one.
My request was to try, with the schematic I gave, to have (as in origin entered on point C) STILL a copletely DC coupled amplifier.
Infact I don't need anymore the big complexity of the power supply, of the servo etc... only to avoid the output caps...
I can very much easily use a SRPP followed by a cathode follower (to lower at maximum possible the output impedence) with two capacitor of 1,5 or 2,2 uF and entering points A and B.
Of course the problem now will be direct to the technology of the two coupling capacitors and I can assure you that the differences between a polipropylen and a paper in oil, is clearly evident and are not dreams...
Very good high frequencies of the paper in oil but a little confused bass, exactly the contrary for the poliprop who has got a very detailed and clear bass freq but is very "sparkly" and disturbing on the high freq...
Dear Ilimzn, all, have you any comment on these consideration?
Thank you so much, regards, G.M.
thank you for your detailed explanation and for your schematic!
The point is clearer to me now.
Everything was born from some experiences:
I noticed very clearly and without any doubt, that the sound has got an improvement (is more neutral, clear, controlled) if (with exactly the same schematic BJT stage and Tubes stage), I entered to point A and B (of course with two capacitors), instead of point C (without capacitor or with one capacitor, exactly the same type I used when entered to points A and B).
So if I have to use the two capacitors on poits A and B, that is seems to be - at the end - the best solution, I would go to change the tubes stage to a simplest one.
My request was to try, with the schematic I gave, to have (as in origin entered on point C) STILL a copletely DC coupled amplifier.
Infact I don't need anymore the big complexity of the power supply, of the servo etc... only to avoid the output caps...
I can very much easily use a SRPP followed by a cathode follower (to lower at maximum possible the output impedence) with two capacitor of 1,5 or 2,2 uF and entering points A and B.
Of course the problem now will be direct to the technology of the two coupling capacitors and I can assure you that the differences between a polipropylen and a paper in oil, is clearly evident and are not dreams...
Very good high frequencies of the paper in oil but a little confused bass, exactly the contrary for the poliprop who has got a very detailed and clear bass freq but is very "sparkly" and disturbing on the high freq...
Dear Ilimzn, all, have you any comment on these consideration?
Thank you so much, regards, G.M.
********* R I G H T M E S S A G E ***************
Dear ilimzn, all,
thank you for your detailed explanation and for your schematic!
The point is clearer to me now.
My request and my intention was to try, modifyingwith the schematic I gave, STILL to have a copletely DC coupled amplifier also without enetered on point C.
-----------------------------------------------------------------------------------
Everything was born from some experiences:
I noticed very clearly and without any doubt, that in my circuits, the sound has got an improvement (is more neutral, clear, controlled) if I entered to point A and B (of course with two capacitors), instead of point C (without capacitor or with one capacitor, exactly the same type I used when entered to points A and B).
-----------------------------------------------------------------------------------
If I have to use two capacitors to enter points A and B, and this seems to be - at the end - the best solution, I would go to change the tubes stage to a simplest one.
Infact I don't need anymore the big complexity of the dual power supply, servo on the tubes stage etc... only to avoid the output caps!
I can very much easily use a SRPP followed by a cathode follower (to lower at maximum possible the output impedance), with two capacitor of 1,5 or 2,2 uF - 630v and entering to points A and B.
Of course the problem now will be direct to the technology of the two coupling capacitors and I can assure you that the differences between a polipropylen and a paper in oil, is clearly evident and are not dreams...
Very good high frequencies for the paper in oil but basses are little confused, exactly the contrary for the poliprop that has got a very detailed and clear bass freq but is very "sparkly" and disturbing on the high freq...
Dear Ilimzn, all, have you anything to say about my consideration?
Thank you so much, regards, G.M.
Dear ilimzn, all,
thank you for your detailed explanation and for your schematic!
The point is clearer to me now.
My request and my intention was to try, modifyingwith the schematic I gave, STILL to have a copletely DC coupled amplifier also without enetered on point C.
-----------------------------------------------------------------------------------
Everything was born from some experiences:
I noticed very clearly and without any doubt, that in my circuits, the sound has got an improvement (is more neutral, clear, controlled) if I entered to point A and B (of course with two capacitors), instead of point C (without capacitor or with one capacitor, exactly the same type I used when entered to points A and B).
-----------------------------------------------------------------------------------
If I have to use two capacitors to enter points A and B, and this seems to be - at the end - the best solution, I would go to change the tubes stage to a simplest one.
