An Engineer in Wonderland – Sziklai pair excitement

As a temporary measure, I am knocking together a power supply for my LED ceiling.

sziklai-pair-mosfet-200.jpgAs a temporary measure, I am knocking together a power supply for my LED ceiling.

The idea is to check-out the choice of brightness levels and the power circuit before I go the whole hog and make a microcontroller version.

I need to control 0-1A through an opamp and some sort of transistor, and the parts bin has yielded a CA3140 and a BUZ11 mosfet.

As the whole thing needs to run off 5V max (as microcontrollers will not run off much more than this) and I am a little worried that the venerable 3140 is not going to be able to deliver enough volts to turn the mosfet on, particularly as the mosfet has an 0.5R source resistor for current sensing.

The 3140 is a remarkable little beast for its age.

Its p-fet input stage (with a 1T? input impedance!!) allows it to work with inputs above and below the negative rail, it will run from 4-36V single supply, and the output is designed to sink down close to the negative rail.

sziklai-pair-mosfet-boost-200.jpgThe fly in the ointment for me is that the output uses an npn emitter follower to source current and will therefore not get anywhere near the 5V rail.

I am hoping a resistor between the output and rail will increase drive sufficiently, maybe with a diode (pictured) if the output gets confused by the resistor.

If this fails, there is already a TIP34 pnp bipolar attached to the heatsink from an earlier project and, in what will be a first for me, I might just try a Sziklai pair (aka, complementary Darlington) – at the expense of base current getting into the sensing resistor.sziklai-pair-200.jpg

This is quite an attractive idea, and modern as well as old, because IGBTs are essentially Sziklai pairs with the input npn replaced by a n-mosfet.

If that turns out instead to be a bad idea, there are some ancient 2N3055s lurking in a draw with which I can knock together a Darlington pair, although hole drilling will be required to mount the hallowed TO3 packages.sziklai-pair-darlington-200.jpg

Then there is the further option of two voltage rails, 12V for the opamp and the BUZ11, further dropped to 5V for the microcontroller.

BTW, the great Bob Pease has a terrific video covering current sinks which, with his usual brilliance, has incredible precision and temperature compensation from a handful of components.

Can’t wait to get into the workshop.





  1. Thanks for your thoughts Mr Kurt.
    I tried the mosfet over the weekend, running the whole thing on 12V to get enough drive, and it regulates except when it bursts into oscillation at higher current settings (with a 150R gate resistor, and worse with no resistor).
    I noticed that the great Mr Pease has some sort of RC network across his op amp output, and another across the load
    As my layout is not great and my stability-inducing superpower has waned (OK, I never had even a stability-inducing mini-power), I am going to cheat and try the Sziklai (yes, I have been waiting to write that) pair to cut the capacitive load on the op amp.
    Other than that, the 3140 trimmed down to a nice low input offset voltage and the levels are looking good.

  2. Dear Alice,
    I think you’re on the right paths.
    re: Plan B (mosfet with pull-up at the gate): Seems okay as long as the op-amp can handle the pull-up voltage.
    If not… maybe a BJT buffer amp would be okay? Use a voltage divider in the feedback path to permit closing the loop… or just use that current sense resistor?
    re: Darlington & Sziklai pairs: seems okay, but you’ll want to add some resistors at some of the bases to give the leakage current a place to go. The only downsides of using pairs like this is the voltage drop (and cost and board space). I’m assuming that voltage drop isn’t an issue. Or maybe the delay in the control loop is something to worry about?
    I’ll have to check the Bob Pease video. He’s a heck of a clever guy.
    p.s. how long have you been waiting for an excuse to put “Sziklai” in print? 🙂

Leave a Reply

Your email address will not be published. Required fields are marked *