Here is an idea for an automatic bias control of a PP output stage.
It has probably been already explored, but here is my take anyway.
I think the schematic is self-explanatory enough; there are two options: bypassed or not.
When the base spreader is bypassed, the behavior exactly mimics that of a conventional circuit, except it is automatic.
If the bypass cap is removed, the circuit operates in real time, and becomes a non-switching class B.
The effect is noticeable in the reduced THD figure. Note that there is some ringing, because I didn't take the care to add any compensation, but these are not difficult issues
It has probably been already explored, but here is my take anyway.
I think the schematic is self-explanatory enough; there are two options: bypassed or not.
When the base spreader is bypassed, the behavior exactly mimics that of a conventional circuit, except it is automatic.
If the bypass cap is removed, the circuit operates in real time, and becomes a non-switching class B.
The effect is noticeable in the reduced THD figure. Note that there is some ringing, because I didn't take the care to add any compensation, but these are not difficult issues
Attachments
Nelson Pass has a patent on it. I am sure it can be updated by now. The big Adcom used it. I was thinking about using a IR detector to look at the output transistor case. Maybe a little less lag.
He also has a circuit to compensate for the driver stage so the bias servo using a thermister only had to figure out the output stage.
He also has a circuit to compensate for the driver stage so the bias servo using a thermister only had to figure out the output stage.
Basically it is very interesting and I commend your long time efforts on this topic.
However I am non proficent on the simulator and from the schematic I cannot see how it works.
The left side opto couplers have an open output (goes to a-b?) and the right side are even short. They could be replaced by two LEDs for that.
I am sure I miss something, however.
On the general idea I am more concern about the linearity of the OPTO response. Those stuff normally have an output that passes quite sharply from interdition to saturation. Maybe you might have to work with the base pin of the opto transistor, that in this opto model seems exposed.
However I am non proficent on the simulator and from the schematic I cannot see how it works.
The left side opto couplers have an open output (goes to a-b?) and the right side are even short. They could be replaced by two LEDs for that.
I am sure I miss something, however.
On the general idea I am more concern about the linearity of the OPTO response. Those stuff normally have an output that passes quite sharply from interdition to saturation. Maybe you might have to work with the base pin of the opto transistor, that in this opto model seems exposed.
Interesting, it would be nice to see how he implemented it
Yes they're in parallel with R5
Problem with any random LED is that they won't match the characteristics of the active opto.
Opto's come in 4-pack, which means good matching and tracking with temperature
The linearity is completely irrelevant here, because they arent in the noble path: they're in in an orthogonal control loop having no direct effect on the output
Last edited:
OK, starting to understand now.
The circuit does not thermally compensate for bias, it justs does auto-bias when needed in the crossover region.
U2 and U4 are OK, especially if the 4N25 is a quad pack, good practical idea.
But when is it that the U1 and U3 internal LEDs are not lit?
It seems to me that they are always forward biased with a good current something like (50-2Vled-Vdiode)/10K i.e. about 4mA.
What did I get wrong?
Are the Schottky diodes necessary for the basic functioning of the circuit?
The circuit does not thermally compensate for bias, it justs does auto-bias when needed in the crossover region.
U2 and U4 are OK, especially if the 4N25 is a quad pack, good practical idea.
But when is it that the U1 and U3 internal LEDs are not lit?
It seems to me that they are always forward biased with a good current something like (50-2Vled-Vdiode)/10K i.e. about 4mA.
What did I get wrong?
Are the Schottky diodes necessary for the basic functioning of the circuit?
Last edited:
Yes, but of course the automatic bias works against any type of perturbation, thermal included
The 4N25 isn't a quad, that is just what I had available in the simulator, but in reality quads can be found easily
When the corresponding transistor passes a sufficient current, they aren't lit, but that's not where interesting things happen: it is when the current becomes too low in any of the transistor: the opto then raises the current to a minimum value, about 180mA as you can see on the real time version
No, their role is simply to avoid wasting power whilst keeping a high enough sensitivity not be annoyed by matching or stability problems, but the details are unimportant and, if you build a real amplifier based on this concept, you have to understand how it works and design it from scratch: I used convenient values in my example, but they are certainly far from optimum in many ways
I have found the Pass patent:
Patent US4752745 - Opto-isolated bias circuit for operating push-pull amplifiers in class A and ... - Google Patents
It works very differently from my idea, and is suited to class A. In class AB, it will bring more problems than it solves
Patent US4752745 - Opto-isolated bias circuit for operating push-pull amplifiers in class A and ... - Google Patents
It works very differently from my idea, and is suited to class A. In class AB, it will bring more problems than it solves
I have used the same Schottky "trick" to keep the efficiency high of a multiple pair CFP output, putting them in parallel to each OP Re.
