Use of integrated darlington transistors (i.e. driver and output transistor on one die) for the output stage is frownded upon because:
- most integrated darlingtons fail to impress regarding bandwidth and beta linearity when compared to the pairing of a first rate power transistor and a first rate driver (if there are exceptions, I am not aware of them)
- one is robbed of the possibility to use a shared emitter resistor in the driver stage which speed up the turn-off of the output transistors
We also know that a triple EF (aka triple darlington) has lowest distortion because it loads the VAS less and has sufficient current gain throughout the operating regime.
So how about using an integrated darlington in TO126 or TO220 case for the driver stage, thus making a triple EF? I don't remember ever seeing an example. Are there obvious downsides? Of course, there cannot be a shared emitter resistor for the predrivers now, but then the drivers carry less current than the output devices and are hopefully faster.
- most integrated darlingtons fail to impress regarding bandwidth and beta linearity when compared to the pairing of a first rate power transistor and a first rate driver (if there are exceptions, I am not aware of them)
- one is robbed of the possibility to use a shared emitter resistor in the driver stage which speed up the turn-off of the output transistors
We also know that a triple EF (aka triple darlington) has lowest distortion because it loads the VAS less and has sufficient current gain throughout the operating regime.
So how about using an integrated darlington in TO126 or TO220 case for the driver stage, thus making a triple EF? I don't remember ever seeing an example. Are there obvious downsides? Of course, there cannot be a shared emitter resistor for the predrivers now, but then the drivers carry less current than the output devices and are hopefully faster.
Darlington's don't allow for a speed-up capacitor, often important to HF distortion performance, as they can greatly reduce switch-off distortion. With more modern fast drivers and output devices this is less vital, but its such an easy win for an output stage.
A main weakness of the CFP output topology is that there is no (easy) way to speedup switch-off (thus fast transistors are needed). With an EF output topology its makes very good sense to use a speedup cap and get the best out of the topology and place less reliance on the device specifications.
A main weakness of the CFP output topology is that there is no (easy) way to speedup switch-off (thus fast transistors are needed). With an EF output topology its makes very good sense to use a speedup cap and get the best out of the topology and place less reliance on the device specifications.
The biggest weakness of darlingtons is being stuck with the internal base-emitter resistor connections (and values). When using an EF3, you usually don’t even want an Rbe on the predriver, even if you could live with the one on the driver. Many moons ago there used to be some types that didn’t have the resistors, but none of them were good for more than 60 volts. What would be nice are 4-terminal darlingtons without the resistors, but that’s just too much to ask.
@ Mark:
My proposal is to use an integrated Darlington for the first two stages of an EF3. I would still get to use a common emitter resistor between the emitters of the Darlingtons (and hence the bases of the output devices). I could even parallel a speedup cap there though it is pretty rare in that location, the cap being more common across the VBE multiplyer.
You can always use a pretty low value resitor between each base and rail in a CFP. The main issue there is the Miller effect in the output transistor.
@ wg:
Yes, I have wished for four terminal Darlingtons before! Looking at this Sanken model
https://www.semicon.sanken-ele.co.jp/sk_content/2sb1647_ds_en.pdf
the 70 R resistor would not hurt that much in the first stage of an EF3, would it?
My gripe with that model (and many others) is the lousy ft curve (you would do better with a 30 - 40 MHz output transistor and a ~150 MHz driver) but more so the very high Miller capacitance compared to said combination.
I haven't found any 120 V+, 1-2 A, 150 MHz + Darlingtons to use in the first/second stage, though.
Well, there's this:
https://www.onsemi.com/pdf/datasheet/tip110-d.pdf -> TIP 112/117
- hfe (Ic) curves from hell
- min hfe at 1 MHz of 25 (i.e. roundabout 25 MHz ft)
- 200 pF Miller capacitance in the PNP
-> normalized to max Ic, this part is much worse than the Sanken
and these are even worse:
https://www.onsemi.com/pdf/datasheet/mjf6388-d.pdf
Seems there just aren't any fast medium power Darlingtons. Lack of demand?
My proposal is to use an integrated Darlington for the first two stages of an EF3. I would still get to use a common emitter resistor between the emitters of the Darlingtons (and hence the bases of the output devices). I could even parallel a speedup cap there though it is pretty rare in that location, the cap being more common across the VBE multiplyer.
You can always use a pretty low value resitor between each base and rail in a CFP. The main issue there is the Miller effect in the output transistor.
@ wg:
Yes, I have wished for four terminal Darlingtons before! Looking at this Sanken model
https://www.semicon.sanken-ele.co.jp/sk_content/2sb1647_ds_en.pdf
the 70 R resistor would not hurt that much in the first stage of an EF3, would it?
My gripe with that model (and many others) is the lousy ft curve (you would do better with a 30 - 40 MHz output transistor and a ~150 MHz driver) but more so the very high Miller capacitance compared to said combination.
I haven't found any 120 V+, 1-2 A, 150 MHz + Darlingtons to use in the first/second stage, though.
