I've just been outside in the shed playing amplifiers for the last hour after I posted the last bit. It's the middle of winter here! 
Brrr!
A better explanation of what is happening is this - say I have the bias set to 200mA. Either fet can swing to 399.99 mA and the other swings down to almost zero. When a fet gets to 401mA the other is driven to zero and the gate drive drops from the threshold voltage to zero fairly quickly because the off fet cannot swing any lower and this looks like an error voltage so the local feedback tries it's hardest to do something.
Actually, now that you mention it, the gate drive is pretty slow when the waveform dives to zero and returns, but that is measured on the driver side of the gate resistor. I think part of the problem is I only have about 8mA current sinking capability and perhaps the output capacitance of the P-channel driver is too much for it. It is a MTP2P50E 500v 2A 6R. Also, at the current levels we are talking about, the transconductance of it is woeful. The curve is almost horizontal still!
I brought home today several BC640's, 80v, 1A, 140 Hfe @ 10mA, Ft 50Mhz or something like that. Maybe I should go and give them a try. Thanks heaps for your interest in my project
GP.
Brrr!A better explanation of what is happening is this - say I have the bias set to 200mA. Either fet can swing to 399.99 mA and the other swings down to almost zero. When a fet gets to 401mA the other is driven to zero and the gate drive drops from the threshold voltage to zero fairly quickly because the off fet cannot swing any lower and this looks like an error voltage so the local feedback tries it's hardest to do something.
Actually, now that you mention it, the gate drive is pretty slow when the waveform dives to zero and returns, but that is measured on the driver side of the gate resistor. I think part of the problem is I only have about 8mA current sinking capability and perhaps the output capacitance of the P-channel driver is too much for it. It is a MTP2P50E 500v 2A 6R. Also, at the current levels we are talking about, the transconductance of it is woeful. The curve is almost horizontal still!
I brought home today several BC640's, 80v, 1A, 140 Hfe @ 10mA, Ft 50Mhz or something like that. Maybe I should go and give them a try. Thanks heaps for your interest in my project
GP.
Righteo. After sleeping on things it seems to me the basic problem is assymetric drive to the main fet gates despite symmetrical drive from the phase splitting input tranny. This is a result of the assymetric transfer curves because the fet that is starting to conduct is going up a steepening curve whereas the offgoing fet is going out on a flattening cureve that eventually cuts off. The local feedback tries to linearise both of these *independently* and therein lies the problem.
The usual way when driving things push pull with a transformer is that when one gate or grid goes up the other one goes down *by the same amount*. Here things are getting snagged because both gates have a certain amount of independence. What I am going to do is in the schematic where T2a and T2b will be added and these force the ac component of the gate drive to be equal and opposite. We only need 1 or 2 volts swing so it needn't be a big tranny.
Something is intuitively telling me that there is something similar to a long tailed pair here, and the way they are balanced can be used here too. Could somebody please read my mind and tell me what I am trying to think of? A pair of resistors in series from gate to gate and sensed at the halfway point should always have a steady potential if the drive is symmetrical; somehow that could be used to straighten things out. Gotta think some more.
GP.
The usual way when driving things push pull with a transformer is that when one gate or grid goes up the other one goes down *by the same amount*. Here things are getting snagged because both gates have a certain amount of independence. What I am going to do is in the schematic where T2a and T2b will be added and these force the ac component of the gate drive to be equal and opposite. We only need 1 or 2 volts swing so it needn't be a big tranny.
Something is intuitively telling me that there is something similar to a long tailed pair here, and the way they are balanced can be used here too. Could somebody please read my mind and tell me what I am trying to think of?
A pair of resistors in series from gate to gate and sensed at the halfway point should always have a steady potential if the drive is symmetrical; somehow that could be used to straighten things out. Gotta think some more.GP.
