I constantly try to educate myself on operating principles of various circuits and just the other day I came across an interesting – and very puzzling – one. Take a look at the power amplifier schematic of Traynor 200KB:
http://www.traynoramps.com/downloads/servman/sm200kb.pdf
At first glance the circuit may look kind of simple, just an amp with an ordinary opamp-based voltage amplifier section - and to increase the voltage swing the opamp is “bootstrapped” by modulating its supply rails according to output signal swing. (Why is the PNP side bootstrap different by the way?) Anyway, further looking got me really puzzled:
Take a look at the bias circuit, in fact the whole branch driving the PNP output Darlingtons. What is the idea of this configuration?
What is the circuit around transistors Q15 and Q18 doing? My guess is that it is some kind of DC servo that keeps the inverting input of U5B in about zero voltage.
Q19, Q28 and Q16 must be some additional protection based on the thermistor, right?
Then, take a look at the feedback network: How is the gain defined when the resistor ratio is so odd. (It’s not an error). It also looks like the input is bootstrapped (via R141) but with such odd topology I really can’t be certain.
Often when I run across an odd circuit that I don’t understand I simulate it. This time it didn't help: I ran a SPICE simulation (of a simplified circuit that omitted the combined VI/thermistor protection and OTA-based limiters) and it shows that the amp has only slightly higher gain than unity, which doesn’t make any sense since I firmly believe that the amplifier truly doesn’t have an input sensitivity of about 30Vpeak (which is what the simulation showed me). The preamp couldn’t even deliver it. I’m obviously overlooking some very important point.
How can I find more information about this circuit? Books from Slone and Self discuss only very basic Lin circuit derivatives and most white papers about OpAmp power buffers or such seem to be no help either. What is this kind of circuit called? Has it been patented? Why use it in the first place? I faintly recall seeing this circuit topology before in some insanely complex public address amplifier (can’t recall which but I don’t mean the other Yorkville sound products). Any help is truly appreciated.
http://www.traynoramps.com/downloads/servman/sm200kb.pdf
At first glance the circuit may look kind of simple, just an amp with an ordinary opamp-based voltage amplifier section - and to increase the voltage swing the opamp is “bootstrapped” by modulating its supply rails according to output signal swing. (Why is the PNP side bootstrap different by the way?) Anyway, further looking got me really puzzled:
Take a look at the bias circuit, in fact the whole branch driving the PNP output Darlingtons. What is the idea of this configuration?
What is the circuit around transistors Q15 and Q18 doing? My guess is that it is some kind of DC servo that keeps the inverting input of U5B in about zero voltage.
Q19, Q28 and Q16 must be some additional protection based on the thermistor, right?
Then, take a look at the feedback network: How is the gain defined when the resistor ratio is so odd. (It’s not an error). It also looks like the input is bootstrapped (via R141) but with such odd topology I really can’t be certain.
Often when I run across an odd circuit that I don’t understand I simulate it. This time it didn't help: I ran a SPICE simulation (of a simplified circuit that omitted the combined VI/thermistor protection and OTA-based limiters) and it shows that the amp has only slightly higher gain than unity, which doesn’t make any sense since I firmly believe that the amplifier truly doesn’t have an input sensitivity of about 30Vpeak (which is what the simulation showed me). The preamp couldn’t even deliver it. I’m obviously overlooking some very important point.
How can I find more information about this circuit? Books from Slone and Self discuss only very basic Lin circuit derivatives and most white papers about OpAmp power buffers or such seem to be no help either. What is this kind of circuit called? Has it been patented? Why use it in the first place? I faintly recall seeing this circuit topology before in some insanely complex public address amplifier (can’t recall which but I don’t mean the other Yorkville sound products). Any help is truly appreciated.
That's what I see (and not being an expert ;-) :
The opamp input is bootstrapped, otherwise it would run out of it's input voltage range. So we have both positive and negative feedback here, and the total feedback must be negative, of course.
Q20 sets the bias, quite conventional.
The termistor with associated circuit modulates current sources Q19/Q16, which shift overcurrent protection via Q14/Q17 to lower current threshold when the outputs run hot.
