# Charging a big capacitor from a weak power supply

Let's say I need to charge a 4m7 (4,700uF) capacitor to 5V, from a 1.5V battery with an internal resistance of 100 ohms. I'm trying to use a typical joule thief with something like soft-start, to limit the initial capacitor current.

I found a problem with unstable capacitor voltage, changing in different periods (about 10s). Sometimes the voltage drops to 2 volts, sometimes it rises to 3V.

Where is the problem? I can't find it.

Edit:

To make it clear, Q1 makes a pulsed voltage, after D1 is high voltage DC, which start to charge C1 via R3. When voltage on cathode of D1 is higher than 2.1V, Q2 opens and charges C1 directly. I added R3 to limit initial charging current to prevent attenuation of joule thief oscillation because of high initial load of capacitor C1.

After connecting the power supply, it starts nicely and I measure on C1 3.6V. After a few seconds, it slowly and continuously falls to 1.5V or 2V, and then rises to 3V again.

• What's the point of Q2 and it's associated circuitry? May 30, 2019 at 22:13
• When on emitter of Q2 is more than 2.1V, bypass the R3.
– user208862
May 30, 2019 at 22:37
• Q1 is soft start and Q2 delayed rectifier.. They should call Joule Thief the Battery Killer. May 31, 2019 at 0:01
• i.stack.imgur.com/AfE0h.png May 31, 2019 at 1:08
• I have doubt, if it is a discharging of C1 back to boost converter, it become onloaded and shots off.
– user208862
May 31, 2019 at 10:51

I think your circuit is oscillating, but I am not completely sure.

You will find that an inverting joule thief topology will work much better:

The zener is not yet conducting much, and that part of the circuit has to be done better -- as it is, it will not keep the supercap from being overcharged.

This is not your circuit exactly. The "transformer" is completely different. The 2N3904 makes a better joule thief transistor in this case because of it's better turn-off figure, and therefore less switching losses. Also making it a decent fit is the 200mA limit on the 2N3904 and the low power nature of this converter. It's always better to use a schottky diode for the catch diode (D3 above), as it is much more efficient, so the 1N5817 is much better than the 1N4148 and contributes to the success of this simulation.

At this point in the simulation (29 seconds), V2 is putting power out at -14.721mW, the battery internal resistance is wasting 9.63mW, D4 is dissipating at 2.9046mW, Q2=1.3675mW, Schottky D3=0.3021mW, R2 at 1K spends 0.2747mW, and C3 is only being charged at 11.533µW.

• Hah, I opened this question, latter I canceled account. You are right. The inverted boost is much less sensitive to high output capacitance since it allows higher inductor voltage swing when cap is at 0V. Feb 22 at 13:42