Attachments
Circlotron amp etc
Hi Circlotron,
Congrats! You just discovered device non-linearity. The problem you describe, of the two output devices having very different "going up, going down" characteristics is at the root of all attempts to make a linear amp, whether it is with BJT's, FET's, tubes, or what have you. What you haven't mention is trying to overlap the two device curves around the take over region so that the nonlinearity of one device is compensated by the nonlinearity of the other. There is only a very narrow area where this works, and then still not 100%. That is also the reason why in Class AB amps there is an optimum bias current for least distortion. Bias below AND above this point give worse results, until you get in Class A country of course. Self explains this much better in his book.
Cheers, Jan Didden
Hi Circlotron,
Congrats! You just discovered device non-linearity. The problem you describe, of the two output devices having very different "going up, going down" characteristics is at the root of all attempts to make a linear amp, whether it is with BJT's, FET's, tubes, or what have you. What you haven't mention is trying to overlap the two device curves around the take over region so that the nonlinearity of one device is compensated by the nonlinearity of the other. There is only a very narrow area where this works, and then still not 100%. That is also the reason why in Class AB amps there is an optimum bias current for least distortion. Bias below AND above this point give worse results, until you get in Class A country of course. Self explains this much better in his book.
Cheers, Jan Didden
Latest idea. Maybe just a bit on the rude side, but worth a try. The idea is that the negative going signals to the driver bases (turning incoming fet on) are unaffected, but the positive going ones (turning outgoing fet off) are attenuated by approximately 1% as they go above zero + ~0.2v by the 200k pulling down on the 1k going up. This means that the driver and main fet will never be turned fully off, hopefully just pulled on a little bit more as it gets dragged over by the incoming fet. That way the feedback loop of the inactive fet always has something to do.
A refinement would be to make the inactive side drive simply a small negative offset when the drive is positive, not a percentage offset of instantaneous signal level. I'll build it up about about 6 hours from now, it's still lunchtime at work.
A refinement would be to make the inactive side drive simply a small negative offset when the drive is positive, not a percentage offset of instantaneous signal level. I'll build it up about about 6 hours from now, it's still lunchtime at work.
Attachments
It is certainly true that there is an "optimum" bias point
for complementary bipolar followers, optimum from the
standpoint of measured distortion.
Interestingly, although biasing at higher levels gives rise
to somewhat more distortion, listeners almost always seem
to prefer the sound of the heavier bias.
I suggest you try this yourself and see if you get the same
result.
for complementary bipolar followers, optimum from the
standpoint of measured distortion.
Interestingly, although biasing at higher levels gives rise
to somewhat more distortion, listeners almost always seem
to prefer the sound of the heavier bias.
I suggest you try this yourself and see if you get the same
result.
Hey, thanks for the feedback guys.
Just branching off toward the definitely agricultural approach in amplifier design for a moment, if we sense the AC component of the gate / source voltage in a voltage follower, i.e. the error signal and insert a magnified version of the error in series with the gate drive then we should be able to reduce nonlinearity on each fet independently by the same factor as the voltage magnification of the feedback. Here is a cct I will try out this weekend using only transformers for this feedback. It could become the worlds worst oscillator of course, but I'll give it a go.
GP.
Just branching off toward the definitely agricultural approach in amplifier design for a moment, if we sense the AC component of the gate / source voltage in a voltage follower, i.e. the error signal and insert a magnified version of the error in series with the gate drive then we should be able to reduce nonlinearity on each fet independently by the same factor as the voltage magnification of the feedback. Here is a cct I will try out this weekend using only transformers for this feedback. It could become the worlds worst oscillator of course, but I'll give it a go.
GP.
Attachments
It worked. Amazing...
Well I put this thing together and the gate transformers were actually 1:6 stepup, not 1:11, (1:10 turns ratio). With a 4v peak sinewave from ct to one end of the phase splitter tranny and driving 4 ohm loudspeaker at 400Hz (lots of 4's there) the signal at the source was 3.9v peak, and the signal at the gate was 4.5v peak. So then, the AC error was reduced from 0.6v to 0.1v, a 6:1 improvement as per the 6:1 error voltage stepup via the transformers.
