I'm trying to build a discrete driver to switch a power transistor (Qg=12nC, Vg=0->6V). I want to push the limits to see what is possible with discrete components (so no driver IC) so I'd like it to have a rise/fall time close to 20ns. This comes down to Ig=12nC/20ns = 0.6A of drive current, assuming constant gate current during the switching. I thought of using a typical BJT push-pull circuit but I'm struggling to find the parts.


simulate this circuit – Schematic created using CircuitLab

Am I using wrong specifications? Is using BJTs a bad idea and would MOSFETs (maybe multiple stages) be the better choice here?

The specifications I looked for:

  • Considering I need at least 0.6A of gate current, but I'd prefer more: Ic>=1A

  • Vg=0->6V: Vce,br>10V

  • tr/tf gate <20ns: tprop+tr+tf+tstor BJT < 20ns

The main problem is the BJT time specifications. Some BJTs don't have these spec'ed, and the best BJTs I find have rise/fall times of 30 ns. Not even mentioning storage times (but I know these can be lowered by using baker clamps and speedup caps)... However I know BJT push-pull is often used in commercial drivers. Are these specs just not possible for discrete components or am I looking in the wrong places?

Thanks in advance!

  • \$\begingroup\$ This is the only case I can recall where I disagree with Olin. It takes time to clear out the charges given how I think you are using these. It just won't happen. Look for RF transistors -- where you can get under 1pF of charge storage. Not recommending, but just pointing out a common example -- the 2N5770. Even then, I've never been able to get that topology working near those speeds without carefully thinking out how to empty the charge fast (caps with access to low impedance driving are part of the design.) Perhaps look up a reference design, if you can find one. \$\endgroup\$
    – jonk
    Dec 13, 2017 at 18:23

1 Answer 1


Just about any small signal NPN and PNP will work in this role. I use MMBT 4401/4403 for jellybean applications like this, but many many would do.

Since these parts only need to pass the high current in short pulses, the max continuous current spec doesn't usually apply anymore. Don't get extra wimpy transistors, but just about anything ordinary will do.

However, if you're worried about this level of detail, then you should be using a FET driver IC. That's what they are for, and they are specified for the maximum pulse current, for exactly the reason you need it.

  • \$\begingroup\$ Thank you for your answer. So if I understand correctly the only thing I should look out for are for the BJTs to have as small time constants as possible? However these are usually spec'ed for their usual use (e.g. the MMBT4401 you mentioned has tr=15ns for Ic=150mA). Can I assume these values will roughly stay the same for Ic,pulse=1A? Also, the online stores I visit (Farnell/Digikey) don't let you filter on times but only on fT. Even though the fT is a small signal model value, can I assume this gives an idea of its speed? \$\endgroup\$
    – John HTS
    Dec 13, 2017 at 13:09
  • \$\begingroup\$ @eeN: When you are using a part in a way not specified in the datasheet, you don't really know. You are also using them in emitter follower mode, where common emitter rise and fall times are probably not relevant. Still, faster common emitter should mean faster emitter follower. Emitter follower should be faster, but you don't know how to relate to common emitter. Again, if all this matters, then you should be using a FET driver where these things are specified. \$\endgroup\$ Dec 13, 2017 at 13:57
  • 1
    \$\begingroup\$ ... or actually conduct the experiment ("I want to push the limits to see what is possible with discrete components"), and see what comes out of it. Empiricism > Speculation. \$\endgroup\$
    – Dampmaskin
    Dec 13, 2017 at 14:42
  • \$\begingroup\$ Differential circuits can avoid charge-storage. With immense steady-state power consumption. \$\endgroup\$ Dec 14, 2017 at 4:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.