Infact I don't need anymore the big complexity of the dual power supply, servo on the tubes stage etc... only to avoid the output caps!
I can very much easily use a SRPP followed by a cathode follower (to lower at maximum possible the output impedance), with two capacitor of 1,5 or 2,2 uF - 630v and entering to points A and B.
Of course the problem now will be direct to the technology of the two coupling capacitors and I can assure you that the differences between a polipropylen and a paper in oil, is clearly evident and are not dreams...
Very good high frequencies for the paper in oil but basses are little confused, exactly the contrary for the poliprop that has got a very detailed and clear bass freq but is very "sparkly" and disturbing on the high freq...
Dear Ilimzn, all, have you anything to say about my consideration?
Thank you so much, regards, G.M.
gianmaria said:
Yes, i understood that. But even so, you cannot avoid AC currents passing through the diode string. The reason you can hear a marked difference when they do is that the dynamic resistance of the diodes is fairly high because a relatively low DC current is passing through them (<1mA) and, with AC signal, is not even nearly constant. This is something that you will not see with coupling caps as they will essentially be in series for AC signals, across point A and B. A capacitor between points A and B would be necessary either way, to give the AC currents a low and more linear impedance. It's just that with two caps from the tube stage to points A and B, you get a cap across points A and B 'for free'.
The situation above could be improved by replacing the resistors to + and - SS suply rails with current sources. If nothing else, you could use dual bootstrapping from the output (since your output stage has voltage gain = 1). This would effectively increase the input impedance of the SS stage as seen by the tubes, which will mean better performance from your caps, but it also requires two more caps (lower voltage), fortunately, these have a far lower influence on sound than the coupler caps.
Another possibility, that requires DC coupling with the tubes, and therefore:
1) Dual power supplies for the tubes
2) -B for tubes carefully filtered and quality cap decoupled to GND
3) DC servo for the whole amp
4) Startup and DC protection via relais or similar
...would be to put the diodes and pot (vith adjusted values) in circuit with the Rk of the tube follower. Point A would connect to it's cathode, point B to the cathode resistors. You MUST use a quality cap between A and B still to prevent diodes from passing AC current (it will go through cap instead). The cap only sees a small voltage (about 3.4 V) so the choice of caps is much wider.
The resulting amp would not have a point C but would be DC coupled and would avoid the problem of low current through diodes and hence high rd of the diodes and associated distortion for LF, while the C in paralel with diodes takes care of HF.
Apart from that, you are back to AC coupling, and simplified stages. You will probably still need a DC servo for the SS stage.
How about using a paralel combination of different technology caps? Or even one type from tube output to point A and another to point B? This way you could tailor the sound to your liking.
Hi gianmaria,
At this point I think it's time for you to build and experiment. Ilimzn has made some good points you should consider, his solution is closer to your wishes than mine would ever be. 'Tis why I don't say much now.
Through building this you will discover all truths in time.
-Chris
At this point I think it's time for you to build and experiment. Ilimzn has made some good points you should consider, his solution is closer to your wishes than mine would ever be. 'Tis why I don't say much now.
Through building this you will discover all truths in time.
-Chris
In answer to your original question as how to get rid of the diodes, this is called a diamond buffer:
http://www.audiodesignguide.com/my/Cool_Follower1.GIF
You could make the output a triple for more current gain by using just one driver and connect the outputs like the Threshold S300 series of amplifiers:
http://www.diyaudio.com/forums/attachment.php?postid=38107&stamp=1024679649
http://www.audiodesignguide.com/my/Cool_Follower1.GIF
You could make the output a triple for more current gain by using just one driver and connect the outputs like the Threshold S300 series of amplifiers:
http://www.diyaudio.com/forums/attachment.php?postid=38107&stamp=1024679649
Hi GianMaria
your topology was covered by Aloia some times ago.
try this italian link and go to post n.10
http://www.audiocostruzioni.com/forum/topic.asp?ARCHIVE=&whichpage=1&TOPIC_ID=520
ciao
Federico
your topology was covered by Aloia some times ago.
try this italian link and go to post n.10
http://www.audiocostruzioni.com/forum/topic.asp?ARCHIVE=&whichpage=1&TOPIC_ID=520
ciao
Federico
djk said:
I keep finding that A Ciuffoli's designs are full of assumptions which very seldomly hold true.
