Darlington and Triple darlington

[lumanauw]

With EF output stage, usually we need another EF driver stage--this is Darlington. With 2stage driver--this is triple darlington.

Once I asked a "guru" here why he don't like to use drivers for EF output stage. Or maximal he only use darlington, but not going to triple darlington.
Other designer here likes triple darlington or even quadruple darlington, to have big current gain in EF output stage.

The answer is very short, but until now I dont understand it clearly. He just said "It add another component in the signal path"

Imagine we have a NPN transistor. Signal is put to Base, output is taken in Emitor. This is a voltage follower. In theory it has 100% feedback (no voltage gain, but with current gain for output stage)

What kind of distortion that a simple EF can have? Does this simple "voltage follower" adds phase shift or what, that makes it bad, that makes the "guru" even dont want to use it even for EF driver?

If such a voltage follower is considered lowering the audio quality reproduction, what about a cascode transistor (used to split dissipation). Does this also included in "not-wanted signal degenerator" ?

[sajti]

It's interesting question. I use triple darlington in my desing. EF has distortion, Mr. Self made some measurings. However the distortion of the emitter follower highly depended by the load. So I guess, that the predriver has not too big distortion.
I found that the double darlington needs higher current VAS to drive it. Say 8-10mA for higher power amplifier. For this VAS we have to use power transistors, which have high capacitance inside. To drive it there is necessary to use emitter follower between the input stage and the VAS.
With triple darlington lower current VAS is enough, so it's easy to drive from the output of the input stage. So finally the number of the stages is same, only the location on the emitter follower is the question.

sajti

[Mr Evil]

There are compromises involved going from single transistor to Darlington to triple. The major benefit is increased input impedance of the output stage, reducing current requirements from the previous stage, and reduced output impedance which reduces load-dependance. This can quite dramatically improve linearity. On the downside, the extra transistors mean an added delay, adversely affecting phase shift at higher frequencies, and the extra transistor introduces some distortion of its own. There are other minor benefits and drawbacks too.

While reducing the number of components is a worthwhile goal, it does not mean that extra components automatically reduce sound quality. If that were true, then no one would ever build anything other than single transistor amps. You need to understand the pros and cons of adding extra transistors to a specific design, be it triple darllington or cascode. In some instances more will be better, in others it will be worse.

Personally I prefer to use MOSFETs for the output stage, rendering the entire subject of triple Darlingtons moot..

[Nelson Pass]

The choice between the two depends on quite a few things,
among them:

Are you running BJT's?

How much current are you trying to deliver to the load?

How much current can your front end source?

How high is the beta of your output device at high current?

How sophisticated is your frequency compensation system?

I've seen very nice versions of both, but I note from experience
that it is more difficult to frequency compensate a triple, which
can parasitically oscillate on its own, independently of the rest
of the system.

Single darlingtons are usually better in situations like the A40
Class A amplifier at 40 watts. Triples start being useful at
200 watts and above, or if you are planning on delivering more
than 10 amps, in other words a Krell.

[lumanauw]

Hi, Mr.Evil,

Personally I prefer to use MOSFETs for the output stage, rendering the entire subject of triple Darlingtons moot..

Yes, you are right. Mosfet can "block" what happens with the load from entering the VAS / signal stage. Mosfets are good for classA. This time I'm making class AB where bipolars are more suitable, so I have to deal with this question

Hi, Mr.Pass,

Are you running BJT's?

Yes, I'm making AB amp using pararel bipolar for final stage

How much current are you trying to deliver to the load?

The rail is +/-70V, and I wanted to stable up to 2ohm. That makes about 35A max?

How much current can your front end source?

I havent design completely, I will be using single differential. But from experimenting, If I use ordinary 3 stage power amp, it will be 2mA ccs (1mA to each leg) cause this configuration seems sound good. Haven't design the VAS yet. Will it need EF after the differential voltage drop?
Other option is to use Folded Cascode.

How high is the beta of your output device at high current?

At high current? The transistor is C5200+its PNP. Is this high current beta the same as the beta shown in datasheet, or we have to test it manually?

How sophisticated is your frequency compensation system?

Since I haven't design anything yet, I dont know. The up limit is 3 stages power amp, with option of folded cascode.

What about this output stage that I attach? It uses mosfet for the driver, connected to ouput via 22ohm resistor, and it performs darlington. What's the pro/con of using mosfet in this mosfet-bipolar darlington system compared to all bipolars triple or quadruple darlington?

[sss]

i think the main problem with tripple output stage is stabilization , the gain (beta) is so high so it starts to oscilate .
its the same as making an amp with five gain stages
anyways , i think the compound config is much better then darlington.

[sajti]

Originally posted by sss
i think the main problem with tripple output stage is stabilization , the gain (beta) is so high so it starts to oscilate .
its the same as making an amp with five gain stages
anyways , i think the compound config is much better then darlington.

As I used triple darlington many times, I can comment, that it's not complicated to avoid oscillation.
Just apply some base stopper resistors. I use 1-4.7ohms for the output devices (it good too if You have more output device connected parallel), 10-47ohms for the drivers, and 100-470ohms for the predrivers. I never had problem with the oscillation.
Another good thing for the triple darlington, that it provide very low output impedance without large amount of feedback. And also very easy to provide constant output impedance over the whole frequency range

sajti

[Circlotron]

Originally posted by lumanauw
The rail is +/-70V, and I wanted to stable up to 2ohm. That makes about 35A max?

If the load is purely inductive (worst case, but you have to allow for these kind of things - a more reasonable real-world figure is 45 degrees) the load current will lag the voltage by 90 degrees so at the zero crossing of the output voltage at full power you will have 70/2=35 amps *and* 70 volts across the output transistors *at the same time*. 30x75=2250 watts instantaneous dissipation.

You are going to need =lots= of transistors.

[sajti]

Originally posted by Circlotron

If the load is purely inductive (worst case, but you have to allow for these kind of things - a more reasonable real-world figure is 45 degrees) the load current will lag the voltage by 90 degrees so at the zero crossing of the output voltage at full power you will have 70/2=35 amps *and* 70 volts across the output transistors *at the same time*. 30x75=2250 watts instantaneous dissipation.

You are going to need =lots= of transistors.

Yes. He needs at least 6pairs of MJL21193/94, or 8pairs of 2SA1943 / 2SC5200.

sajti

[Circlotron]

Certainly. At 70V one of those transistors can only handle 3 amps.
http://www.onsemi.com/pub/Collateral/MJL21193-D.PDF