I'd take a vas as a transimpedance stage so I'd like it to see a constant load resistance regardless its output voltage. I'd think the 2sc5200 combined with a driver as shown in your chart would do a far better job than that Darlington could.
Alan: For low distortion you want hfe to be constant. That means if the output stage is supplying a sinusoidal current to the load, it will also draw a sinusoidal current from the VAS. Variations in hfe cause nonlinear loading of the VAS.
You also want hfe to be high, since any load on the VAS, linear or nonlinear, reduces the loop gain and increases the distortion of the VAS itself. Fans of output triples argue that if hfe is high enough, the current drawn by the output stage from the VAS can just be neglected altogether. They don't care so much if hfe varies, as long as it's always huge.
This starts to get interesting once you take into account the various snubbers used to stabilise triple output stages. Hopefully if these are well designed, they function as local feedback and trade magnitude of gain for linearity of it.
You also want hfe to be high, since any load on the VAS, linear or nonlinear, reduces the loop gain and increases the distortion of the VAS itself. Fans of output triples argue that if hfe is high enough, the current drawn by the output stage from the VAS can just be neglected altogether. They don't care so much if hfe varies, as long as it's always huge.
This starts to get interesting once you take into account the various snubbers used to stabilise triple output stages. Hopefully if these are well designed, they function as local feedback and trade magnitude of gain for linearity of it.
Thanks
I was thinking about that also. When current is low, who cares!!! Using darlington as the big power transistor is out because if you parallel darlingtons, there are two transistors mismatch between the dalingtons. The only place to use darlington is the driver transistor. Then you have those two transistor matching. The only ones that have matching problem are the big power transistors.
Run the VAS higher current, this will swamp out the variation of the hfe.
BTW, any tips on preventing 3EF from oscillating? I know base stop resistor of 4.7ohm on every power transistor right at the base pin is a must. What else?
Thanks
I try to dissuade Builders from using a Darlington as the switch.
BTW, the loss through a Darlington is a lot worse than double.
Saturated transistor typically has a Vce of 0.03V to 0.08V
A Darlington typically has a Vce of 0.7V to 1V when turned full on.
That loss times the passing current is wasted power and excessive heating.
4r7 is probably too high.
A useful range to investigate to find the optimum would be 1r0 to 10r for the output devices.
Around 1r5 to 2r7 are probably typical of optimum for an output device, when some negative impedance needs canceling.
Drivers may need around ten times this stopper resistance.
Pre-drivers may need around fifty times this stopper resistance.
A Darlington certainly makes a poor switch. But with so many good and cheap MOSFETs and IGBTs around, there is no reason to be using any sort of BJT in a switching power application any more. They are still used in some low-end PC power supplies and lamp ballasts, in a self-resonant circuit that looks simple but is hard to design for reliable operation.
Yet we still see many Builders adopting them in ignorance, despite attempts to enlighten.
One audio circuit which is ideal for Darlington power transistors (and for which I have used Darlingtons myself) is: a supply rail splitter. Also called a Virtual Earth generator.
This is a circuit which turns a floating 30VDC power supply, into two ground-referenced power supply outputs: +15V, -15V. For preamp and line-level applications, the load current is less than an ampere per rail.
Why is a Darlington ideal for this application? Because it is guaranteed to operate at high Vce (namely, 15V), far far away from saturation. Just exactly where you'd want to operate a Darlington. The two Vbe drops are not troublesome either, since the emitters are at ~ ground. The bases are two Vbes above and below ground (thus giving enormous supply headroom), meaning that a low cost, non-rail-to-rail opamp can be used.
I chose the TIP142 / TIP147 complementary Darlingtons, and they worked beautifully. Very low output impedance, very good tracking between +15V and -15V, and very good transient response to a load current step-change. Very low cost too!
This is a circuit which turns a floating 30VDC power supply, into two ground-referenced power supply outputs: +15V, -15V. For preamp and line-level applications, the load current is less than an ampere per rail.
Why is a Darlington ideal for this application? Because it is guaranteed to operate at high Vce (namely, 15V), far far away from saturation. Just exactly where you'd want to operate a Darlington. The two Vbe drops are not troublesome either, since the emitters are at ~ ground. The bases are two Vbes above and below ground (thus giving enormous supply headroom), meaning that a low cost, non-rail-to-rail opamp can be used.
I chose the TIP142 / TIP147 complementary Darlingtons, and they worked beautifully. Very low output impedance, very good tracking between +15V and -15V, and very good transient response to a load current step-change. Very low cost too!
You guys keep talking about switch using darlington. I just want to clarify you guys are not talking about the output transistor anymore. This is only for switching, nothing to do with my question for power amp.
BTW, nobody comment on the idea of using darlington for the driver transistor to drive the power BJT. I calculated the current variation of the driver transistor should be no more than 100mA. Some more if you consider charging and discharging the input capacitance of the power transistors. But it's not going to be that high.
So the hfe variation in say 300mA range is not high at all.
BTW, nobody comment on the idea of using darlington for the driver transistor to drive the power BJT. I calculated the current variation of the driver transistor should be no more than 100mA. Some more if you consider charging and discharging the input capacitance of the power transistors. But it's not going to be that high.
So the hfe variation in say 300mA range is not high at all.
Class AB output devices aren't switching, hmmm what are they doing then.
Last edited:
You only talked about lost of 1.2V across the darlington instead of 0.5V at the extreme end. That does not seems to be a big problem. So you loss 0.7V swing.
