Boosting 0.3V to 2V or higher?

[godfreyl]

The amplitude of oscillation is limited by the fact that the gate draws current when it is more than about 0.5V positive, relative to the source.

If you can get the circuit to oscillate, the amplitude can be increased by adding a high value resistor, bypassed with a capacitor, in series with the gate. This allows the gate to float to a negative voltage, allowing greater voltage swing.

I got the results below in a simulator. Reality is always different, but it does look encouraging.

 

[D.A.(Tony)Stewart]

for "maximum power transfer" try to match the impedance of source and the load.

If current is say 10µA@0.2V then Rs~20kΩ with available power=V*I=2µW

Consider a red 5mm LED rated at say 20mA @1.6V, hence, RL~80Ω , V*I= 32mW.

We know load RL is mismatched and won't work, but if it is pulsed so that it is running dim and equiv dc current is say 2mA @ 1.4V RL~700Ω, V*I=2.8mW is still way too much load for this situation.
offtopic rant...
We also notice the nonlinear diode knee shaped V-I curve has an effective slope which we can approximate here as (1.6-1.4V) / (20-2mA)=11Ω which we call the Effective Series Resistance ESR. The higher the power rating of the LED, the lower the ESR. 1W LED devices are usually 0.5Ω or better. Keep this in mind when choosing series current limiting resistors, when add an R, you are also compensating for variations in ESR to make the total more stable and thus predictable current because ESR is also very non-linear and is an indicator of the power dissipate as well as size and quality of the chip.

 

[dselec]

I think mtwieg solved the problem.
Thanks m... u did it any interference or am or FM transmission or even the transition caused by connecting the dc volt source to the chip will get things going will be collected and transferred to the fet gate and oscillation starts.
and godfrey your circuits looks good too..thank u 2.

 

[mtwieg]

I remember a few years ago I went to a trade show which was right after Linear released the first version of the LTC3108, along with rest of their energy harvesting lineup. I asked about the details about the oscillating transistor (they used a depletion MOSFET symbol in their schematic, which can't be right unless it's a special zero Vgs type), but they wouldn't give away any hints (though they did say it wasn't a zero Vgs FET). They actually showed a neat photo of their proof-of-principle circuit for the step up oscillator, which was all discrete components soldered to copper clad. The transistor was in one of those TO-18 metal cans, with the package markings censored out. I'm pretty whatever the IC uses, it must actually be a MESFET or a JFET, and the datasheet just hides that, and it's the secret to their good performance.

But in any case, you shouldn't need anything terribly fancy since you have 0.3V to work with, not 20mV.

 

[FvM]

for "maximum power transfer" try to match the impedance of source and the load.
If current is say 10µA@0.2V then Rs~20kΩ with available power=V*I=2µW
Consider a red 5mm LED rated at say 20mA @1.6V, hence, RL~80Ω , V*I= 32mW.

The calculation is much easier. Assuming 0.2V/10uA is already the maximum power point, you know that you have a little bit more than 1 uA to run your LED. That's it. Just decide, if the brightness satisfies your needs.

 

[dselec]

artluv said
"Pretty much.
The current at the source is in the 50µA-1mA range.
At the output i want to see a charged capacitor or battery - 5V would be about perfect, 2V sufficient."
thats max 200mwatt input even if the circuit has 50% loss he will/can charge 2v with 50ma... it may take him a day or 2 but it will save him if he got stuck in the desert.
I say its a start up idea for travelers, maybe a small gadget on the hat /cap signalling SOS.

 

[D.A.(Tony)Stewart]

The LTC3108 is designed for very low voltage sources starting from 20mV such as thermopile , Peltier effect or PV cell sources. It does a good job at 25% efficiency with 200mV input at 4.5V out @1000µA (4500µW out) implying that input current is 90mA. (18,000µW)

It won't work for the theoretical question in this thread.

10µA at 0.2V ( 2µW ) is not enough to harvest for any practical purpose.
Any conversion to a higher Voltage without external power is impractical.

