Replacing a standard output transistor with a Darlington

No, I don't want to do this, but on a dead Yamaha receiver, I was asked about an appropriate replacement for blown outputs. I was prepared to make suggestions when he came back and said he'd found some Darlington outputs, popped them in, and now he's experiancing sonic bliss.:rolleyes:

My comment was that he's got no bias current, and that, in general, replacing a standard output transistor with a Darlington was a bad move. Others are trying to tell me that it's a wonderful idea, and besides the lack of bias current I can't think of a lot in-depth reasons why you shouldn't do this (I personally think the lack of an ability to bias the transistor is reason enough).

So, why aren't Darlington's used more for output transistors? My own thoughts are that, if crossover issues are a point of contention with a Vbe of 0.6V, that they are much more so for a Vbe of 1.2V and more. But beyond that, I'm out of steam. Difficulty with thermal issues and runaway? Other instabilities? I'd love to hear some additional technical thoughts on the issue...
 
I dunno, I've always kinda liked darlingtons. Motorola had a good app note on audio power amps with 'em. Back in college I had a big (for the time) Integral System power amp, 100Wrms/ch, that used just two (somewhat stressed) darlingtons per amp. Sure, the bias will be off if somebody just pops 'em in, but feedback cures a lot of ills. They can be slow and can have trouble turning off, but used properly it's not a big issue. I'm not sure why they're not more popular. OTOH, they're no guarantee of sonic bliss.
 

linuxguru

Member
2007-09-21 6:28 am
If you just replace one O/P transistor with a darlington, not only will the bias be wrong, but the operating point will also be unsymmetric class-B and the DC offset will be way off.

However, you can use darlingtons in well-designed O/P stages, and they can sound very good indeed. For instance, some entry-level Sony receivers (DE-485?) use Sanken darlingtons (2sb1620/2sd2488) specially labeled for Sony , and those stages sound exceptionally detailed - it is probably helped along by an fT of 70 MHz.

I have used OnSemi 2N6052/59 as well as Philips TIP142/147, and they are quite acceptable for Class A/AB/B use - however, they may need base stoppers to tame instability. ST BDX33C/34C are also worth taking a look at - inexpensive and reasonable SOA.
 

ilimzn

Member
2005-02-11 1:25 pm
Zagreb
In general, what happens if you replace a discrete transistor with a darlington, is this:
- bias must be adjusted to compensate for more Vbe drops
- Another stage is added which may reduce the phase margin and induce oscillations - however, this is very dependant on the actual parts chosen. Along with Ft, the things to look at are switching characteristics because the built-in resistors between bases and emitters may not be suitable for the particular implementation.
- More current gain is usually available at audio frequencies, which increases the input impedance of the output stage, maiing the VAS work into a lighter load. Absolutely speaking, however, gain variations with output current and frequency tend to increase. In the end, the improvement may not occur where you would have a seizable benefit.

So, in general it is possible. Usually, if one would do this sort of thing, it would be replacing the output and driver transistors with integrated darlingtons (which is what they were really designed for).

That being said, there are probably hundreds of darlington types but really very few are usefull for a properly executed design. In most cases there are B-E shunting resistors which may not at all be optimal for a given implementation, and, such transistors are usually used for switching applications, which means that even if you find reasonably quick ones, they have a vanishing SOA at voltages and currents you would expect just looking at the Icmax and Vcemax specs.

There are exceptions, though.
Sanken makes a number of power darlingtons based on either tripple diffused or LAPT processes, specifically intended for audio use.
Old school 2N6052/59, TIP142/147, BDX33/34, BDV64/65 have shunt resistors as opposed to the Sanken parts, but also usable SOA for smaller amps.
MJ11015/16 look good on the surface, with 120V and 30A capability, but the SOA is actually useful only for much lower power outputs (compare with sieblings like the MJ15003/4).
Most of these see a rather huge variation in current gain depending on collector current, but should work fine in the usual range one would find in a medium power (50W) amp regarding bias current and load current However, comparing it with a combo of a good driver and beta sustained or LAPT power transistor for audio, it will look quite dissapointing.
I have used both the BDV64/65 and MJ11015/16 in class AB amps, taking care to provide good SOA limiters, but I have found them to be very good in low power (20-25W) class A amps where the power rails are within usable SOA limits without doing anything special, and the high idle current keeps the gain and Ft nice and high.
 

d3imlay

Member
2004-09-09 11:34 pm
Ohio
The early Phase Linear 700B's used a Darlington. They later switched to a standard bipolar. Seems as though I read that load sharing was better with lower gain devices. I've replaced the Darlington's in 700b's with MJ15024 without any troubles. I don't remember if I had to mod the bias circuit though.
 

Donk

Member
2007-09-28 10:21 am
d3imlay said:
The early Phase Linear 700B's used a Darlington. They later switched to a standard bipolar. Seems as though I read that load sharing was better with lower gain devices. I've replaced the Darlington's in 700b's with MJ15024 without any troubles. I don't remember if I had to mod the bias circuit though.


Are you sure?
 
Any amplifier using the classic two cascoded emitter follower output stage could be modified to use darlingtons resulting in a triple emitter-follower.

Bias, thermal compensation and frequency compensation would have to be redesigned, though, and not everybody can do this. There is not such a thing as a successful direct replacement, that would result in crossover distortion, oscillation and some blown darlingtons in the long term.

The main drawback of monolithic darlingtons is that the inner base is not accessible and thus the outer (driver) transistor can't be biased properly (class A) and the inner base charge can't be properly depleted when fast current zero-crossing events are demanded (like in trebble with no bass). Monolithic darlingtouns should have been conceived with a fourth terminal...
 

wg_ski

Member
2007-10-10 5:21 pm
Every PL400 or 700 of any vintage I've ever seen used 40327/4003 predrivers and RCA house #66546 drivers. A wide array of outputs, depending on whether they were original or what they were replaced with. I beleive the original spec callled for RCA house # 1B05. In any case, it's 3 stages of current gain *without* resorting to using darlingtons. I'd hate to try and make that stable with 4 stages, especially on the negative rail!