Is that even possible?
I've challenged myself just for fun of mind bending.
The circuit presented below starts at 2mV, operates down to 0.81mV and draws about 5mA.
The operating point while capturing the video was 0.84mV, 5mA which corresponds to power consumption of 4.2uW.
As a load it represents the resistance of 0.168 ohms.
To provide such low and well controlled supply voltage I used bench power supply with current limiting to create voltage drop across 10mohm shunt resistor.
Short video:
Photos:
![](http://rfscientific.eu/sites/default/files/articleimage/img_3343_m.jpg)
![](http://rfscientific.eu/sites/default/files/articleimage/img_3345_m.jpg)
The schematic is presented below:
![](http://rfscientific.eu/sites/default/files/articleimage/1mv_led_blinker_sch.png)
The challenge was divided in two parts: design a low voltage oscillator and design power efficient LED blinker.
The oscillator is a Meissner (Armstrong) type which generates sinusoidal output voltage using transformer feedback.
The Cgs together with transformer's secondary forms a LC tank.
The active component is depletion mode MOSFET. This one is however worth a deeper look.
It is IXTH16N10D2 made by IXYS which has max Rdson <=64milliohm and a typical transconductance of 11 Siemens.
D2, C4, R6 shifts the gate voltage towards -1.5V where the oscillator operates more efficiently.
Regarding the transformer: high turns ratio is required in this application.
LT and others use 1:100 commercial transformers.
I decided to choose the other way. I used commonly available current transformers.
I used two 2500:1 50/60Hz AP-2500 devices from Talema.
Two secondaries are connected in series while primary is just one thick wire passing both CTs.
The inductance of the winding wound on permalloy or nanoperm core is outstanding: 300Henries measured @100Hz just for one coil.
This fact and the high input capacitance of the depletion mode MOSFET result in low frequency of the oscillation:
close to 60Hz which fortunately fits well the operating range of the current transformer.
![](http://rfscientific.eu/sites/default/files/articleimage/img_3335_m.jpg)
*The oscillogram acquired across transformer's secondary just before emitting the light pulse.
Diodes are low leakage schottky BAS70-04.
Low leakage is the main requirement.
Several other types of diodes I tried caused too much loading on the oscillator.
The blinker was selected from dozens of simulations of online schematics.
Lot of interesting circuits can be found here
http://techlib.com/electronics/flasher.htmlThe blinker has truly remarkable features which I was not aware before.
It operates down to 700mV. I'm talking about presence of pulses, as You can imagine LED is not visible at such low voltage.
The period of generated pulses is practically independent from supply voltage.
It is extremely simple and power efficient.
The key point here is that base current for Q2 needed in transition moment is taken from C1.
The circuit tolerates wide range of components' values.
Last important component is the LED. The Vf should be as low as possible plus it shall be high brightness type.
ALMD-EG2E-XZ002 BROADCOM (AVAGO) was choosen.
C7 was initially a 1Meg resistor added to prevent quenching of the oscillations after emitting the pulse of light.
The capacitor however is a low loss solution.
As always any suggestions for improvements are appreciated.
Literature for those who are interested:'Comparison of 3 self-starting step-up DC:DC converter
topologies for harvesting energy from low-voltage and
low-power microbial fuel cells'
https://hal.archives-ouvertes.fr/hal-00638259'A Bipolar 13 mV Self-Starting and 85% Peak
Efficiency DC/DC Converter for Thermoelectric
Energy Harvesting' MDPI Energies
'A low-voltage boost converter using a forward converter with integrated Meissner oscillator'
DOI:10.1088/1742-6596/476/1/012081
LTC3108 datasheet