Frequency response was a bit ordinary though. Heh heh. Output voltage dipped to about 2/3 in the 2-4 kHz range. Actually the gate transformer windings had no resistive loading so the stray inductance & capacitance of the windings might have caused this. The phase splitter tranny has interwinding capacitances of about 5nF(!) and yet it behaves itself quite well provided it is driven from a moderately low impedance. As mentioned before, the way this trans is wound is with hexfilar twisted wire, ~525 turns and 2 of the strands are paralleled to from the primary and 4 strands are put in series to form the gate to gate secondary. The secondary has 4x 220 ohm resistors across it to damp it, 1 resistor across each winding section.
Anyway, time for a big rev up! I attached a 6 ohm 50 watt resistor as a load because that is all I could find, and it got to about 16.6 watts output and both of the Hexfets popped. I did notice a bit of oscillation on the waveform peaks before it quit and I think the gate transformer were ringing or something and causing the fets to cross conduct. Anyway, that was the end of that. The sound was surprisingly good for such a dreadful setup.
The next thing to do, at the risk of getting drummed out of diyaudio.com is to ditch the gate transformers and replace them with a pair of opamps with a gain of 10 and going up bit by bit. They will do a similar job as the transformers, comparing the difference between the gate signal and the source signal and dialling in some sort of correction. The two independent opamp supplies will be tied to their respective hexfet sources so the demands on the opamps will be eased greatly. They will only be required to handle the error component, not the entire audio signal. We'll see just how badly this one flops.
GP.
Well I put this thing together and the gate transformers were actually 1:6 stepup, not 1:11, (1:10 turns ratio). With a 4v peak sinewave from ct to one end of the phase splitter tranny and driving 4 ohm loudspeaker at 400Hz (lots of 4's there) the signal at the source was 3.9v peak, and the signal at the gate was 4.5v peak. So then, the AC error was reduced from 0.6v to 0.1v, a 6:1 improvement as per the 6:1 error voltage stepup via the transformers.
Frequency response was a bit ordinary though. Heh heh. Output voltage dipped to about 2/3 in the 2-4 kHz range. Actually the gate transformer windings had no resistive loading so the stray inductance & capacitance of the windings might have caused this. The phase splitter tranny has interwinding capacitances of about 5nF(!) and yet it behaves itself quite well provided it is driven from a moderately low impedance. As mentioned before, the way this trans is wound is with hexfilar twisted wire, ~525 turns and 2 of the strands are paralleled to from the primary and 4 strands are put in series to form the gate to gate secondary. The secondary has 4x 220 ohm resistors across it to damp it, 1 resistor across each winding section.
Anyway, time for a big rev up! I attached a 6 ohm 50 watt resistor as a load because that is all I could find, and it got to about 16.6 watts output and both of the Hexfets popped.
I did notice a bit of oscillation on the waveform peaks before it quit and I think the gate transformer were ringing or something and causing the fets to cross conduct. Anyway, that was the end of that. The sound was surprisingly good for such a dreadful setup.The next thing to do, at the risk of getting drummed out of diyaudio.com
is to ditch the gate transformers and replace them with a pair of opamps with a gain of 10 and going up bit by bit. They will do a similar job as the transformers, comparing the difference between the gate signal and the source signal and dialling in some sort of correction. The two independent opamp supplies will be tied to their respective hexfet sources so the demands on the opamps will be eased greatly. They will only be required to handle the error component, not the entire audio signal. We'll see just how badly this one flops. GP.
Phase Funnies
Hello Graham,
Sorry to hear that it went splatt.
I was going to mention much earlier that I think you need to take care to load the secondaries properly, elsewise because of parastics they can ring (resonate) and cause a peaking response (phono carts same deal), and phase funnies may cause what you are now experiencing.
I think that if you load the secondaries correctly you are likely to have more success.
Regards, Eric.