Q18/Q15 are in parallel and look like a clamp to keep the opamp in closed loop when the amp clips.
The parallel leg with D25 etc for the neg. bootstrapped supply looks strange, I suspect they found a transient or startup problem and this is the catch for it.
Quite a strange circuit, at any rate...
- Klaus
The opamp input is bootstrapped, otherwise it would run out of it's input voltage range. So we have both positive and negative feedback here, and the total feedback must be negative, of course.
Q20 sets the bias, quite conventional.
The termistor with associated circuit modulates current sources Q19/Q16, which shift overcurrent protection via Q14/Q17 to lower current threshold when the outputs run hot.
Q18/Q15 are in parallel and look like a clamp to keep the opamp in closed loop when the amp clips.
The parallel leg with D25 etc for the neg. bootstrapped supply looks strange, I suspect they found a transient or startup problem and this is the catch for it.
Quite a strange circuit, at any rate...
- Klaus
Ok, an update: This was one of those circuits that do not simulate very well in SPICE. I got it running after replicating the whole OTA section; then I started removing stuff. Eventually the only thing that was left from the additions was a simple OpAmp buffer stage in between the ideal voltage supply (that served as the input signal source) and input node of the amp. With this in place the simulation converged well and the circuit’s gain increased notably: The input sensitivity went from 30Vpeak to about 2Vpeak. I had replicated the input’s series resistance and AC coupling already but obviously that wasn’t enough. The wonders of SPICE and its too ideal performance.
I still don’t understand how the circuit defines its gain, though. Klaus, you mentioned positive feedback, could you elaborate that? I see that the 2Vpeak input voltage is actually “pumped up” to 4.5Vpeak already at the node that ties input to common via R45 or muting FET. Am I seeing effects of positive feedback? Could you also explain me the Q15/Q18 clamp? - I just don't get it...
Your post confirmed my suspicions about the over current protection, I was thinking that the additional circuit might do something like that.
All in all, things started to make more sense once the simulation started running ok and allowed experimenting. The “clamp” can be removed so it obviously isn’t a key part of the topology like I first thought. (One big distraction removed). I suppose the other OpAmp is simply there to increase the drive current of the PNP Darlingtons so that the drive current is about equal to that of the NPN pair. In simulation it can be removed as well and after that the circuit starts to look pretty basic (aside the strange feedback loop). But why is the bias circuit/PNP base referenced to the modulated (low) negative rail instead of the –45V rail?
I still don’t understand how the circuit defines its gain, though. Klaus, you mentioned positive feedback, could you elaborate that? I see that the 2Vpeak input voltage is actually “pumped up” to 4.5Vpeak already at the node that ties input to common via R45 or muting FET. Am I seeing effects of positive feedback? Could you also explain me the Q15/Q18 clamp? - I just don't get it...
Your post confirmed my suspicions about the over current protection, I was thinking that the additional circuit might do something like that.
All in all, things started to make more sense once the simulation started running ok and allowed experimenting. The “clamp” can be removed so it obviously isn’t a key part of the topology like I first thought. (One big distraction removed). I suppose the other OpAmp is simply there to increase the drive current of the PNP Darlingtons so that the drive current is about equal to that of the NPN pair. In simulation it can be removed as well and after that the circuit starts to look pretty basic (aside the strange feedback loop). But why is the bias circuit/PNP base referenced to the modulated (low) negative rail instead of the –45V rail?
It looks like I’m finally starting to see some light in this: I was familiar with the basic technique of bootstrapping opamps to increase their maximum signal swing but this way to do it was/is new to me. It seems that the circuit topology is known as “suspended supply operation”. So far I found only few examples showing a simplified schematic of the basic configuration. Does someone know any good application note or whitepaper that would discuss this circuit in greater detail?
