Request for Experience (Current Pulse FET driver)

[mdwebster]

Greetings to all,
I am a recently-minted, like-new, rarely-used, still in original-packaging electrical design engineer and I've run into my first real stumbling block 6 months into my first job out of college.

I'm looking to build a (relatively) constant current source for pulsing a pair of IR LED's from a nominally 5V source (USB or 4AA batteries depending on use). The pulse supplier is a microcontroller off of a pin that can source and sink 20mA but only up to around the micro's regulated Vdd, which is 3.3V. Pulse rate is 36kHz with 25% duty cycle. I'm aiming at a ~350mA pulses which should develop ~3.4V drop across the pair of LED's I'm using (TSHF5210).

The circuit that I came up with (looks fairly good in simulation) utilizes the micro output running into a 75 Ohm resistor which goes to a 2.7V zener running to ground. At the resistor-zener junction is the gate of an N-MOSFET, NTK3043N whose source is at ground and drain hooked up to the bottom of one of the IR LED. There is another IR LED in series with that one before being attached to the battery voltage (4.5-6.5V expected range).

The questions I have are:
a) Is using a zener to regulate a FET like this a decent idea? I don't need it to be super-accurate and the spec-sheet for the zener claims 5% accuracy. I'm guessing I need to allow 5 or 6 mA of current to get the zener to regulate decently.

b) Does anyone know of a FET with a lower turn-on voltage? I'm looking at an equivalent circuit on a handheld remote that I would like to pulse even higher (~500mA) but only has access to 2.6-3.2V from batteries) and I can't find a FET that will turn on that hard with that low of a voltage to save my life. Currently, I'm using a design that I inherited from my predecessor that uses a pair of NPN transistors to pulse the current, but I'm running into problems with the main power transistor's turn-off time (I think it's being driven into saturation), so I wanted to see if it was possible to transfer the FET circuit over to this other device.

c) Would it be better to use a P-FET "up top", say, with an open-drain output on the micro that could pull it down to ground to turn it on and a pull-up resistor to turn it off?

My humble appreciation for any advice or help ...
Mike Webster

 

[mneary]

I gather you're regulating the gate voltage to control the drain current. Problem is, the NTK3043N threshold can vary by 0.9V, and there's only the 'typical' value given for transconductance, no limit values. You would probably get better current regulation by saturating the MOSFET and controlling the LED current with a resistor (even with an unregulated source).

The Vbe of a BJT is much more dependable; you can get very good current regulation using an emitter resistor, and putting a regulated drive on the base. Instead of a zener, you could use a voltage divider to save money.

I would try a small NPN, with one ohm in the emitter. One volt on the base (3.3V with a divider consisting of 110R in series and 47R to ground.) You'll have adjust the 47R a bit to account for base current.

 

[crust]

Most of your circuit seems ok. Why don't you just put a current limiting resistor in series with the LEDs, get rid of that 75 ohm and the zener and just fire the gate directly from the micro?

 

[mdwebster]

Thanks for the replies.
The reason I wasn't going with a straight resistor limiter was because I was trying to have a circuit that was relatively constant current across the battery lifetime.
The fact that the threshold varies by ~.9V is a killer though, I didn't realize that from reading the spec sheet. Thanks for that at least, that would have come back to bite me in the ass in a few weeks time.

The simulations I've run using a single BJT, with a 1hm: emitter resistor and a voltage divider from the drive to the base showed a pretty significant voltage dependence.

Ha! I just re-ran the simulation and I must've been using some really messed up values for my resistor divider because this looks a hell of a lot better than last time. Thanks again ... Mike

Edit:
Ahh, the bad voltage dependence was when I was working on a similar driver for a remote control where the overhead voltage is much lower (2.5-3.2V) and the required current is a bit higher (500mA) (but the period is shorter at 455kHz). In that case, there is no regulator and the drive voltage from the micro varies directly with the battery voltage and the whole thing goes out the window. With that device, I have two transistors set up where one regulates the base current to the other by means of a sense resistor in the main drive current path. That seems to be working okay all things considered, but maybe I'll try my zener idea here since the zener is cheaper than the 2nd transistor...

 

[mneary]

Be sure that you use a BJT that still has enough gain at the proposed collector current. The 2N2222 for example, looks almost good enough. (It's a trap.) I assume that this is for production.

Also be sure to simulate over temperature; the BJT is good, but not perfect. I might assume an IR remote will always be used in a 25-35C environment but you know your customer spec.

Zener cheaper than transistor? I wouldn't have thought.

By the way, I picked 1 ohm as a trade off between voltage loss (350 mV) and sensitivity to Vbe changes. You might find there are better values depending on your goals. If you can tolerate more variation, a lower value would decrease your losses.