Bernard, Harry or Jocko may have some good info here.
Hello Graham,
Sorry to hear that it went splatt.
I was going to mention much earlier that I think you need to take care to load the secondaries properly, elsewise because of parastics they can ring (resonate) and cause a peaking response (phono carts same deal), and phase funnies may cause what you are now experiencing.
I think that if you load the secondaries correctly you are likely to have more success.
Regards, Eric.
Bernard, Harry or Jocko may have some good info here.
Happy days are here again
SO now the CDA definitely pulls into high gear... After the belly flop of the other day, and especially just before that, trying all sorts of schemes with a Sziklai pair setup things were becoming decidedly un-fun I always wondered what things would go like if I stuck a pair of opamps in there to do the work and despite the risk of "slings and arrows" I gave it a shot fully expecting it to be a mess of oscillations and instability. Well, I wasn't disappointed this time (i.e. it worked ok!) I can crank the quiescent current all the way down to 5mA and it still produces a visually perfect 400 Hz sinewave with no xover distortion. YeeHah. I'll still run it at 200mA as I planned though. On a good recording it sounds really clean and smooth, to my ears anyway. It sounds like FUN, so it's doing exactly what it is supposed to do. Now finally I can get on with the final construction of the thing. Will post a pic of the test setup later.
What I will need now is a preamp of some sort to drive this thing. If I raise the resistors on the secondary side of the input transformer to 1K each then the input impedance of the primary side will be 250 ohms. It will need a 20v p/p drive. Can anybody suggest something fairly basic and "no frills" to drive this. All I need is bass and treble control and an input for a CD. Nothing flash.
GP.
SO now the CDA definitely pulls into high gear... After the belly flop of the other day, and especially just before that, trying all sorts of schemes with a Sziklai pair setup things were becoming decidedly un-fun
I always wondered what things would go like if I stuck a pair of opamps in there to do the work and despite the risk of "slings and arrows" I gave it a shot fully expecting it to be a mess of oscillations and instability. Well, I wasn't disappointed this time (i.e. it worked ok!) I can crank the quiescent current all the way down to 5mA and it still produces a visually perfect 400 Hz sinewave with no xover distortion. YeeHah. I'll still run it at 200mA as I planned though. On a good recording it sounds really clean and smooth, to my ears anyway. It sounds like FUN, so it's doing exactly what it is supposed to do. Now finally I can get on with the final construction of the thing. Will post a pic of the test setup later.What I will need now is a preamp of some sort to drive this thing. If I raise the resistors on the secondary side of the input transformer to 1K each then the input impedance of the primary side will be 250 ohms. It will need a 20v p/p drive. Can anybody suggest something fairly basic and "no frills" to drive this. All I need is bass and treble control and an input for a CD. Nothing flash.
GP.
Attachments
Ain't gunna bump no more...
When a complementary symmetry amp powers up the output usually thumps one direction momentarily. The circlotron topology can't do this by nature of it's balanced configuration. But you can get an "internal" thump with consequent uncontrolled current flow through both fets in series. The speaker doesn't see this current and so it doesn't make a sound.
Almost the whole time during prototyping of this thing I would have a 2R2 1/4 watt resistor in series with the power supply rails so if anything went wrong it would pop the resistor. I must have replaced it 30 or 40 times! Cheaper than fuses and I could measure the quiescent current using it too. A certain number of times it conked out at switch-on and it seems the opamps would overshoot their steady state settings and cause both hexfets to turn on full. This is very naughty. So what I have done is add a clamp diode going to each fet gate and a large cap that gets slowly charged up. At switch-on there is about a 1 second delay and then the quiescent current comes up to full value in a further 1 second. It is absolutely smooth and silent.
What I am thinking about is to have a setup where the cap can be discharged by a 1k resistor and a pair of relay contacts. Power on and then a moment later open the contacts when the main rails and opamps have settled down. At power-down, close the contacts first and then kill the power. If there is no program signal for a defined interval then short the contacts so quiescent current goes to zero and saves power. If a fuse has gone so that you can detect a voltage across it, don't open the contacts at all, or shut them if the amp was already running when it happened, and bring up a front panel fault led. Lots of possibilities that a little micro will make happen.
