Hi all. I was musing on emitter followers driven by opamps this afternoon (when I was supposed to be doing something else 
) and all of a sudden this circuit idea popped into my head. Has anyone seen this before? Or have I finally invented something entirely new?
In the diagram Q1 emitter is held at approximately half rail and driven by the emitter lead. If for example Q1 emitter rose, it's base would be stationary and it's collector would rise as Q1 operates as a grounded base cct. Consequently Q2's base is driven upward pulling it's emitter upward and making an output signal and at the same time pulling Q1's base upward. This last event tends to subtract from the upward movement of Q1 emitter input so we have negative feedback. Q1 emitter moves up and it's base follows it almost exactly.
So then, we have a cct with a gain of 1 and all the internal gain is swallowed up by feedback. Late this afternoon I threw together a mosfet version first, using a 2N7000 for Q1. It is fairly fast,switching in 6 nS or so. Anyway, I used *no* gate resistors and the thing did NOT oscillate! That feedback loop must be FAST!
It worked but it's needs a bit of refining that I didn't have time to do. It may have some promise as something useful. The Vbe drift is almost entirely dependent on cold Q1, not hot Q2 so it should be easy to have stable bias in a comp symm version. And with so much gain ploughed into a tight feedback loop it should be very linear. On a square wave input the rise time was about 200nS and the fall time 50nS for a 15 volt step. That fall time is 300v per uS. I's jes' bustin' fer termorra ta come so's ah kin git orn an' try sum sterf.
) and all of a sudden this circuit idea popped into my head. Has anyone seen this before? Or have I finally invented something entirely new?In the diagram Q1 emitter is held at approximately half rail and driven by the emitter lead. If for example Q1 emitter rose, it's base would be stationary and it's collector would rise as Q1 operates as a grounded base cct. Consequently Q2's base is driven upward pulling it's emitter upward and making an output signal and at the same time pulling Q1's base upward. This last event tends to subtract from the upward movement of Q1 emitter input so we have negative feedback. Q1 emitter moves up and it's base follows it almost exactly.
So then, we have a cct with a gain of 1 and all the internal gain is swallowed up by feedback. Late this afternoon I threw together a mosfet version first, using a 2N7000 for Q1. It is fairly fast,switching in 6 nS or so. Anyway, I used *no* gate resistors and the thing did NOT oscillate!
That feedback loop must be FAST! It worked but it's needs a bit of refining that I didn't have time to do. It may have some promise as something useful. The Vbe drift is almost entirely dependent on cold Q1, not hot Q2 so it should be easy to have stable bias in a comp symm version. And with so much gain ploughed into a tight feedback loop it should be very linear. On a square wave input the rise time was about 200nS and the fall time 50nS for a 15 volt step. That fall time is 300v per uS. I's jes' bustin' fer termorra ta come so's ah kin git orn an' try sum sterf.
Attachments
There are future...in audio design...
Hi Ciclotron!!
Why not change Q2 for a darlington...because as Q1 is a comum base stage it don't have current gain...so only Q2 is providing current gain!
The power suplly also must be well smothed as the base of Q2 is referenced to the + suplly and the emiter of Q2 is referenced to ground...so any ripple in the suplly will be amplified...
Regards
PS: I like creativ people!!
Hi Ciclotron!!
Why not change Q2 for a darlington...because as Q1 is a comum base stage it don't have current gain...so only Q2 is providing current gain!
The power suplly also must be well smothed as the base of Q2 is referenced to the + suplly and the emiter of Q2 is referenced to ground...so any ripple in the suplly will be amplified...
Regards
PS: I like creativ people!!
Hey Circlotron isn´t your circuit similar with this circuit ?
If you go on like this you´ll definitely find something entirely new.
Shouldn´t you end your LM317-idea before exploring other galaxies?
BTW: I love "crazy" and simple circuit ideas; they always get me tempted to try. So keep them coming.
Where have you got your 100mH coils from?
Regards
Jens
If you go on like this you´ll definitely find something entirely new.
Shouldn´t you end your LM317-idea before exploring other galaxies?
BTW: I love "crazy" and simple circuit ideas; they always get me tempted to try. So keep them coming.
Where have you got your 100mH coils from?
Regards
Jens
Attachments
Circlotron: When I need a compact voltage regulator, I occasionally incorporate something quite similar to the circuit that you posted. In my case, a constant-current circuit replaces the top resistor (or bottom resistor, for the opposite polarity). I nearly always use bipolars for Q1, but have used both MOSFETs and bipolars for Q2. Should you encounter any ringing or tendancy towards instability, I would suggest inserting a small capacitor from the Q1 collector - Q2 base node to ground. If you use a MOSFET for Q2, adding a small value gate-stopper resistor can also help control any tendancy towards oscillation.
