How can I regulate 1.2v NiMH to output ~1.0v

Thread Starter


Joined Jul 1, 2020
Can anyone suggest some ways to efficiently regulate the power from a 1.2 NiMH battery to be ~1.0v? The battery powers a small motor and the current is only about 40uA. This is for a marathon type experiment. I considered a couple resistors to make a voltage divider, but as the battery runs down the “divided” voltage will drop proportionally and motor will stall – plus the resistors are inefficient. I really need to REGULATE rather than LIMIT the voltage because as the battery begins to drop voltage, I would like to maintain the 1v source as long as possible. Maybe some type of Zener or regular diode with super low breakdown voltage? Thanks for your help!


Joined Jul 11, 2016
confirm your motor requires only 40µA = 40·10 ̄ ⁶ A
  • . . . going switching your efficiency is about 58% . . . 78% with most scenarios
  • Ge BjT-s are leaking 1...2µA . . . tough they can be used down to 420mV
    there might be some small signal Si ones ? i ! guess like 2N1973 ? PMBT6428 ? (i've almost not used such)
    that might do better than commonly used Si ones
  • Joule thief operates down to 750mV (but i donno at what efficiency)
    regulating it to 1V OUTP throughout the 1.2 ... 0.75V INP is all hell (but not totally impossible)
  • ?? special j-Fet or sub-threshold MOS-Fet circuits . . . i guess can't be achieved by conventional commercially avail parts ??
  • MOS-Fet-s are efficient only if they are driven fast (requires higer than 1.2V and i guess more than 40µA)
    . . . driving large inductors somewhat reduces the speed demand . . . but the circuit goes tricky
(nothing else to do = ) some essential tests -- seems the MMBT6428 is not too good
... also the russian КТ203 goes hot fast above Ic >10mA (irrelevant at 40µA target) ,
the МП38 might be more leaky (=conducting=ON) than the simulation shows (e.g. it might be unusable for such low power)
The TEST just compares different bjt-s -- it does not have much to do with switched or non-switched regulation of motor power
(the ehc.- or dpl.- MOS-Fet-s likely require additionally boosted voltage for sufficient driving . . . likely . . . ? at worse case - extra converter and losses)
ss-TEST - L-0a.png
Last edited:


Joined Jan 30, 2016
There is no easy way to do this because 0.2v is below the threshold of any currently publicly available technology. Even the nanoWatt SMPS devices for wearable don't manage that.

The obvious solution is not using a dead technology like NiMH - Li-Ion gives better energy density and a higher voltage in the same size package. With 3.6v to start with and 2.5v terminal volts you've got far more options.

sparky 1

Joined Nov 3, 2018
A very small current from the battery is multiplied just enough to supply a 1.6V zener. Or get the 1V reference.
The constant 1.6 volts is divided with a resistive voltage divider to 1.000 V using a small 100k 10 turn pot.
This reference is compared to capacitor bank that is maintained at 1.000 Volts used to drive the small motor.
The step down has a high efficiency possibly 98% but later as the battery gets too low the step up efficiency becomes the limiting factor. The efficiency steppng up battery source when it gets down to 780 mV when it cannot fill the capacitor because it is tired.
Last edited:


Joined Sep 17, 2013
Looks like the current draw of that circuit, via R1,R2,R5, would exceed the 40uA motor current by some margin? That would presumably be a problem for a 'marathon experiment' :(.


Joined Jan 30, 2016
Quite possibly.

By inspection the LH side is ~( 1.2v - (Vf_D1 + Vce_Q2) )/10k = (1.2-(.41+.2))/10k = 0.6/10k = 60uA

the central current mirror is ~(1.2 - Vbe_Q1-Vce_Q2)/100k = (1.2-.6-.2)/100k = 0.4/100k = 4u

The RH side is Vbe_Q4/100k = 6uA

Then the output stage is 1v/384k = 3uA

Total ~43uA

So 50% efficiency approx. Better than the ADR510 but still not great for a marathon!


Joined Jan 30, 2016
My estimate was worst case. Yours may be closer - and I note you've changed a few resistor values - but its still an estimate based on assumptions on the accuracy/validity of the models used for those transistors. My own experience down at these levels suggests there can be a lot of variation between theory/model and reality! Have you actually built and characterised one to validate the model?


Joined May 20, 2015
I didn't think I was talking to pedants. You can use agreed transistor pairs in small enclosures. But even using non-negotiated transistors, you can change the output voltage by changing the value of one of the divider resistors. I didn't know that every microampere was important to you. The circuit is also capable of operating at lower currents! That's why I reduced the current. I think 10 µA is an acceptable loss. In any case, you will need to adjust the output voltage.


Joined May 20, 2015
Here I have replaced transistors in my scheme with other ones (and naturally their spice models are different, different manufacturers. The result has changed little). I did not make any selection.


Joined Jan 30, 2016
I'm not a pedant, I'm a realist based on 45y of experience! I agree this simulation isn't hugely sensitive to transistors - I get the same results with LTSpice, but I only trust models that I've validated on the bench - simulation is just a better way of getting the starting point, or deciding if its a route to follow...

The OP wants a solution for some as yet unspecified marathon event. So this has got to work not just at 1.2v but all the way down to 0.9v.

Run your model for an input voltage dropping from 1.2v to 0.9v and see what happens (I just did... )


Joined May 20, 2015
At the beginning I simulated in the range of supply voltages from 0.9 to 1.5 V, but then I thought that at discharge from 1 V to 0.9 V at such low currents less than five percent of the total energy would be received. A discharge of up to 0.9 V can be performed at sufficiently high (relatively) currents. And my scheme will behave sneaky and discharge the battery quickly. 2020-07-02_14-49-23.png


Joined Jan 30, 2016
Yes, for the marathon you need to use up the cell to terminal voltage which is 0.9v... but that circuit starts using a lot of current when it goes out of regulation. Maybe it can be tuned to 0.9v and the OP can accept 0.9v as terminal rather than 1v?


schmitt trigger

Joined Jul 12, 2010
I am going to ask what could be a rhetorical question.

All the statistical data that I have seen for Marathon times, the vast majority of runners finish it between 3 and 5 hours. Of course, high performance runners take much less than that, while some stragglers can take up to 7+ hours.

Let's say that the regulator circuit is very inefficient, and the total current draw becomes 1 mA.

Wouldn't any NiMh cell, even a lowly AAA one, be perfectly capable of supplying 7 mAHr?

Don't get me wrong, I am fascinated by the discussion of ultra low power regulation. The point that I am attempting to drive is that the TS may be overthinking this "issue".

Additionally, like member ci139 above, I feel the TS is withholding crucial information from us. What type of motor only consumes 40 microwatts? And driving what type of load? Is this some sort of electric wristwatch movement?

If the latter, speed accuracy is achieved by the movement, and is fairly independent of the battery voltage.