BTW, I haven't had any suggestions for a preamp as per the last post! Anyway, here it is as an attached file rather than a pic as before so I can read the download counter and gauge the level of interest here. Is there the remotest possibility that anyone else will have a go at making one of these amps?
GP.
When a complementary symmetry amp powers up the output usually thumps one direction momentarily. The circlotron topology can't do this by nature of it's balanced configuration. But you can get an "internal" thump with consequent uncontrolled current flow through both fets in series. The speaker doesn't see this current and so it doesn't make a sound.
Almost the whole time during prototyping of this thing I would have a 2R2 1/4 watt resistor in series with the power supply rails so if anything went wrong it would pop the resistor. I must have replaced it 30 or 40 times! Cheaper than fuses and I could measure the quiescent current using it too. A certain number of times it conked out at switch-on and it seems the opamps would overshoot their steady state settings and cause both hexfets to turn on full. This is very naughty. So what I have done is add a clamp diode going to each fet gate and a large cap that gets slowly charged up. At switch-on there is about a 1 second delay and then the quiescent current comes up to full value in a further 1 second. It is absolutely smooth and silent.
What I am thinking about is to have a setup where the cap can be discharged by a 1k resistor and a pair of relay contacts. Power on and then a moment later open the contacts when the main rails and opamps have settled down. At power-down, close the contacts first and then kill the power. If there is no program signal for a defined interval then short the contacts so quiescent current goes to zero and saves power. If a fuse has gone so that you can detect a voltage across it, don't open the contacts at all, or shut them if the amp was already running when it happened, and bring up a front panel fault led. Lots of possibilities that a little micro will make happen.
BTW, I haven't had any suggestions for a preamp as per the last post! Anyway, here it is as an attached file rather than a pic as before so I can read the download counter and gauge the level of interest here. Is there the remotest possibility that anyone else will have a go at making one of these amps?
GP.
I'm pretty sure they are bought from an overseas buying house. I'll find out. The ones I have are all second hand 'crumbs that have fallen from the table'. Farnell have the IXYS ones but you just don't want to know the price
Actually, I have a bunch of IRF540's and you are welcome to have a handful if you like. Free that is, but on the condition that you WILL use them for a circlotron amp. They are 100v 28 amp 70mR I forget the watts, TO-220. They would have about the same capability as a 2N3055 with the same rails. Good for experimenting till you work your way up to something that I will hear clear across the Nullarbor
GP.
Actually, I have a bunch of IRF540's and you are welcome to have a handful if you like. Free that is, but on the condition that you WILL use them for a circlotron amp. They are 100v 28 amp 70mR I forget the watts, TO-220. They would have about the same capability as a 2N3055 with the same rails. Good for experimenting till you work your way up to something that I will hear clear across the Nullarbor
GP.
Circlotron etc
Hi Circlotron,
That is one interesting circuit you developed there! I wish I had the time to try it out. Maybe later.
Two questions I have though. I noted the 10K gate resistors. They may be the reason for your limited frequency range. Any particular reason they are so large? I would probably try 220R or so here.
The gain of the opamps in the active feedback. This is limited to 15 times. The opamp really just gives out output voltage until its -input equals its +input, but it is hampered by the limited gain. It would do a much better job if it could run open-loop.
I guess you probably got oscillations when you increase the 15k over a certain value, let alone delete it. Do you have any info on this aspect of the design?
Cheers, Jan Didden
Hi Circlotron,
That is one interesting circuit you developed there! I wish I had the time to try it out. Maybe later.
Two questions I have though. I noted the 10K gate resistors. They may be the reason for your limited frequency range. Any particular reason they are so large? I would probably try 220R or so here.