In his diamond buffer, the bias current depends (quite heavily) on the rail voltages, and by what divine providence he guarantees the idle current to be 'optimal class AB' remains a mystery. It is also HARDLY a never before seen design. Even Walt Jung that is often credited with it, does not credit himself with coming up with the design.
A diamond buffer would be a good idea, but for it to work properly at high power, it requires the first stage to be fed from supply rails via current sources - this insures non-switching operation of the first stage if the source current is properly chosen. With current sources, the two parallel capacitors will either not be needed at all or could be replaced by one of vastly smaller capacitance.
Regarding the Threshold, it uses a CFP pair with multiple second stage - nothing new there either. CFPs can be used instead of the output BJTs in the diamond buffer, keeping in mind that one CFP (with all it's resistors) replaces a single BJT in the diamond buffer, WITHOUT it's emitter resistor.
Using a CFP has the advantage of a shallower and wider gm 'notch' at crossover. However, because of the above, you effectively have two Re per side in series (0.18 ohm in AC's diagram) in order to get stable bias current AND current sharing, one in every emitter of the power BJTs (final stage) and another one in the 'effective' emitter of the CFP (connection of emitter of first and collector of second stage of the CFP).
The typical cookbook approach for thermally stabilising a CFP by (in this case) putting the first stage of the buffer on the same heatsink as the first stage of the CFP does not assure proper thermal tracking and absence of runaway! With dual Re's, all transistors on the same heatsink, it will work - but you will have some power loss due to the double Re's compared to simple emitter follower.
You're the one that asked for a single drive point without diodes, the diamond circuit does that.
"With current sources, the two parallel capacitors will either not be needed at all or could be replaced by one of vastly smaller capacitance."
Without the caps the output stage is very limited in current unless you bias it fully into class A. With the caps it can move into class AB.
The current sources should be constructed with a diode connected transistor as the voltage reference. This diode should then be thermally linked to the outputs. As the outputs heat up so will the diode, and thus decrease the reference voltage for the current sources.
You can get the basic idea from the LC End Millenium or the Hawk A18 as part of the DC servo design they have. At this point the Threshold CFP output stage could be used, and it would have less drop from the rail than a triple EF.
"With current sources, the two parallel capacitors will either not be needed at all or could be replaced by one of vastly smaller capacitance."
Without the caps the output stage is very limited in current unless you bias it fully into class A. With the caps it can move into class AB.
The current sources should be constructed with a diode connected transistor as the voltage reference. This diode should then be thermally linked to the outputs. As the outputs heat up so will the diode, and thus decrease the reference voltage for the current sources.
You can get the basic idea from the LC End Millenium or the Hawk A18 as part of the DC servo design they have. At this point the Threshold CFP output stage could be used, and it would have less drop from the rail than a triple EF.
djk said:
No, it wasn't me, it was Gianmaria. I would have used a different topology with the bias chain in the cathode circuit of the follower.
The caps need to be there to bootstrap the bias voltage otherwise the current available to turn on the output BJTs lowers as the output voltage increases, at some point the output BJTs will current-limit the output. In class A, this point would be for input voltages equal to power rails or more, but since the point is defined by load current, even that would not insure proper operation if complex loads were used.
It would also have been possible to bootstrap the second stage from it's output, this would also create virtual current sources to it's bases, as well as avoid problems with hard cap discharge at clipping (see below).
This was precisely why I said that if current sources were used instead of the 10W resistors, this would not be a problem (and less heat would be generated).
The advantage of the capacitors is that the voltage at the output stage bases can swing higher than the power rails, resulting in lower losses - up to the point of saturation of a driver transistor, which will at that point clam the input signal by forward biasing it's CB junction.
You pay for that by having to pass audio current through an electrolytic cap, as well as some other considerations at clipping.
Alternatively a two-transistor source can be used, the BJT with the sensing resistor between B and E should be on the heatsink.
Actually, that is not a given. Combining a diamond buffer and CFP for good thermal stability requires extra emitter resistors, so you get a bit more loss at full output current. That is, comparing to the original diamond buffer. For tripple EF, you get the drop at the final Re + 2x Vbe - you can drive the first stage of the EF to satuiration since one Vbe extra over it's supply rails is not a problem for the tube driving stage.
