Member
Joined 2009
Paid Member
the bottom line - can you design and build an amplifier with power Darlington devices for the output that you like listening to ? - I know I can 🙂
How do you overcome thermal matching of the 2Vbe variation in you parallel darlingtons?
A Darlington can never perform better than a well designed discrete two transistor pair.
A well designed two transistor pair can perform better than a Darlington in some types of operation.
I will say again.
A Darlington is an "assembly cost reduction" product.
A Darlington is not a performance enhancing product.
See post5 for a different version of the same message.
A well designed two transistor pair can perform better than a Darlington in some types of operation.
I will say again.
A Darlington is an "assembly cost reduction" product.
A Darlington is not a performance enhancing product.
See post5 for a different version of the same message.
Last edited:
You don't. If you want to build an amplifier that needs 600 watts worth of SOA in each output device you don't use darlingtons.
Or you match them - as Classe Audio does for sure, using 4 pairs per channel in parallel.
As I said over and over, how do you thermal match the 4 darlington? You can static match them on the bench for Vbe. But in real world working situation, each of the 4 in parallel are not in the same temperature exactly. Then you will have 2 X -2.2mV = -4.4mV per deg C instead of -2.2mV/deg C of a discrete BJT. There is no way out of this.
Even if you design a special so the equilibrium temp will settle into the same temp. But dynamically, the temperature will vary and you might kill the darlington before equilibrium temperature can reach.
Also, how much harder to match 2Vbe of darlington compare to one Vbe?
Here is a copy of the AB100 schematic.
This design was manufactured by two companies, and was reliable (or at
least the circuit was). The goal was to provide a relatively simple, easily
manufactured, bullet-proof design. It got some good reviews.
There was no matching of devices. Thermal tracking was provided by
Q8 being in thermal contact with the heat sink.
You can see that thermal issues are largely solved by the use of 1 ohm
emitter resistors. The 1 ohm figure is greater than that considered optimal
(see Self on the subject) but then it was biased higher than his optimal
current figure anyway.
Paralleling alleviates the issue of driver current being lower than we like since
there are as many of them as output devices.
😎
This design was manufactured by two companies, and was reliable (or at
least the circuit was). The goal was to provide a relatively simple, easily
manufactured, bullet-proof design. It got some good reviews.
There was no matching of devices. Thermal tracking was provided by
Q8 being in thermal contact with the heat sink.
You can see that thermal issues are largely solved by the use of 1 ohm
emitter resistors. The 1 ohm figure is greater than that considered optimal
(see Self on the subject) but then it was biased higher than his optimal
current figure anyway.
Paralleling alleviates the issue of driver current being lower than we like since
there are as many of them as output devices.
😎
Attachments
Hi Mr. Pass
I am glad big gun like you join in. Yes, if you use 1ohm emitter resistor, that solve all the problem. But how about the gm doubling that people talk about? That will cause crossover distortion. Can you elaborate a little?
Also, completely off the subject. But I want to ask famous people like you while you are here. I read your schematic, I also read the schematic of my most favorite amp that got me into audiophile.....YBA. Also I read quite a bit of the schematics of famous ones. All the schematics are very simple, not much of the current mirror on the differential pairs, not necessary symmetrical VAS..............But here we are, trying to squeeze in all the fancy circuits. I am really confused. Is there truth about "less is more"?
Thanks
Last edited:
And how would you match the tempco of a discrete darlington, double or tripple? This is really not an issue. A Vbe multiplier also happens to multiply the Vbe tempco. So, 4 Vbe per pair just as if it was a discrete design.
Having multiple output pairs (discrete!) together with multiple driver pairs, while not common, is by no means unheard of.
The trick is, increase the emitter resistors. There will be some additional imbalance due to the emitter resistors of the drivers terminating into the emitters of the output devices rather than the output line, but the device itself also has higher transconductance, which the increased emitter resistor tames quite well - now it works on both Vbes not only that of the output.
Yes, and it also uses the emitter resistor value together with the 3 diode clamp on the input to limit current - which may well be another reason why it's a higher value. A long time ago I used MJ11015/16 darlingtons in a multiple pair situation and 0R47 took care of the current sharing issue just fine.
Big difference. You use ONE driver transistor to drive 4 power BJT. You don't have to match the driver as there's only one. In darlington, it is as if you have 4 separate drivers.
Read Oliver's about minimize crossover distortion when re' of the power BJT equal to RE. You want to use value of RE to drop only 26mV at idle current to prevent gm doubling. Of cause, if you ignore the crossover distortion and set the RE=1ohm or higher, you have much less of a matching problem.
Here's a classic app note from Motorola (70s vintage) using their power Darlingtons in audio amplifier applications.
http://www.w5dor.com/AppNotes/AN483B-15-to-60-Watt-Audio-Amps.pdf
I ended up buliding a version of the differential input circuit using discrete Darlingtons, as I couldn't get hold of the Motorola integrated devices. If I remember correctly, I used 2N3716 and 2N3789 along with 2N3439 and 2N5416. I also used leds for output stage bias. It worked and didn't blow up, though I would do things quite differently these days, ~40 years down the line.
http://www.w5dor.com/AppNotes/AN483B-15-to-60-Watt-Audio-Amps.pdf
I ended up buliding a version of the differential input circuit using discrete Darlingtons, as I couldn't get hold of the Motorola integrated devices. If I remember correctly, I used 2N3716 and 2N3789 along with 2N3439 and 2N5416. I also used leds for output stage bias. It worked and didn't blow up, though I would do things quite differently these days, ~40 years down the line.
I scan through the apt notes quickly. It's using only one pair of darlingtons. It will work perfectly. There is no need to thermal match like if you have more than one pair in parallel.
why don't you build one and see for yourself......
there is a sea of difference between reading about it and getting
experience firsthand....
i built a super leach amp using the MJ11032/MJ11033 darlingtons in
the output stage, at slightly lower rails of bout 70 volts,
the amp performed superbly to my liking...
that was almost 20 years ago....last i heard from the owner, the amp still lives....
That's where the difference in engineering. You want to first look at the potential problems. Two question I raised and all I heard is "it works".
FYI, I am laying out the OPS and next is IPS and VAS. It takes time. But that doesn't stop me in asking. Read Cordell's book and Oliver's optimization.
Obviously people are interested in this thread because there are much better darlingtons, I showed one that is 50MHz. Still the problem is the thermal compensation.
no need to reinvent the wheel every-time you want to build a car...
there is a great body of knowledge from those who came before us...
you just have to use the search button....
you still worry about thermal compensation when the man himself said it is not so mych an issue in practical amps as it is in your head....
listen to Nelson Pass, he has built thousands upon thousand amps, he knows whereof he speaks....
if you worry too much about issues in your mind, you will never get to build anything at all...
i urge you to look up on Ostripper's threads and that of Apexudio, these two gents are fine examples to learn from....
there are already a handful of armchair theoreticians here, do not be another one, please build something...
there is a great body of knowledge from those who came before us...
you just have to use the search button....
you still worry about thermal compensation when the man himself said it is not so mych an issue in practical amps as it is in your head....
listen to Nelson Pass, he has built thousands upon thousand amps, he knows whereof he speaks....
if you worry too much about issues in your mind, you will never get to build anything at all...
i urge you to look up on Ostripper's threads and that of Apexudio, these two gents are fine examples to learn from....
there are already a handful of armchair theoreticians here, do not be another one, please build something...
- Status
- Not open for further replies.