The gain of the opamps in the active feedback. This is limited to 15 times. The opamp really just gives out output voltage until its -input equals its +input, but it is hampered by the limited gain. It would do a much better job if it could run open-loop.
I guess you probably got oscillations when you increase the 15k over a certain value, let alone delete it. Do you have any info on this aspect of the design?
Cheers, Jan Didden
Hi Jan.
The 10k resistors are a bit of an overkill for sure. I was just paranoid about the whole thing breaking into ultrasonic oscillation so I overdid it somewhat. I will in fact lower the value somewhat as time goes by, and see how it goes. Same goes for the opamp - just wanted to start off gentle. 15k was the first value I tried and so I just left it for the moment. Same goes for the other opamp R's and C's. Open loop would be the ideal way to go in some ways although it must have a gain of 1 at DC so the bias voltage fed in through the input transformer comes out the opamp multiplied by x1 I think there is a lot of scope for winding up the gain at audio frequencies, but above 20kHz or so I think it would be asking for trouble to keep the high gain. Hence the 470pF local feedback cap. What I will do is post a spreadsheet app I wrote last night that graphs the response and gain of the opamp when you fiddle the various R's and C's. I have a few little resistor sized inductors that I might put in there too to try and maintain high gain in the audio band but roll it off quicker >20kHz.
Frequency response was only really awful when I tried that silly cct a couple of posts back that used those extra pair of transformers to step up the error voltage to the gates. It's reasonable at the moment. I don't know just how reasonable because I don't have an oscillator so I have to run it from sinewaves generated from my pc sound card.
GP.
The 10k resistors are a bit of an overkill for sure. I was just paranoid about the whole thing breaking into ultrasonic oscillation so I overdid it somewhat. I will in fact lower the value somewhat as time goes by, and see how it goes. Same goes for the opamp - just wanted to start off gentle. 15k was the first value I tried and so I just left it for the moment. Same goes for the other opamp R's and C's. Open loop would be the ideal way to go in some ways although it must have a gain of 1 at DC so the bias voltage fed in through the input transformer comes out the opamp multiplied by x1
I think there is a lot of scope for winding up the gain at audio frequencies, but above 20kHz or so I think it would be asking for trouble to keep the high gain. Hence the 470pF local feedback cap. What I will do is post a spreadsheet app I wrote last night that graphs the response and gain of the opamp when you fiddle the various R's and C's. I have a few little resistor sized inductors that I might put in there too to try and maintain high gain in the audio band but roll it off quicker >20kHz. Frequency response was only really awful when I tried that silly cct a couple of posts back that used those extra pair of transformers to step up the error voltage to the gates. It's reasonable at the moment. I don't know just how reasonable because I don't have an oscillator so I have to run it from sinewaves generated from my pc sound card.
GP.
Opamp feedback stuff
Sounds a bit risky to me. Any drift will be multiplied by the opamp closed loop gain. But I suppose the phase response will be a bit better behaved. Anyway I did a little bit of messing around tonight and put the 15k up to 100k and the 470pF to 1nF and it seemed to go ok. Dropped the 1nF to 100pF and instant oscillation. Back to 1nf. Wound the bias back to induce xover distortion and looked at the opamp outputs. During the time when both fets are non-conducting the wave swings quick to hurry up and start the incoming fet. As soon as the fet starts conduction the opamp output changes from a near vertical transition to the arched top of a sinewave, but at that very junction there was a bit of ringing - about 10% of the total signal amplitude. I lowered the gate resistors from 10k to 1k and it reduced by 95%. Shoved a bit of music through the thing and it sounded really nice. Now that the gate resistors are 1k the soft start cap should be increased to 1000uF. The speadsheet I promised is not quite ready for human consumption yet. Basically things are going well. Lots of niceness continues to radiate out of the amplifier.
GP.
Sounds a bit risky to me.