 

[dselec]

Yep at 1:20 ratio u need atleast 2ma /50mv input at no output(shown by graph)... maybe diff ratio will get diff.results like 1ma with 100mv input .....so atleast 2 cell needed.
ok another idea take 10 low 100uf leakage capacitors connecte them in paralell and after they are charged connect them in serial getting enough current and voltage.
now all members in here lets start doing a circuit that will charge 10 caps for say 10sec then they will be connected in serial transferring the charge to the batt with constant current charge for 0.1 sec on and on until cell fully charged... how about it can we all do that and help this poor fellow out in the desert ?:smile:

... just remembered someone talked about joule (sorry don't remember who) joule will help here
and maybe we can use the ltc too....errr godfrey will help too
also check this site http://www.dicks-website.eu/fetosc/enindex.htm

 

[Artlav]

ok another idea take 10 low 100uf leakage capacitors connecte them in paralell and after they are charged connect them in serial getting enough current and voltage.
now all members in here lets start doing a circuit that will charge 10 caps for say 10sec then they will be connected in serial transferring the charge

Actually, that's a good idea.
I don't need a cold start capability.
Is it possible to make a charge pump-like voltage multiplier that would extract more power than it consumes, provided that there is already an existing reserve?
The device i think of powering consumes on average 10µA at 2.3V, and includes a microcontroller, so switch timing won't be a problem.

Simplified (maybe excessively so, but that works manually) what it would take is this:

The source's connection moves back and forth along the smaller capacitors, and the big capacitor is charged summarily.
That would be easy with relays, but these would take too much power.
There are solid-state relays out of two MOSFETs which should take approximately no power at all, but need optocoupling that takes 2-5mA.

So, is there any practical way to do this kind of switching, or practical way of doing some other kind of charge pump?

 

[FvM]

A serious problem with the charge pump idea is that the FET switches need a minimal gate voltage for operation. Once you have charged the output capacitor to certain level, e.g. 1.5 V, a CMOS charge pump circuit would be able to maintain the output voltage.

I still think that a self oscillating (depletion mode) FET oscillator as anaylzed by godfreyl is an appropriate and feasible solution, even at the intended low current level.

 

[Artlav]

A serious problem with the charge pump idea is that the FET switches need a minimal gate voltage for operation. Once you have charged the output capacitor to certain level, e.g. 1.5 V, a CMOS charge pump circuit would be able to maintain the output voltage.

Can it be supplied from the capacitor it is charging?
That one should have more than 1.8V at all times, initially charged by battery, then maintained by the charge pump from the 0.3V cell.

I still think that a self oscillating (depletion mode) FET oscillator as anaylzed by godfreyl is an appropriate and feasible solution, even at the intended low current level.

Maybe, but i'm yet to succeed in making one - getting/making the right transformer turned out to be somewhat tricky.
So i'm exploring other options as well.

 

[godfreyl]

Why don't you want to use the Linear Tech chip that was mentioned earlier? That seems to be the easiest way.

 

[FvM]

Why don't you want to use the Linear Tech chip that was mentioned earlier? That seems to be the easiest way.

The no-load input current of LTC3108 will be in a mA range. Artlav gave this specification:

The current at the source is in the 50µA-1mA range.

Can it be supplied from the capacitor it is charging?
That one should have more than 1.8V at all times, initially charged by battery, then maintained by the charge pump from the 0.3V cell.

Yes, but the circuit needs to be started from an external voltage source, if the capacitor gets discharged. Apart from this problem, a charge-pump circuit could be build from low voltage CMOS analog switch ICs. You need 20 SPDT switches and a clock oscillator for a x10 multiplier. Alternatively, you can try cascaded charge-pump multipliers.

 

[sak.attish]

Dear Artlav
your problem can be resolved with a simple voltage Doubler Circuit .



Best Regards

 

[dselec]

All good guys .. once u get the iF capacitor charged (mechanical)we can use that charge to make the electronics to do the #29 post automatically to keep things gong on for ever controlling the charge and discharge voltage for the 1uF cap.

 

[Artlav]

The amplitude of oscillation is limited by the fact that the gate draws current when it is more than about 0.5V positive, relative to the source.