2 refs:
http://www.edn.com/index.asp?layout=article&articleid=CA45890
and Linear AN-67 p58
for op amp ps modeling problems
Analog AN138
and http://focus.ti.com/lit/an/sboa027/sboa027.pdf
http://www.edn.com/index.asp?layout=article&articleid=CA45890
and Linear AN-67 p58
for op amp ps modeling problems
Analog AN138
and http://focus.ti.com/lit/an/sboa027/sboa027.pdf
Hi teemuk,
For the gain calculation (or simulation), on can reduce the whole stage to a simple ideal (supplyless) opamp circuit, with both positive and negative feedback.
We have the positive feedback resistors, that is 15.5k (14k+1.54k) and 2.74k and the negative feedback R values are 4.12k and 681R. Let a=4.12k/(4.12k+681)=0.858 and b=15.5k/(15.5k+2.74)=0.850, these are the feedback factors, negative and positive resp.
We know that at the opamp Vn=Vp in the closed loop condition.
Vn=aVo
Vp=Vi + b(Vo-Vi) (here the bootstrap from Vo shows up)
aVo = Vi + bVo -bVi
aVo - bVo = Vi - bVi
Vo(a-b) = Vi(1-b)
Gain= Vo/Vi = (1-b)(a-b)
Plugging in the number gives a gain of 18.
This is very sensitve to resistor tolerance because we divide through the difference of the factors. And when b > a the circuit latches to one of the rails, too much positive feedback.
The imporant thing is that Vp is lifted and is close to Vo, well within input voltage bounds.
The bias VBE-multiplier is hooked to the bootrapped rail because then a simple resistor gives a quite good current source to feed it. Otherwise a transistor current source would be needed, with extra dissipation. The second opamp is only a current buffer.
The clamp works in the following way:
Assume the opamp output goes positive so that Q18 starts to conduct. Then D46 becomes forward biased and injects a positive current into the summing node (Vn) which reduces output voltage. Think of Q18 as a variable resistor which shunts the main feedback resitor. Same Q15 for negative opamp output (referred to the amp's output), but at a different (lower) level, because of the drop on the bias network for the lower half of the output . This asymmetry can also be seen in the clipping indicator network around the optocoupler.
jcx is right that simulation of opamps *with* exact behaviour of their supply lines and corresponding limits has to be carefully checked. Most models go wrong as far as the supplies are concerned.
- Klaus
For the gain calculation (or simulation), on can reduce the whole stage to a simple ideal (supplyless) opamp circuit, with both positive and negative feedback.
We have the positive feedback resistors, that is 15.5k (14k+1.54k) and 2.74k and the negative feedback R values are 4.12k and 681R. Let a=4.12k/(4.12k+681)=0.858 and b=15.5k/(15.5k+2.74)=0.850, these are the feedback factors, negative and positive resp.
We know that at the opamp Vn=Vp in the closed loop condition.
Vn=aVo
Vp=Vi + b(Vo-Vi) (here the bootstrap from Vo shows up)
aVo = Vi + bVo -bVi
aVo - bVo = Vi - bVi
Vo(a-b) = Vi(1-b)
Gain= Vo/Vi = (1-b)(a-b)
Plugging in the number gives a gain of 18.
This is very sensitve to resistor tolerance because we divide through the difference of the factors. And when b > a the circuit latches to one of the rails, too much positive feedback.
The imporant thing is that Vp is lifted and is close to Vo, well within input voltage bounds.
The bias VBE-multiplier is hooked to the bootrapped rail because then a simple resistor gives a quite good current source to feed it. Otherwise a transistor current source would be needed, with extra dissipation. The second opamp is only a current buffer.
The clamp works in the following way:
Assume the opamp output goes positive so that Q18 starts to conduct. Then D46 becomes forward biased and injects a positive current into the summing node (Vn) which reduces output voltage. Think of Q18 as a variable resistor which shunts the main feedback resitor. Same Q15 for negative opamp output (referred to the amp's output), but at a different (lower) level, because of the drop on the bias network for the lower half of the output . This asymmetry can also be seen in the clipping indicator network around the optocoupler.
jcx is right that simulation of opamps *with* exact behaviour of their supply lines and corresponding limits has to be carefully checked. Most models go wrong as far as the supplies are concerned.
- Klaus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.