Well, there's this:
https://www.onsemi.com/pdf/datasheet/tip110-d.pdf -> TIP 112/117
- hfe (Ic) curves from hell
- min hfe at 1 MHz of 25 (i.e. roundabout 25 MHz ft)
- 200 pF Miller capacitance in the PNP
-> normalized to max Ic, this part is much worse than the Sanken
and these are even worse:
https://www.onsemi.com/pdf/datasheet/mjf6388-d.pdf
Seems there just aren't any fast medium power Darlingtons. Lack of demand?
Last edited:
Many of the Sanken darlingtons are specifically targeted to audio applications where reduced parts count is desired. The TIP112/7 aren’t exactly tailor made for high performance audio amps.
EF3’s aren’t as common as they used to be. It was one way of getting around the horrible beta droop that early epitaxial base output transistors had - throw enough composite beta at the problem and linearity of it matters less. Today’s outputs and drivers are a lot better for hFE linearity, so most mass market receivers, even “good” ones don’t resort to using triples anymore. When you do see them, the cost of an extra TO-126 pair per channel isn’t as much of an issue - you see them when you’re already driving 4 pairs of outputs (or more) and want to drive 2 ohm loads. Stuff that people are willing to pay for. That’s not the majority of applications. If something doesn’t sell millions of pieces a year, it gets dropped. That’s what ended up happening to the old D40 through D43 series from GE/RCA. They went the way of the dinosaur. Nice darlingtons, plenty fast. But not enough voltage to make 100 watt HT receivers out of, and you don’t need an EF3 on a +/-30 volt supply.
EF3’s aren’t as common as they used to be. It was one way of getting around the horrible beta droop that early epitaxial base output transistors had - throw enough composite beta at the problem and linearity of it matters less. Today’s outputs and drivers are a lot better for hFE linearity, so most mass market receivers, even “good” ones don’t resort to using triples anymore. When you do see them, the cost of an extra TO-126 pair per channel isn’t as much of an issue - you see them when you’re already driving 4 pairs of outputs (or more) and want to drive 2 ohm loads. Stuff that people are willing to pay for. That’s not the majority of applications. If something doesn’t sell millions of pieces a year, it gets dropped. That’s what ended up happening to the old D40 through D43 series from GE/RCA. They went the way of the dinosaur. Nice darlingtons, plenty fast. But not enough voltage to make 100 watt HT receivers out of, and you don’t need an EF3 on a +/-30 volt supply.
Darlington OPS was popular in the past. Here is one way from the early 80's to drive Darlington EF stage, using the available types with internal Rbe. Incidentally, this approach will also work well with vertical MOS as they share a similar transfer signature. 
Yeah, I noticed that many vintage amps well into the 80s used triple EFs, even those with RET or LAPT output devices. I think even if you don't need to drive 2 R loads, having a triple is a good idea if you want to get into -100 dB or less distortion territory. The question is how much excess phase you incur which might force you to compensate more aggressively. Maybe the penalty isn't realistically there as you also get better isolation from the output's feedback capacitance.
Re AVRs, last time I looked into this, all models including the very top of the line Denon and Marantz models were using EF2s with integrated Darlingtons. The notable exception were some Onkyos with a "three stage inverted Darlington" output. They started in the upper middle range, e.g. TX-NR 3007 - 3010 and even 1010. Not sure about their current line.
Re AVRs, last time I looked into this, all models including the very top of the line Denon and Marantz models were using EF2s with integrated Darlingtons. The notable exception were some Onkyos with a "three stage inverted Darlington" output. They started in the upper middle range, e.g. TX-NR 3007 - 3010 and even 1010. Not sure about their current line.
Personally I’ve never had a bit of problem compensating EF3’s. And I’ve always gotten superior performance to anything else. There are only two legitimate gripes against them - the extra 2Vbe voltage swing loss and the lest-than-perfect thermal tracking which causes more bias drift and slower settling after a change in dissipation compared to EF2 or CFP. That and the cost, but we don’t worry about two 50 cent transistors when building amps.
Bob Cordell has a nice solution for that in his ed. 2 amplifier book: reverse the polarity of the drivers to make the pres and drivers some sort of diamond buffer. Don't think he came up with this but he puts it in as an example. Does away with both of the problems you mention.
I’ve seen that done before too - using CFP’s for the drivers. Removes the first order thermal dependency on the 2nd stage. But there is no way to stabilize the first stage dissipation. It will undergo wild swings, and normally you use a small signal transistor, because the first stage of a CFP needs to be FAST. Whether or not it helps with the thermal tracking depends on how high the self heating is in that predriver. If you’re really beating on it because your output stage (and driver stage) beta are falling into the low single digits, you’ve made the problem worse than it would have been if you used an EF3 with a high bias in the first stage. If you’re using nice LAPT outputs (or at least sustained beta types) and good drivers the arrangement does what you’re proposing. But those cheap modern Made in China Crown amps do it with MJ15024 outputs and TIP50 drivers, exactly what do do not want to do here. They would have less thermal issues if it were just run as an EF3 with the drivers on the output heatsink and 6 to 10 mA in the predrivers, like all the old-school pro amps do.