But using them as in your circuit, I fail to see how the Opto "senses" when the OP is drawing current. With the Schottky in place the maximun voltage variation that it can see is about Vd, so a few hudred mV. It seems to me too less to turn off the internal LED.
I mean U1 LED anode is always at +V1 (or , at least, at +V1 -Vd1) and the cathode of U4 LED is always at -V2, so it looks that U1 always "ON" (output saturated). Withour D1 /D2 and adjusting R3, it should work.
Where am I wrong?
Great concept, BTW.
But using them as in your circuit, I fail to see how the Opto "senses" when the OP is drawing current. With the Schottky in place the maximun voltage variation that it can see is about Vd, so a few hudred mV. It seems to me too less to turn off the internal LED.
I mean U1 LED anode is always at +V1 (or , at least, at +V1 -Vd1) and the cathode of U4 LED is always at -V2, so it looks that U1 always "ON" (output saturated). Withour D1 /D2 and adjusting R3, it should work.
Where am I wrong?
Great concept, BTW.
When they are across the Re's, they bring serious non-linearities, unless you use some tricks like Broskie's class C.
Here, in the collectors their influence is minimal, through Early effect or similar.
With the opto bias, no emitter resistance is required, which is best for linearity
Remove the schottkys if you prefer, they change nothing to the operation of the circuit.
Consider the pair of LEDs as a virtual diff amp, without explicit transistors
I bet that is why they quit using it. The motivation was probably to not have to factory set bias. Any hand-operation is expensive. I think the 585 used it.
here's an alternative recently published in Linear Audio:
Volume 6
Practical Electronic Control of Class AB output stage quiescent current Vol 6
Daniel Joffe
Controlling the quiescent current in class AB output stages has long been the Achilles heel of such amplifiers. Low crossover distortion depends critically on correct and stable quiescent current. Crossover distortion cannot be completely eliminated in AB designs due to the output devices switching in and out of conduction on alternate signal half cycles. The thermal stability of the bias control is also important for reliability of class AB amps with respect to thermal runaway. Daniel Joffe has designed a novel control circuit that prevents output devices to switch off completely, maintaining a small but critical quiescent current over the whole signal cycle. The circuit has no critical elements, uses only low-cost small-signal devices and can be grafted onto almost any amplifier.
Volume 6
Practical Electronic Control of Class AB output stage quiescent current Vol 6
Daniel Joffe
Controlling the quiescent current in class AB output stages has long been the Achilles heel of such amplifiers. Low crossover distortion depends critically on correct and stable quiescent current. Crossover distortion cannot be completely eliminated in AB designs due to the output devices switching in and out of conduction on alternate signal half cycles. The thermal stability of the bias control is also important for reliability of class AB amps with respect to thermal runaway. Daniel Joffe has designed a novel control circuit that prevents output devices to switch off completely, maintaining a small but critical quiescent current over the whole signal cycle. The circuit has no critical elements, uses only low-cost small-signal devices and can be grafted onto almost any amplifier.
OK, since you participate to this thread and you are the author of article in question, can you give an example circuit of your technique we can examine and discuss (otherwise it's just advertising for Linear Audio and yourself)
How is this going to work without a very big integrator? That 100u across the spreader is very bad for distortion.
Hello Elvee. Please show a schematic as pdf so as to print, study and understand this interesting circuit. Thank you.
No need for integrator, and if you think 100µ is bad for distortion, you have two options: the two extremes, no cap at all and you get NSW class B, or 10,000µ or more: at 2.5V working voltage it is neither big nor expensive
One more thing to consider: with the control active, a capacitor will not be a mere capacitor anymore: if it is too small, its behavior will be modified by the control loop
Not as pdf, but as png:
But I remind once more that the practical details as shown are just a support for the concept: if I implement this scheme in reality (which will probably happen some day), it will be completely different, because circuits that work really well generally tend to be very different from those that are easy to understand like this one.
Finally, one could find a use for the transistors of the unused optocouplers. I havent figured out a really interesting job for them yet, but this deserves more thoughts....
Attachments
Last edited:
Elvee. Please note the lettering on the optocoupling transistors. Both are a and b; but not a c as the front end shows.
The connections to a and b nodes are correct. c is just a reference point, unconnected to anything else for the time being.
I failed to discover where the optical part was hiding
also looks rather pointless for those of us who don't have a fab - the circuits all use area ratio "current sense/replication" of well matching devices on a single substrate
- Status
- Not open for further replies.