Any drift will be multiplied by the opamp closed loop gain. But I suppose the phase response will be a bit better behaved. Anyway I did a little bit of messing around tonight and put the 15k up to 100k and the 470pF to 1nF and it seemed to go ok. Dropped the 1nF to 100pF and instant oscillation. Back to 1nf. Wound the bias back to induce xover distortion and looked at the opamp outputs. During the time when both fets are non-conducting the wave swings quick to hurry up and start the incoming fet. As soon as the fet starts conduction the opamp output changes from a near vertical transition to the arched top of a sinewave, but at that very junction there was a bit of ringing - about 10% of the total signal amplitude. I lowered the gate resistors from 10k to 1k and it reduced by 95%. Shoved a bit of music through the thing and it sounded really nice. Now that the gate resistors are 1k the soft start cap should be increased to 1000uF. The speadsheet I promised is not quite ready for human consumption yet. Basically things are going well. Lots of niceness continues to radiate out of the amplifier. GP.
Circlotron amp etc
Circlotron,
Yes, it all makes sense doesn't it.
In fact, lowering the gate resistor may allow you to increase the opamp gain, as there will be less loop phase shift to cause instability. And even if 100pF is unstable, any value substantially below 1nF would increase performance.
Jan Didden
"Any sufficiently advanced technology is indistinguishable from magic" [Robert Heinlein]
Circlotron,
Yes, it all makes sense doesn't it.
In fact, lowering the gate resistor may allow you to increase the opamp gain, as there will be less loop phase shift to cause instability. And even if 100pF is unstable, any value substantially below 1nF would increase performance.
Jan Didden
"Any sufficiently advanced technology is indistinguishable from magic" [Robert Heinlein]
Changing direction
I've been giving the whole thing a bit of thought, and I think for the time being at least I will skip using any active components between the driver tranny and the Hexfet gates and just feed them directly as per the cct right near the beginning of this thread. I want to stop being peeved and grieved! Maybe I'll resume the path I have been following sometime later on.
So... if I raise the driver trans secondary damping resistors to 1k each (a more reasonable figure BTW), up from 220 ohms each, then the total resistance across it will be 4k. The turns ratio is 1:4 stepup, so the impedance ratio is 1:4 squared, i.e. 1:16 so the input side will see 1/16 of 4000 ohms or 250 ohms.
With the rails I have, the loudspeaker output should swing + / - 40v (80v p/p) and seeing the outputs are source followers we need to feed the gates with a little more than that, say 90v p/p. Our nifty little transformer does it's stuff in the stepup department as only a transformer can do, so we only need 22.5 volts p/p into it, but into an impedance of 250 ohms. This works out to ~250 mW, not really very much.
Now... what to drive it with then? Several things spring to mind. I have a brand new pair of 6CM5 horizontal output tubes that I could run in class A triode config, 1 per channel. Or perhaps a scaled down version of a Zen output stage. Seeing this amplifier is going to be the best in the Universe I would then have to rename it Zen-ith would I not
SO then, what should I drive it with? Over to U.
GP.
I've been giving the whole thing a bit of thought, and I think for the time being at least I will skip using any active components between the driver tranny and the Hexfet gates and just feed them directly as per the cct right near the beginning of this thread. I want to stop being peeved and grieved! Maybe I'll resume the path I have been following sometime later on.
So... if I raise the driver trans secondary damping resistors to 1k each (a more reasonable figure BTW), up from 220 ohms each, then the total resistance across it will be 4k. The turns ratio is 1:4 stepup, so the impedance ratio is 1:4 squared, i.e. 1:16 so the input side will see 1/16 of 4000 ohms or 250 ohms.
With the rails I have, the loudspeaker output should swing + / - 40v (80v p/p) and seeing the outputs are source followers we need to feed the gates with a little more than that, say 90v p/p. Our nifty little transformer does it's stuff in the stepup department as only a transformer can do, so we only need 22.5 volts p/p into it, but into an impedance of 250 ohms. This works out to ~250 mW, not really very much.