On the Diamond buffer, you have to rely on bootstrapping (capacitors in AC's diagram) or provide higher supply rails and current sources from them, to the first stage of the buffer - the driver for the buffer can't do a thing here. Even with bootstrapping, you only get the bias voltage to shift up over the supply rail, and that's approximately 2x Vbe. Increasing driver voltage will do nothing as the CB junction of the first diamond buffer stage becomes forward biassed and clamps the input voltage. Assuming ideal output BJTs, 1xVbe over rail voltage will saturate the output transistor. Anything over that will be fed into the rail and output, and will partially discharge the large capacitors, through forward biassed BC and BE junctions of the output BJT, and through the driver BJT - this will be a large amount of current, though limited by the base current available to the first stage BJT. The bigger problem is the discharge of the caps, which means that for some time later the bufer will be unbderbiassed, and produce distortion. The rail loss of the stage will equal output BJT saturation loss plus voltage drop on an equivalent of 0.18/3 ohm, resistor.
Things become more complex for a CFP used as the second stage of the diamond buffer. The CFP current starves it's output transistors exactly at saturation, at Vc = Vb, assuming zero saturation voltage on the first stage of the CFP. Thius is helpful for clipping recovery, but implies you have 1 x Vbe loss by default. You also have one effective 0.06 ohm Re for the second stage of the CFP (0.18/3) for current sharing and one more in place of the 0.18/3 that would be found in the simple 2-stage implementation of the diamond buffer - without this, there would be no thermal stability. The rail loss equals voltage drop on the two equivalent Re's, i.e. Iout x (0.12), plus 1 x Vbe. For this, the input voltage to the CFP needs to be Iout x 0.06 + Vbe + Vout, i.e. Iout x 0.06 below the power rail. In this case the bootstrap experiences potentially more severe discharge as anything over that voltage discharges the large caps through the driver of the CFP and one 0.06 ohm Re, and the driver of the diamond buffer. Again, the current is limited by the base current available to the first stage of the diamond buffer, but the cap can discharge lower, and for very high output currents, even charge in reverse. So, the output stage can theoretically slip into class B or even slightly into class C for a short while.
So, comparing all three, you get (Vre = voltage on Re):
2 x Vbe + Vre for EF
Vre for straight diamond buffer
2 x Vre + Vbe for CFP diamond buffer,
Keeping in mind that diamond buffers may become underbiassed for a short while after clipping severely.
The straight diamond buffer has smallest losses but lowest current gain, and is reasonably easy to stabilise thermally.
The CFP diamond buffer leads over the EF as long as Vre < Vbe, but has high current gain and smoothest transfer curve, but can be difficult to stabilise thermally (T versus Ibies is a complex function).
The EF has high gain but is less linear than the CFP diamond buffer, but is dead easy to stabilise thermally, and does not have problems at clipping, but may not have a single signal entry point.
As far as coupling goes, it should be noted that all of these need some sort of capacitor somewhere, even when DC coupled to the tube stage. For both diamond buffers, this will be the cap between bases of the output stage (either single BJT or CFP). Straight diamond buffer needs large caps, CFP diamond buffer and EF need small caps. ALL would benefit from current sources to provide input biassing current, instead of simple resistors off of the supply rails, which may relax the requirement of the caps. In the case of the EF, the current source may be provided by the circuit in the last stage of the tube driver, if DC coupling is used, the diamond buffers do not have that option.
Let me just finish off that in my own design I intend on using MOSFETs, which largely removes all problems with rail loss, leaving the source follower as the most logical configuration.
Hi!
Please remind my two circuits...
The differences entered to point A and B (join file BJT2) and entering on point C (file BJT1), with exactly the same conditions, circuits etc., will be not only very clear when you listen, but will be also different at mesurements.
Infact using Spectralab, FFT analisys, in the same conditions as I repeat, I obtained the following results:
- with the "BJT 2" we have at 10 volt rms output (on 8 OHm resistive), around 0,1% THD+N with 2nd armonic predominant on 3rd armonic.
- With the "BJT 1" we have THD+N 0,07% with 3rd armonic predominant (also at 10 volt rms output on 8 OHM resistive)!
ONLY BECAUSE THE SIGNAL PASS or NOT PASS THROUGH DIODES!!
Ilimzn (thank you) has already explained why there are the motivations... but I couldn't belive to my ears.
I repeat I try many time the experiences to be sure, and always the same result.
Please remind my two circuits...
The differences entered to point A and B (join file BJT2) and entering on point C (file BJT1), with exactly the same conditions, circuits etc., will be not only very clear when you listen, but will be also different at mesurements.