This circuit works well and has shown itself to be completely trouble-free in practice.
best, jonathan carr
This circuit works well and has shown itself to be completely trouble-free in practice.
best, jonathan carr
Circlotron said:Hi all. I was musing on emitter followers driven by opamps this afternoon (when I was supposed to be doing something else
In the diagram Q1 emitter is held at approximately half rail and driven by the emitter lead. If for example Q1 emitter rose, it's base would be stationary and it's collector would rise as Q1 operates as a grounded base cct. Consequently Q2's base is driven upward pulling it's emitter upward and making an output signal and at the same time pulling Q1's base upward. This last event tends to subtract from the upward movement of Q1 emitter input so we have negative feedback. Q1 emitter moves up and it's base follows it almost exactly.
So then, we have a cct with a gain of 1 and all the internal gain is swallowed up by feedback. Late this afternoon I threw together a mosfet version first, using a 2N7000 for Q1. It is fairly fast,switching in 6 nS or so. Anyway, I used *no* gate resistors and the thing did NOT oscillate!
It worked but it's needs a bit of refining that I didn't have time to do. It may have some promise as something useful. The Vbe drift is almost entirely dependent on cold Q1, not hot Q2 so it should be easy to have stable bias in a comp symm version. And with so much gain ploughed into a tight feedback loop it should be very linear. On a square wave input the rise time was about 200nS and the fall time 50nS for a 15 volt step. That fall time is 300v per uS. I's jes' bustin' fer termorra ta come so's ah kin git orn an' try sum sterf.
...nice work circlotron..
Yessir, three bags full sir!
Try it - it's a real rocket.
Nahhh. I always have several ideas on the go at the same itme so if I get a bit stale with one then I just swap to the other. Anyway, your wish is my command. I was in actual fact going to try an LM317 in this cct just for a laugh but it worked so well that I completely ditched any further attempts with discrete stuff. This cct is a killer! The supply voltage limit is set by the main mosfet and the little pnp bipolar. Another good thing is that the '317 doesn't handle the entire signal voltage swing, it only handles the difference between input and output of the main mosfet. That means the slew rate requirements of the '317 are greatly reduced. It has an internal gain of 80dB and this is multiplied by the gain of the mosfet and all of this is ploughed into linearising the mosfet.joensd said:
Try it - it's a real rocket.
Attachments
Another refinement that I haven't tried yet but will soon is to have a -15v negative floating rail as well, with the zero volt (+) point going to the output and the neg rail going to the collector of Q1 instead of Q1 going to ground as it is now. This way the collector-emitter voltage will be virtually constant. As the Q1 emitter goes downward for example with signal, the main fet source goes down and therefore the two floating rails go downward all by the same amount taking Q2 collector with it. The fact that the upper positive floating rail goes downward in this case by the same amount as the signal means that the mosfet gate pullup resistor voltage drop doesn't change and therefore the current in that vertical line all the way down through to Q1 is constant. Therefore Q1 Vbe stays constant during signal excursions too. See, the cct only *looks* simple. 
The point is, Q1 has constant current through it, constant voltage across it , and constant Vbe. Hopefully this is the recipe for the closest approach to zero distortion? I don't have a distortion meter, so could someone perhaps tell me how valid is this approach?
The point is, Q1 has constant current through it, constant voltage across it , and constant Vbe. Hopefully this is the recipe for the closest approach to zero distortion? I don't have a distortion meter, so could someone perhaps tell me how valid is this approach?... been there ...done that...
Hi Circlotron,
I used this configuration back in the early eighties as the input stage of the Tesserac Audio mc head amp that I designed ( with Ed Portelli). It also makes a terrific low impedance microphone pre-amp input stage.
It's a nice circuit for specific applications - I still haven't found a better one for mc input use.
ciao
James
Hi Circlotron,
I used this configuration back in the early eighties as the input stage of the Tesserac Audio mc head amp that I designed ( with Ed Portelli). It also makes a terrific low impedance microphone pre-amp input stage.
It's a nice circuit for specific applications - I still haven't found a better one for mc input use.
ciao
James
Complementary Symmetry
This is how I would do it for a comp symm version. Note the bottom reg is a '337 not a '317 like the top one. It has a different pinout too. One slight disadvantage of this cct is the feedback is taken from the sources, not the final output terminal so the 0.1 ohm source resistors are outside the feedback loop. Therefore the amp will have a minimum output impedance of 0.1 ohms. This can be lowered if you take a second feedback loop from the output terminal though.