Now... what to drive it with then? Several things spring to mind. I have a brand new pair of 6CM5 horizontal output tubes that I could run in class A triode config, 1 per channel. Or perhaps a scaled down version of a Zen output stage. Seeing this amplifier is going to be the best in the Universe I would then have to rename it Zen-ith would I not
SO then, what should I drive it with? Over to U.
GP.
Hi everybody! (Hi Dr. C!)
Well then. After being "deluged" by all the suggestions as to how to make a driver for my phase splitter tranny, I copy the example of Bill Oddie of "The Goodies" fame in an episode where he was mistreating Black and White Beauty, a non-descript horse (that VERY obviously had two people inside it), and I proclaim "Circlotron Doomsday Amplifier? She's mine! And I'll treat her any way I want! Bwuhahahaha!!
So then. What I cooked up on the weekend was a buffer amplifier with an opamp (what's that noise? Oh, it's just Mr. Pass running outside with his hand over his mouth) driving a complementary pair. The biasing and temperature compensation is always a bit of a nuisance for this kind of thing because it is a small version of an output stage in reality. What I did then was to tie the bases together and assuming they are both at zero volts, I pulled the upper emitter below zero with 100mA, and the lower emitter above zero with 100mA. I used a 12v battery for this because the supply has to be floating, and the 110 ohms makes the 100mA stay virtually constant no matter what the temperature.
Provided you don't try and pull too much current to the load, the thing stays in class A. With a dual trace scope I compared the bases to the output and it was virtually identical. Changing the load from 220 ohms resistive to 10 ohms paralleled with it at about 4v p/p o/p made NO visible difference i.e. way less than 1mV.
There is about 100 nano-seconds delay from opamp input to emitters output so phase shift is no problem so I was able to use the full gain of the opamp to make it linear, though it was pretty good anyway.
The next thing to do is drive the tranny and the output stage and close another feedback loop to a gain stage before the driver. The transformer has a phase shift of about 9.5 degrees at 15kHz. Is this good or bad? What should I do to my feedback loop to keep out of trouble with this transformer? Would adding a cap across the feedback resistor be the right thing to do here to give it a bit of phase lead? Now I am sailing in uncharted waters.
GP.
Well then. After being "deluged" by all the suggestions as to how to make a driver for my phase splitter tranny, I copy the example of Bill Oddie of "The Goodies" fame in an episode where he was mistreating Black and White Beauty, a non-descript horse (that VERY obviously had two people inside it), and I proclaim "Circlotron Doomsday Amplifier? She's mine! And I'll treat her any way I want! Bwuhahahaha!!
So then. What I cooked up on the weekend was a buffer amplifier with an opamp (what's that noise? Oh, it's just Mr. Pass running outside with his hand over his mouth) driving a complementary pair. The biasing and temperature compensation is always a bit of a nuisance for this kind of thing because it is a small version of an output stage in reality. What I did then was to tie the bases together and assuming they are both at zero volts, I pulled the upper emitter below zero with 100mA, and the lower emitter above zero with 100mA. I used a 12v battery for this because the supply has to be floating, and the 110 ohms makes the 100mA stay virtually constant no matter what the temperature.
Provided you don't try and pull too much current to the load, the thing stays in class A. With a dual trace scope I compared the bases to the output and it was virtually identical. Changing the load from 220 ohms resistive to 10 ohms paralleled with it at about 4v p/p o/p made NO visible difference i.e. way less than 1mV.
There is about 100 nano-seconds delay from opamp input to emitters output so phase shift is no problem so I was able to use the full gain of the opamp to make it linear, though it was pretty good anyway.
The next thing to do is drive the tranny and the output stage and close another feedback loop to a gain stage before the driver. The transformer has a phase shift of about 9.5 degrees at 15kHz. Is this good or bad? What should I do to my feedback loop to keep out of trouble with this transformer? Would adding a cap across the feedback resistor be the right thing to do here to give it a bit of phase lead? Now I am sailing in uncharted waters.
GP.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.