Infact using Spectralab, FFT analisys, in the same conditions as I repeat, I obtained the following results:
- with the "BJT 2" we have at 10 volt rms output (on 8 OHm resistive), around 0,1% THD+N with 2nd armonic predominant on 3rd armonic.
- With the "BJT 1" we have THD+N 0,07% with 3rd armonic predominant (also at 10 volt rms output on 8 OHM resistive)!
ONLY BECAUSE THE SIGNAL PASS or NOT PASS THROUGH DIODES!!
Ilimzn (thank you) has already explained why there are the motivations... but I couldn't belive to my ears.
I repeat I try many time the experiences to be sure, and always the same result.
Attachments
Gianmaria, you may want to try three things:
1) Try using two caps as in BJT2, on BJT1, but in parallel coming into point C. This way you get the same coupling for both cases. his will probably reduce distortion but only by a negligible amount.
2) reduce 56k from supply rails to 27 or 22k, also re-adjust the bias (with the same setting of the pot it will be higher for 22k or 27k resistors), on BJT1. You would now need 2x the capacitance of the coupling cap for equal LF response, try it if you have the larger cap.
3) Add two caps (0.47 - 2.2u or even more, low voltage is OK as there are only a few V there) between points A and C, and B and C on BJT1. With these caps, BJT1 should behave like BJT2. Of course, you now have 2 more caps, but they can be low voltage, and you can optimize the single HV coupling cap from the tube stage on it's own.
Still, i think BJT2 is the simplest...
1) Try using two caps as in BJT2, on BJT1, but in parallel coming into point C. This way you get the same coupling for both cases. his will probably reduce distortion but only by a negligible amount.
2) reduce 56k from supply rails to 27 or 22k, also re-adjust the bias (with the same setting of the pot it will be higher for 22k or 27k resistors), on BJT1. You would now need 2x the capacitance of the coupling cap for equal LF response, try it if you have the larger cap.
3) Add two caps (0.47 - 2.2u or even more, low voltage is OK as there are only a few V there) between points A and C, and B and C on BJT1. With these caps, BJT1 should behave like BJT2. Of course, you now have 2 more caps, but they can be low voltage, and you can optimize the single HV coupling cap from the tube stage on it's own.
Still, i think BJT2 is the simplest...
I've been following this thread over the past couple of days because I too have an interest in building a Hybrid Amp More or less from scratch. I've experiment some and have studied all the schematics that I can actually find on the internet. But, I have a few questions that this post has brought up for me.
First, does using CCS's above and below the Vgs multiplier essentially make the input inpeadance of the output stage lower so that it is easier to drive with a tube front end? Example:
http://www.homestead.com/whaan/files/page6463.html
Second, How much current would an average single pair Source follower with a simple Vbe multiplier output stage need to drive it? (SK 1529 SJ200 for example)
Third - what would be the lowest voltage that you would suggest running a 6922 front end with? (trying to keep this as low as I can so i can use Trannies I already have, ie max voltage would be about 120V positive supply)
Sorry to Hijack this thread, but I felt that my questions sort of fit. I've prototyped a low voltage hybrid using a 12AX7 with a simple AB sorce follower output with Mosfets. While the highs and the mids sounded fine, the bass just didn't cut it. I sort of figured that the tube just didn't have enough current to drive the output. The tube stage was a simple plate follower loaded by a JFET CCS.
Thanxs and I hope this thread stays alive, as I've learded a bunch from it.
-Doug
First, does using CCS's above and below the Vgs multiplier essentially make the input inpeadance of the output stage lower so that it is easier to drive with a tube front end? Example:
http://www.homestead.com/whaan/files/page6463.html
Second, How much current would an average single pair Source follower with a simple Vbe multiplier output stage need to drive it? (SK 1529 SJ200 for example)
Third - what would be the lowest voltage that you would suggest running a 6922 front end with? (trying to keep this as low as I can so i can use Trannies I already have, ie max voltage would be about 120V positive supply)
Sorry to Hijack this thread, but I felt that my questions sort of fit. I've prototyped a low voltage hybrid using a 12AX7 with a simple AB sorce follower output with Mosfets. While the highs and the mids sounded fine, the bass just didn't cut it. I sort of figured that the tube just didn't have enough current to drive the output. The tube stage was a simple plate follower loaded by a JFET CCS.
Thanxs and I hope this thread stays alive, as I've learded a bunch from it.
-Doug
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.