complementary feedback pair question

[Sleeper]

Sziklai pair (also known as a "complementary feedback pair" (CFP) or "compound transistor") is a configuration of two bipolar transistors, similar to a Darlington pair.[1] In contrast to the Darlington arrangement, the Sziklai pair has one NPN and one PNP transistor, and so it is sometimes also called the "complementary Darlington".

I've been looking around I'm not finding an answer. Hope someone can clue me in.

In your typical 2520 op amp circuit, and it's variants, Gain Blokes, Hybrids etc. there is a BD139 and a BD140 CFP that drives the output.
in between these transistors is usually a pair of resistors, sometimes 1ohm,  often 3r6  and just as often 10r

I need to substitute what I have in my box vs. whats on a schematic.
I'm 99% sure I can go up or down a few ohms and everything will still work fine, but I can't find any criteria to decide.
All the articles I found about the Sziklai pair just ignore these parts, but they must do something.
?
do they set the output impedence? raise the transistor bias? Keep the transistor within operating temperatures
It's usually part of the feedback loop - even if it's not on the op amp itself, it's in the op amp's circuit ???

Thanks
Sleeper

 

[PRR]

Since they are series to the output and load, they need to be "small" compared to load. Worst-case tends to be loudspeaker amps, where 0.5 Ohm emitter resistors feed 8 Ohm loads.

If D103 D104 were known (relative to Q101 Q102) size geometry and doping, then you could compute R144 R145 to get a current-ratio to the current in D103 D104 (which is known via R142 R143 and known supply voltages minus low diode drops).

In fact we don't know all the facts, so we select a trial value and see what current we get in Q101 Q102.

An old-old derivation suggests that, for good class AB crossover, there should be >50 milliVolts DC across each emitter resistor. More may be fine, less is scritchy.

But there is a clever trick which also affects the resistor values. Short the output, feed an input. As Q101 Base is pulled up, eventually D103 D104 pull up to the point that D106 conducts to the (shorted) output. Q101's Base won't go higher than "three" diode-drops above the output. Take-away Q101's base-emitter drop, you have at-most two diode drops across 2.7 Ohms. Pencil 1.2V across 2.7 Ohms, 444mA maximum current in Q101. (Same thing other-way for Q102.)

MJE180/170 is a fat pair which can handle 500mA easily. (However you also check the power dissipation, which "could" be 15V*0.44A or a half-dozen Watts!! In AC/audio-coupled duty, half that per device, still a bunch.)

If the emitter resistors are 10 Ohms then 1.2V/10r= 120mA max.

I don't see a need for 2.7 Ohms unless you are driving 37 Ohms.

10 ohms should be fine for 170 Ohm loads.

You find these in "heroic" mixers from days when opamps were expensive and you got the MOST out of each one. 37 Ohms suggests a 600-match distribution amp with *16* unshortable outputs. 170 is in-line for driving a 1:2 OT to +31dBm.

Typical modern loads are maybe 600, more often 10K. Levels rarely need to be +8dBu. I'd say you could use 22 or 33 Ohms.

 

[JohnRoberts]

Sziklai pair (also known as a "complementary feedback pair" (CFP) or "compound transistor") is a configuration of two bipolar transistors, similar to a Darlington pair.[1] In contrast to the Darlington arrangement, the Sziklai pair has one NPN and one PNP transistor, and so it is sometimes also called the "complementary Darlington".

I've been looking around I'm not finding an answer. Hope someone can clue me in.

Not sure what the question really is... the schematic you posted does not have a quasi-comp darlington connection. This topology was commonly used back in the day to make a power amp output stage with only one sex of power devices. So straight darlington for zig, and quasi-complementary to zag.

In your typical 2520 op amp circuit, and it's variants, Gain Blokes, Hybrids etc. there is a BD139 and a BD140 CFP that drives the output.
in between these transistors is usually a pair of resistors, sometimes 1ohm,  often 3r6  and just as often 10r

I need to substitute what I have in my box vs. whats on a schematic.
I'm 99% sure I can go up or down a few ohms and everything will still work fine, but I can't find any criteria to decide.
All the articles I found about the Sziklai pair just ignore these parts, but they must do something.
?
do they set the output impedence? raise the transistor bias? Keep the transistor within operating temperatures
It's usually part of the feedback loop - even if it's not on the op amp itself, it's in the op amp's circuit ???