If you can get the circuit to oscillate, the amplitude can be increased by adding a high value resistor, bypassed with a capacitor, in series with the gate. This allows the gate to float to a negative voltage, allowing greater voltage swing.

Getting back to that idea, it mostly worked.
I'm getting about 10µA out of it in broad sunlight, with up to 9V of voltage.
That gives efficiency of about 1%?

The problem is, it does not self-start and collapses when the secondary gives less than about 5V - it needs to draw just under 1mA from the source to start oscillating.

The transformer i got is 1:70, 10Ω primary, 7KΩ secondary.
Oscillates at around 1 kHz.
With 2 JFETs i get about twice the output power, but noticeably higher threshold.
Increasing C1 reduces frequency and also allows more power through, but increases consumption.

I was able to get it to self-start with a 820MΩ to 2.2GΩ range of R1, but it gives 1µA at best, and appear to barely oscillate regardless of C2 size (1..100pF).
Oscillation appear to be pings 20ms apart, which makes me suspect mains hum as the source of oscillations.
It is never the less proportional to the source's current in amplitude, so it's not parasitic, and self-starts in the 50-100µA range at the source.

Anything meaningful i can do to improve it?

I got the results below in a simulator. Reality is always different, but it does look encouraging.

What did you use to simulate it?

your problem can be resolved with a simple voltage Doubler Circuit .

Which of the many variants?

Apart from this problem, a charge-pump circuit could be build from low voltage CMOS analog switch ICs. You need 20 SPDT switches and a clock oscillator for a x10 multiplier. Alternatively, you can try cascaded charge-pump multipliers.

That's a lot of not-too-cheap chips, and at 1µA per chip that would already exceed the output if the oscillator idea above is in scale.

I suppose i need chips like MAX4616 or ADG513?
Anything better you can recommend?

 

[FvM]

I suppose i need chips like MAX4616 or ADG513?
Anything better you can recommend?

I rather thought of cheap analog switches from logic IC series. E.g. 74auc2g53, a 0.8..2.7 V DPST switch https://www.ti.com/product/sn74auc2g53

 

[Artlav]

I rather thought of cheap analog switches from logic IC series. E.g. 74auc2g53, a 0.8..2.7 V DPST switch https://www.ti.com/product/sn74auc2g53

Bad example: for each chip - supply current 10µA, input current 5µA, leak current 1µA.
While the power budget is from 10µA to 200µA (at 3V), assuming 100% conversion efficiency.

That site do have some interesting options, like ts5a3160 and ts5a3159, which appear to fit the power budget with room to spare - 10nA supply, 20nA leaks, 2nA input.
Overkill, missed something, any other reason for being a bad idea?

Also, are there any other chips that would fit the bill without being bulk-quantity-only unobtanium?

 

[FvM]

Bad example: for each chip - supply current 10µA, input current 5µA, leak current 1µA.

No, rather shallow datasheet reading and lack of knowledge about electronic components. 10 µA is a maximum leakage current specification, you'll hardly find exemplars with more than a few 10 nA leakage. If you are planning a production design, you possibly need to consider worst case data. It would be easy for the manufacturer to tighten the specification, but the usual application doesn't need it.

In real life, the dynamic power dissipation of CMOS logic will be more critical. The operation frequency has to be selected carefully to minimize losses.

 

[Artlav]

No, rather shallow datasheet reading and lack of knowledge about electronic components. 10 µA is a maximum leakage current specification, you'll hardly find exemplars with more than a few 10 nA leakage. If you are planning a production design, you possibly need to consider worst case data. It would be easy for the manufacturer to tighten the specification, but the usual application doesn't need it.

In real life, the dynamic power dissipation of CMOS logic will be more critical. The operation frequency has to be selected carefully to minimize losses.

That's nice to know, i am in fact not familiar with this kind of chips and took the datasheet at face value.

However, i tried to make a charge pump with the ADG511 chip i had at hand.
The spec says 1µA max supply current at 5V.

I measured 19µA of total consumption at 70Hz, excluding the oscillator.
45µA at 1000Hz.
15µA at 50Hz.
Finally, 140µA at idle.

So, how do you select the operation frequency carefully?