I'm trying to design a UPS circuit for microcontroller backup. Require 3.7-5V, and less than 50mA for 10 seconds every 15 min. Power cuts in my area are never longer than 3hours, and occur rarely.

I'm wondering if the inductor of the oscillator drive part of the joule(!) thief can be made any simpler.. i.e. can be replaced with something else that I do not have to wind myself, or maybe a capacitor...

The simulated output shows 4V at about 70% efficiency on battery. Trickle charge is continuous and fixed at about C/100. Should take 100 hours to charge the cell, which can supply power for a week of downtime. I'm guessing the NiMH will always be above 1.2V, and should not be damaged at 10mA charging current. Joule thief part draws about 2mA with no load, or when powered from mains.

I pieced together this circuit from things on the web, so all suggestions will be appreciated.

LTSpice schematic

added: I was going with the single NiMH because (as has been pointed out by Jasen), there would not be an issue of polarity reversal. Also no fire hazard, no complicated charging circuit (as in lithium), and cheaply available (even older used NiMH would be fine as they would all hold charge for at least a day).

Q3 ensures fairly accurate constant current. I was unable to fix that confidently with a resistor, since I don't know beforehand the voltage and resistance of the NiMH cell. Especially so, since the input voltage after the diode may vary from 3.8V (4.5V USB - 1n4148 drop) to 4.8V (5V USB - 1n5819 drop). But maybe a resistor is possible, as you say. I don't want to let the trickle charge current increase too much as it's always on.

LT1073 costs a dollar, and not readily available (I'm in India). Parts in my circuit cost 5-10 cents and available everywhere. The coupled inductor does mean that I have to wind it by hand everytime though.

I was of course looking for a solution with a single coil inductor (for the boost circuit), and something else to drive the oscillator. That would allow me to use a cheap off the shelf inductor (like they do in the LT1073 and all other chips). But maybe it is too complex to make with a few simple discrete parts...

Another option is the dirt cheap HH004F, but I have been unable to find any complete datasheet (in English or even Chinese).

  • \$\begingroup\$ 3,7 to 5V sounds like an application fro a lithium cell (4.2 to 3.3v) or 4 nimh cells in series (5v to 4.7v) \$\endgroup\$ – Jasen Jan 4 '18 at 20:28
  • \$\begingroup\$ anyway going with what you have designed Nimh is not damaged by deep discharge, only by polarity reversal, and a single cell isn't going to suffer from that. \$\endgroup\$ – Jasen Jan 4 '18 at 20:30
  • \$\begingroup\$ for charging from 5V a single diode and resistor is simpler. what does Q3 gain you? \$\endgroup\$ – Jasen Jan 4 '18 at 20:35
  • \$\begingroup\$ I don't think you want to make a jewel thief, that could get you in trouble... \$\endgroup\$ – alex.forencich Jan 4 '18 at 21:12
  • 1
    \$\begingroup\$ Do you really need to use discrete only designs? There are many off-the-shelf ICs that will do what you want and are much, much simpler to implement. The LT1073 is one of many simple "AA battery to 5V" boost ICs. Also if you can get small lithium polymer cells, then something like a MCP73831 can keep it charged and you could use the output of that lipo cell directly for your MCU. The Featherwing boards at Adafruit use this architecture. Much easier than what you are trying to do. \$\endgroup\$ – Vince Patron Jan 4 '18 at 22:09

Just Don't.

instead use one of the microcontroller pins to drive a simple inductor in boost mode and have a capacitor big enough to keep the processur alive during the sleep phase.

each time you wake check the capacitor voltage and pump a few pulses into it until it's high enough

since you need well under 1mA any micronctroller pin should be plenty strong-enough for the task.


simulate this circuit – Schematic created using CircuitLab

at the top R3 provides C/20 charge current suitable for a 800mAh NiMH cell, D2 and R4 provide power when ther usb charger is connected, allowing the micrcontroller to operate without using the boost converter.

below that, on the left is a boost conveerter driven by poin PA1, set PA! low for a few cycles then set it high-impedance to allow the energy from L1 to feed C1

On the right is a voltage checking circuit, when needed set PB2 high and read PB1 depending on the voltage in C1 PB1 will either read high or low, when you know enough set PB2 low to save energy. . You will find that this circuit switches state well below the marked zener voltage, some experimentation may be needed.

| improve this answer | |
  • \$\begingroup\$ oops misread - 50mA required, might need a transistor in there to pull on the inductor instead of using GPIO diectly. also remove. need to beef other power parts up too. \$\endgroup\$ – Jasen Jan 21 '18 at 0:56
  • \$\begingroup\$ Interesting solution! Although I do wonder about the practicality in this case. 1mF cap will discharge in ~1.1sec from 4.7 to 3V @ 1-2mA. So battery depletes rapidly in case of actual power outage as U1 has to wake up every second. Also, large caps take up space and eventually fail. Besides, why to charge one power supply from another? \$\endgroup\$ – Indraneel Jan 21 '18 at 18:21
  • \$\begingroup\$ what needs 1mA? many microncontrollers sleep at less than 1uA \$\endgroup\$ – Jasen Jan 21 '18 at 19:05
  • \$\begingroup\$ OK, maybe it can sleep for 5 min with a current draw of 6uA. I'll accept your answer, even though I've never actually tested it. Thanks a lot, Jasen. \$\endgroup\$ – Indraneel Jan 23 '18 at 4:38

your 47:10 inductance ratio is tricky, 7:15 turns gets you close. (46:10 inductance ratio)

but if 4:1 is close-enough there might be an off-the-shelf part you can use.

eg: Bourns PM610-10-RC has 6 individual windings so you can make an inductance ratio 4:1 quite easily.

| improve this answer | |
  • \$\begingroup\$ MOQ 150 from electro-sonic , elsewhere 600, dunno if Bourns will give samples. \$\endgroup\$ – Jasen Jan 4 '18 at 21:14
  • \$\begingroup\$ bourns.com/support/request-samples \$\endgroup\$ – Jasen Jan 4 '18 at 21:15
  • \$\begingroup\$ Thanks Jasen, but as you point out, difficult to get in small quantities (and I'm in India, so really difficult to get cheap). Also, it's huge. \$\endgroup\$ – Indraneel Jan 5 '18 at 3:34
  • \$\begingroup\$ there are other inductors with multiplel windings out there... mouser has several listed under the category "pulse transformers" \$\endgroup\$ – Jasen Jan 18 '18 at 4:32
  • \$\begingroup\$ eg: Murata 78604-4C mouser.in/ProductDetail/Murata/78604-4C/… It has a 67uH primary, so the secondary will be about 15.5uH \$\endgroup\$ – Jasen Jan 18 '18 at 4:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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