Thanks
Sleeper

The value of those resistors roughly determine the class A current or how much current the devices draw at idle. The outputs shift to class B for more than a few mA.

As PRR mentioned these need to be small in the context of the output load, but not too small. These are called "emitter degeneration" resistors and hopefully prevent the output stage from thermal run-away (they also promote sharing between parallel output devices). The Vbe or forward voltage drop across a transistor base to emitter will get smaller the hotter the transistor gets. Driving the two output bases with a relatively constant base-to-base voltage means that the hotter the output transistors get, the less the Vbe and more current they draw. The resistors in series with the emitters create a voltage drop as the current increases that stabilized the class A current (hopefully).

I vaguely recall a fancy equation from some old National semi app notes that described how to calculate a resistor value for thermal stability, but to use that formula you need to know the actual thermal resistance of your devices to ambient. So it's easier to just experiment with trial and error.

I'm with PRR start with the largest value you can get away with and still drive your anticipated loads, and get good (low) crossover distortion.  If you have a scope or distortion analyzer feed a low voltage (few hundred mV) sine wave at 20kHz and look for crossover distortion. If you see a step as the sine wave passes through zero drop the resistor values.

You can measure the actual class A current at idle from the voltage drop across the emitter resistors, You shouldn't need more than a few mA of class A current for an opamp. Power amps generally only use 20-30 mA for much larger geometry power devices.

If the power transistors release smoke from getting too hot, replace and make the emitter resistors bigger.  ;D

JR

 

[Sleeper]

Thanks John and PRR, for the informative posts.
THIS is what is so great about this forum. 
I don't know if it's the same in other fields, maybe it is.?

John-I'm not sure if I even really had a direct question, more like general curiosity about the topology.
The actual circuit I'm dealing with is just a garden variety output driver
(as I mentioned, I'm certain it would work with the 5r6 in my box.)
But
it's certainly cool to throw out a question and get answered by a couple of dons
and to top it off, now I even have a method to test my results.
I haven't had a chance to study on this yet, but I will and I hope this helps some other people here too.
Thanks for the info.
Kelly

 

[gridcurrent]

general curiosity about the topology.
The actual circuit I'm dealing with is just a garden variety output driver

The 2 transistors comprise a simple complementary emitter follower, not a complementary feedback pair.
Rod Elliott has a succint description of the cfp:
"The description of a the compound/ Sziklai pair is a configuration of two bipolar transistors of opposite polarities, so will always consist of one NPN and one PNP transistor. The configuration is named after its Hungarian born inventor, George Sziklai. It is also sometimes known as a CFP (complementary feedback pair). The composite device takes the polarity of the driver transistor, so if a compound pair is made with an NPN driver and PNP output device, the overall device behaves like an NPN transistor."

 

[Sleeper]

When I said garden variety, I was mistakenly thinking that those diodes didn't send bias voltage to the junction of the 2 resistors
(whereas they do in the first schemo I posted) but now I've looked at the 2520 schematic again I see this is not the case.
I guess - as far as opamps and output drivers go THIS is the topology dujour.

I'd noticed that the CFP was used differently...
but without that bias voltage (that was there all along) you might see why I was confused.
Oh well, sometimes you go looking after one thing and find another
Thanks

 

[PRR]

> Sziklai pair (also known as a "complementary feedback pair" (CFP) ... I've been looking around I'm not finding an answer.

You have not posted any Sziklai in this thread.

You seem to be mixing three different things.

http://en.wikipedia.org/wiki/Sziklai_pair
http://en.wikipedia.org/wiki/Darlington_transistor
http://en.wikipedia.org/wiki/Push%E2%80%93pull_output

 

[Sleeper]

I think that should be past tense.
"I was mixing three separate things"

Now I'm very clear on the difference... 
Thanks gentlemen.