Lasersaber strikes again. A joule thief king ?

[MarkE]

Conrad with 2V or more and very careful firmware design you should be able to get the MPS430 or similar low power uC down to 0.5uA or less average operating current.  The sacrifices that you have to make getting there may force a larger physical coil.

[conradelektro]

I worked a bit on my MAX931 circuit and could light a white LED a little bit brighter by help of a coil and a BC547C as switch.

Vortex1 was right, driving a LED with the back EMF spike of a coil makes it appear a bit brighter (for the same power consumption as feeding current directly through the LED).

MarkE was right, it needs a bigger coil.

One can now play endlessly with duty cycle and coil impedance to reach an optimum.

The lowest power consumption I could reach is 9 µA at 2.5 V, but the white LED is of course just shining, not really bright.

Note the 1N4148 diode in series with the white LED, it makes the LED a little bit brighter and lowers power consumption because the LED is a bad diode (high reverse current of LED consumes power, it helps most with a red LED).

Removing the Ferrite core from the coil reduced its inductance from 355mH to 48 mH. This makes the LED brighter but power consumption goes up to 30 µA at 2.5 V.

All measurements are stated on the circuit diagram. Note how clean the square wave on the output of the MAX931 is (although I use Mega Ohm resistors). Also interesting is the ringing (over collector emitter of the BC547C) which was also observed by LaserSaber in his 2N1304 circuit.

I found the low power OpAmp TLV2401 http://www.ti.com/product/tlv2401 which has the lowest power draw today. According to its data sheet from TI: "The TLV240x family of single-supply operational amplifiers has the lowest supply current available today at only 880 nA per channel." I might get some, looks interesting, but needs 2.7 Volt and the output is only 0.2 mA (enough to switch a little NPN transitor like a 2N2222, 2N3904 or BC547).

Concerning the leakage of electrolytic capacitors: my measurements showed that the leakage is low at 1V to 2V, around 1 µA, but then it goes up, at 10 Volt it easily is 10 µA to 20 µA.

Next week I try some fly back transformers to light a LED lamp. Keep the suggestions and ideas coming!

Greetings, Conrad

[MarkE]

I worked a bit on my MAX931 circuit and could light a white LED a little bit brighter by help of a coil and a BC547C as switch.

Vortex1 was right, driving a LED with the back EMF spike of a coil makes it appear a bit brighter (for the same power consumption as feeding current directly through the LED).

MarkE was right, it needs a bigger coil.

One can now play endlessly with duty cycle and coil impedance to reach an optimum.

The lowest power consumption I could reach is 9 µA at 2.5 V, but the white LED is of course just shining, not really bright.

Note the 1N4148 diode in series with the white LED, it makes the LED a little bit brighter and lowers power consumption because the LED is a bad diode (high reverse current of LED consumes power, it helps most with a red LED).

Replace the 1N4148 with a small Schottky like a BAT54 or BAT46 and you will decrease the voltage drop while still blocking losses due to the LED's slow reverse recovery time.

Removing the Ferrite core from the coil reduced its inductance from 355mH to 48 mH. This makes the LED brighter but power consumption goes up to 30 µA at 2.5 V.

All measurements are stated on the circuit diagram. Note how clean the square wave on the output of the MAX931 is (although I use Mega Ohm resistors). Also interesting is the ringing (over collector emitter of the BC547C) which was also observed by LaserSaber in his 2N1304 circuit.

A low threshold voltage power MOSFET may do much better than the BC547 NPN.  They aren't that easy to find in leaded parts.  If you can deal with surface mount parts, a DMG1012 from Diodes Inc only needs about 1nC gate charge to fully turn on.  If you are pulsing at 100Hz or less that will be less than 0.1uA to turn the MOSFET on and off.

I found the low power OpAmp TLV2401 http://www.ti.com/product/tlv2401 which has the lowest power draw today. According to its data sheet from TI: "The TLV240x family of single-supply operational amplifiers has the lowest supply current available today at only 880 nA per channel." I might get some, looks interesting, but needs 2.7 Volt and the output is only 0.2 mA (enough to switch a little NPN transitor like a 2N2222, 2N3904 or BC547).

The DMG1012 might help you there.

Concerning the leakage of electrolytic capacitors: my measurements showed that the leakage is low at 1V to 2V, around 1 µA, but then it goes up, at 10 Volt it easily is 10 µA to 20 µA.

Next week I try some fly back transformers to light a LED lamp. Keep the suggestions and ideas coming!

Greetings, Conrad

At 1.8V, 100Hz and 10nF your timing circuit is using about 2uA by itself.  If you have a smaller capacitor like 1nF or 2nF and larger resistor values you could see some improvement there.

[d3x0r]

more progress 5 Hours; fairly bright; 3300uF

[conradelektro]

@MarkE: thank you for your input, the DMG1012 looks very interesting, just the right transistor to be driven by a TLV2401 multivibrator narrow square wave.

I will go in the direction of driving a 220V LED lamp with two AAA batteries and the multivibrator. The 2 V to 3 V from two 1.5 AAA or AA batteries seem to be just right.

What I have to do now is find or wind the right transformer, but I have a few sitting somewhere in my house. I think that a useful emergency light or flash light can be done with 0.1 Watt.

I found that a LED needs at least 1 mW to shine brightly, and there are plenty of Joule Thiefs doing that nicely.

I like what LaserSaber is doing, the 1 µA power draw is amazing, but I dread the winding of different pot core coils. I will wait to see where he goes before running in the same direction.

Greetings, Conrad