Turning off an LM334-based led driver

I want to drive power led from a Li-ion cell.


I have a few specific requirements.

One is that the driver must tolerate a second current source feeding the led, and only supply current when the other source is under 250mA.

It should also not be blown up by this second feed.

And it should shut down to a few µA when switched off.

There do seem to be a few chips that can do this lot, but I thought I would try a home brew version.

Powering a white power LED from a Li-ion cell can take a fair amount of ingenuity as their operating ranges overlap.

– The output of a Li-ion varies from 2.7V to 4.2V.

– LED forward voltages spread from 3.0V to 3.6V or more.

Hence the number of charge-pump drivers with 1x and 1.5x modes.

Incidentally, the efficiency of these has to drop to 70% or less at certain input-output voltage ratios.

There are a few very neat inductive switchers that mode-switch between boost and buck, where the coil is driven by a bridge of mosfets.

However I am going to cheat as I am driving Cree XR-E leds which have a typical voltage drop of 3.4V, and for most of their discharge life Li-ion cells stay above 3.6V.

So, there is 200mV to play with and a linear circuit is possible with 94% efficiency.

In the past I made something similar based around an op-amp and a reference which worked well, but I came across a neat circuit from Nat Semi’s analogue guru Bob Pease that uses an LM334 current reference, which happens to have an inherent hidden 68mV voltage reference.

leddriveronoff250.jpgIf the voltage across the resistor drops below 68mV, the chip draws more current through its top contact. This further turns on the PNP transistor and feeds the load.

BTW, feeding extra current down the LED from the other source should not cause problems as the LM334 sensing pin, according to the data sheet, is good for 5V above the negative pin.

Pease didn’t discuss turning it off, but it seems to me, and I haven’t tried it yet, that an NPN transistor in series with the top contact should do the trick providing its base is switched from rail to rail.

 Actually, if the base gets all the way to the top rail there may be trouble as if the NPN has a lower Vbe current will flow from its emitter rather than the PNP’s base – but the diode should sort that out.

A couple of resistors as a potentiometer should also do the trick, and provide a base pull-down to ensure the NPN turns fully off.

A PNP in the same position would also be a possibility, but this would open up a leakage path from the alternative current source backwards through the two emitter-base junctions when this circuit was switched off.

Any comments?



  1. Ah ha.
    Zetex, based in Oldham, do a range of low saturation voltage PNPs.
    The FZT968, for example, will saturate to 50mV at 500mA with Ic/Ib of 100, and to 70mV at Ic/Ib=200.
    These ‘matrix’ transistors are the brainchild of Zetex’ Dave Casey, and are about the best there is for saturation voltage.
    I got some legged ones free on the front of Electronics World once, they are good.

  2. Dear Alice,
    The Vce(sat) of the top transistor will probably use most (or all) of the 200mV “overhead” between the battery and LED. A FET solution may be possible, but in my experience they need more gate volts to give a decent “on” resistance than is available here, so will need cunning driving.
    A germanium transistor (dates me) could get just below 100mV Vce(sat), but introduces leakage and availability problems.
    Philip Spencer

Leave a Reply

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