This is how I would do it for a comp symm version. Note the bottom reg is a '337 not a '317 like the top one. It has a different pinout too. One slight disadvantage of this cct is the feedback is taken from the sources, not the final output terminal so the 0.1 ohm source resistors are outside the feedback loop. Therefore the amp will have a minimum output impedance of 0.1 ohms. This can be lowered if you take a second feedback loop from the output terminal though.
Attachments
The one you've been waiting for but didn't realise...
And now, the circuit that almost has more transformers than a cat has lives, a full blown (small "c") circlotron / tetrahedron amp using every trick in the book. This is an all stops out , over the top, to the max, full on honker that I am going to throw together tonight. It embodies everything I think is fun in an amplifier. Look at the symmetry in that schematic! Yum. I only hope it works as good as it looks. Wish me luck!
And now, the circuit that almost has more transformers than a cat has lives, a full blown (small "c") circlotron / tetrahedron amp using every trick in the book. This is an all stops out , over the top, to the max, full on honker that I am going to throw together tonight. It embodies everything I think is fun in an amplifier. Look at the symmetry in that schematic! Yum. I only hope it works as good as it looks. Wish me luck!
Attachments
Hi Circlotron,
I love all the iron in your design! A man after my own heart. Intersting circuit too - good luck with it.
On the Tesserac circuit - I'll have to ask Ed if I can reproduce it here as he still owns the company and all it's IPR. I'll keep you updated. It's a nice mc front end and I organised the biasing to allow direct connection of the cartridge with the common base stage having it's current varied to give different input impedances - in this mode I had no need to start adding input capacitance to tame mc top end resonance - just carefully adjust the cb current until it measured flat...
ciao
James
I love all the iron in your design! A man after my own heart. Intersting circuit too - good luck with it.
On the Tesserac circuit - I'll have to ask Ed if I can reproduce it here as he still owns the company and all it's IPR. I'll keep you updated. It's a nice mc front end and I organised the biasing to allow direct connection of the cartridge with the common base stage having it's current varied to give different input impedances - in this mode I had no need to start adding input capacitance to tame mc top end resonance - just carefully adjust the cb current until it measured flat...
ciao
James
It worked. The input transistors needed a 100nF cap from base to emitter to stop things oscillating. Also, I forgot that the Vbe of those transistors in not under control of any feedback loop so it changes with temperature. Luckily it makes the mosfet bias go down with temperature, not up as is usually the case. The bias is very touchy. 1mV change in bias voltage = 1mV change in transformer half primary voltage drop in the source side of things, = 11mA change in bias current. But the good part is once you have set it, it stays put because of the temperature stable voltage of the LM317's. Only the input BJT's let it down. The big question is whether to pursue it further. I might have to see how a complementary symmetry version goes first.
Latest incarnation of things
I keep on coming back to the choke loaded source follower because it seems to work so well and mostly refuses to blow up. The latest modification is to replace the 330 ohm gate pullup resistor with a 30mA constant current source. The '317 used in this position has a slope resistance (how much the current changes for a given change in voltageacross it) of about 250k so the lower '317 thinks it has a 250k load resistor so it's gain is pretty high. Actually the apparent resistance is about 20 times higher than this because the IN terminal of the upper '317 is feed from a bootstrapped (great word, that) supply that swings about 95% of the gate voltage movement so the current change in the top '317 is about 20 times smaller as stated. It therefore looks like a 5 meg load so the lower '317 gain must be pretty high.
I ran the thing with a 14.3 volt main supply and it put out 12.22 watts into 6 ohms before clipping. The 1uF cap from base to emitter of the PN200 is not a typo. The instantaneous Vbe changes very very little so it needs a decent sized cap to have any effect, in this case to stop the thing from oscillating.
To be continued, of course.
I keep on coming back to the choke loaded source follower because it seems to work so well and mostly refuses to blow up.
The latest modification is to replace the 330 ohm gate pullup resistor with a 30mA constant current source. The '317 used in this position has a slope resistance (how much the current changes for a given change in voltageacross it) of about 250k so the lower '317 thinks it has a 250k load resistor so it's gain is pretty high. Actually the apparent resistance is about 20 times higher than this because the IN terminal of the upper '317 is feed from a bootstrapped (great word, that) supply that swings about 95% of the gate voltage movement so the current change in the top '317 is about 20 times smaller as stated. It therefore looks like a 5 meg load so the lower '317 gain must be pretty high. I ran the thing with a 14.3 volt main supply and it put out 12.22 watts into 6 ohms before clipping. The 1uF cap from base to emitter of the PN200 is not a typo. The instantaneous Vbe changes very very little so it needs a decent sized cap to have any effect, in this case to stop the thing from oscillating.
To be continued, of course.
Attachments
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.