I am more concern with the variation of hfe. At the driver, the variation is more like from say from 50mA to 200mA. It is a much smaller change than the output transistor that swing from 50mA to 3A. So the hfe change should be very small.
That's only an opinion, how about getting to the reasoning? I look at the 0.7 volt drop disadvantage, I also look at the variation of hfe with current.
I agree that darlington is not good as power transistor in parallel situation because of the variation of 2Vbe that cannot be compensated.
But now I am talking about using it as pre-driver that you don't need to parallel up. Also the current variation is much much smaller than when used as power transistor. So the hfe variation is very small.
So what other disadvantage in using as pre-driver?
Ah, using darlington as power output transistor in Class AB will lead to turning off BOTH transistor inside the darlington when going into Class B. That will cause double the crossover distortion.
Is that what you mean? That is a very good point I did not think about!!!
I already gave up the idea of using darlington as output transistor in parallel as you have 2 transistors inside each darlington that mismatch with other darlingtons in parallel. That make the thermal and current sharing situation much worst.
Now I am contemplate using darlington for pre-driver stage. It should never be turn off so no cross over distortion to talk about. Also, the current variation is much smaller than when using in output power transistor, so hfe stays quite constant at range of say 50mA to 200mA.
What do you think?
Thanks, that's a very good point for output transistor. I wonder why Nelson Pass used that as described in post #7, was that a class A only?
Last edited:
NO
think about multiple sine waves , then what about audio amps reproducing square waves. looks like the outputs are switching after all.
think about bandwidth and slew rate?
you can gather a lot of new learning by reading a chip-amps data sheet and looking at their test set-ups. start measuring a commercial discrete audio amp using the same set-up. PSRR etc.
think about multiple sine waves , then what about audio amps reproducing square waves. looks like the outputs are switching after all.
think about bandwidth and slew rate?
you can gather a lot of new learning by reading a chip-amps data sheet and looking at their test set-ups. start measuring a commercial discrete audio amp using the same set-up. PSRR etc.
Hi All,
From my perspective, this is a somewhat pointless discussion.
Starting from the thread name - some people do use Darlington power transistors in their amps.
Depends on particular transistors. And hfe or Vbe drop are not the biggest problems here. Some of them are just too slow. Although, some of them are good enough.
Example - MJ11032/11033 pair. Rather good for using instead of EF2. I ran a simulation. They just require certain amendments in compensation.
Now, imagine - you'd like to use more than one pair of them. Most likely, with EF2, you would use one driver and 2-5 output transistors. And you have still got a lot of flexibility in combining, for example bjt-bjt, bjt-fet, fet-bjt, fet-fet (EF, or, say, CFP, if you want). As soon as you decently match the drivers, the outputs matching is not required. With Darlington ones, you will have as many internal "drivers", as many devices you use. Overall Hfe varies in the range of 1K - 18K. So... you will most likely want to match them. I mean - all of them.
Also, with EF2, you can keep the drivers in pure class A, if you want, for example, loading then with CCSs (although in most of the designs, they switch off together with the output ones in class B). With Darlingtons you can't.
So.. it's a matter of reasoning.
Can you build an amp with Darlingtons?
Yes, you can. With some of them.
But why limiting yourself, if you can use the separate drivers and have a lot of freedom in the way you design your OPS?
Cheers,
Valery
From my perspective, this is a somewhat pointless discussion.
Starting from the thread name - some people do use Darlington power transistors in their amps.
Depends on particular transistors. And hfe or Vbe drop are not the biggest problems here. Some of them are just too slow. Although, some of them are good enough.
Example - MJ11032/11033 pair. Rather good for using instead of EF2. I ran a simulation. They just require certain amendments in compensation.
Now, imagine - you'd like to use more than one pair of them. Most likely, with EF2, you would use one driver and 2-5 output transistors. And you have still got a lot of flexibility in combining, for example bjt-bjt, bjt-fet, fet-bjt, fet-fet (EF, or, say, CFP, if you want). As soon as you decently match the drivers, the outputs matching is not required. With Darlington ones, you will have as many internal "drivers", as many devices you use. Overall Hfe varies in the range of 1K - 18K. So... you will most likely want to match them. I mean - all of them.
Also, with EF2, you can keep the drivers in pure class A, if you want, for example, loading then with CCSs (although in most of the designs, they switch off together with the output ones in class B). With Darlingtons you can't.
So.. it's a matter of reasoning.
Can you build an amp with Darlingtons?
Yes, you can. With some of them.
But why limiting yourself, if you can use the separate drivers and have a lot of freedom in the way you design your OPS?
Cheers,
Valery
I think by "switching" AndrewT must have meant operating as a saturated switch in some other application than a Class-AB power amp. There is no Class-AB output stage topology in use in hifi that can saturate the output devices and get the output really close to the rail. The double EF is exactly the same as an integrated Darlington in this respect, they both need 2 Vbe of "headroom". The Sziklai pair is a bit better, but not nearly as good as you'd think.
The switching that happens in the crossover region of a Class-AB amp is not saturated switching. Every high quality amp design I've seen tries pretty hard to avoid saturating anything, as it prolongs recovery from clipping. Bob Cordell goes as far as advocating Baker clamps that hang everything up 3-5V away from the supply rails.
The switching that happens in the crossover region of a Class-AB amp is not saturated switching. Every high quality amp design I've seen tries pretty hard to avoid saturating anything, as it prolongs recovery from clipping. Bob Cordell goes as far as advocating Baker clamps that hang everything up 3-5V away from the supply rails.
- Status
- Not open for further replies.