I don't see how this is a problem? It seems to me you are mixing up 3 things, thermal compensation, current sharing and crossover region problems. These are quite separate issues that converge only due to an overlap in requirements for Re.
Current sharing especially is not a large issue as it's easier to get proper current sharing from high gm devices given the same vale of Re, and a darlington, even in it's most unfavorable operating point for beta, has sufficient gm for this - practically by definition, as there is hardly a single BJT that does not. In fact, by forcing degeneration through Re for both the driver and the power transistor, you may end up with less of a current sharing problem than in traditional discrete darlington output stages where one set of pre+drivers drive more than one set of power transistors.
Regarding gm doubling, what you get is a reduction of the voltage drop on Re if they are increased over Re', which may be required for better stability, but you tell it like it's 'ignoring crossover'. Who is ignoring it? For every Re there is an 'ideal' bias current, which is only ideal in the sense you get the best result for that particular setup, but crossover problems simply CANNOT be avoided, regardless or Re chosen. I would advise a closer look and perhaps some simulation to explore what happens. The internals of integrated darlingtons are often well known, sufficiently that an educated guess can be made how to simulate one using discrete components. What you might find is that while there are differences, with properly chosen conditions, they are not dramatic, as in orders of magnitude (I'm not even going to go into the topic of simulating crossover only at one RESISTANCE on the output when this is never the case in real life...).
What this boils down to is, of course it works. If you have any decent skilll as a designer you can make thigs work. The real question is how mych better can you get a fully discrete design for the same money, honestly. Current audio darlingtons are really made to reduce component count and price. It doesn't take a genius to understand that discrete offers more freedom in design, but there is no point in being able to use ANYTHING discrete and then compare it to a darlington solution cut down to a price/performance point.
One might better look at what actual trade-offs an integrated darlington forces on the designer and how to optimize the design.
Also, there is a subtle irony in the nameing of the thread as we are actually comparing two topologies of darlington transtor outputs, one discrete and optimized, and one integrated and not optimized. It's really down to 'why do not the manufacturers offer an integrated part optimized for audio'.
That being said, comparing (again) 'any' discrete combination with a 'average' integrated darlington part is pointless. Especially for older darlingtons. These are mostly devices with rather poor SOA given their maximum ratings, which have integrated parts not entirely suited for audio use.
A good exercise would be making an amp with one of the Sanken darlingtons, say, simulation it with the discrete equivalents and seeing how much of a compromise there is when the given integrated parts are used instead.
Designing audio circuits is all about compromising with trade-offs. It is about being able to use the components that you have to work with. The Darlington 'scrap amp'😀 I made was useful and it alleviated my boredom for a while at the time. It was useful in that I was able to experiment with some different types of circuit/speaker protection schemes as well. This is very important, I believe, in making a real amplifier circuit but not so much in regards to simulations. To be honest, I did not expect the really good results I got from using such components as they were never intended to be used in this manner, but it was still fun to build it. Is not building stuff the purpose of DIY, eh?🙂Of course sometimes you have to fix it.......yesterday I was playing with that Scrap Amp and accidentally shorted the base of the output transistor to the rail and it blew up.
! Opps, I blew out the TIP100 of one channel, but not anything else so it was back to the scrap heap of old and busted PCBs to search for another. One was found and used to replace the shorted one and there was no problem. Still listening.😎BTW forgot to mention a simple point regardin thermal stability. Use a smaller darlington in the Vbe multiplier 😛 - of course on eneeds to take into account the internal resistors in the darlington pack when calculating sensible values for the Vbe multiplier external resistors.
Most Darlingtons are built on older processes and have poor second breakdown characteristics. That is why I would tend not to use them. Modern single transistors (MJ21193/4; 2sC5200 etc) have better second breakdown.
Crossover distortion may be higher due to the built-in base resistor across the output device.
But you can always reduce crossover distortion by moving the Miller capacitor to the output rail, and adding sprog stopper capacitors on the VAS if necessary (which is likely).
To make this work best you also need input degen as this sets the upper frequency limit, and then increase the gain in the stages following the input!
Oh, that looks like a modified Blameless!
John
Crossover distortion may be higher due to the built-in base resistor across the output device.
But you can always reduce crossover distortion by moving the Miller capacitor to the output rail, and adding sprog stopper capacitors on the VAS if necessary (which is likely).
To make this work best you also need input degen as this sets the upper frequency limit, and then increase the gain in the stages following the input!
Oh, that looks like a modified Blameless!
John
- Status
- Not open for further replies.