25mV Joule Thief powered by peltier merely using our body heat -Free energy 24/7

[magpwr]

Hi Acmefixer,

It's a good question how i came to find 2SK170 to be the best.

There was this crazy experiment i was doing with same toroid used for my 45mV joule thief which i was using Japanese transistor 2SD1450 but it not the self starting version,as shown in youtube "sanjev21".

I did a lot of research and found that mosfet can never be used in joule thief,because it do not release immediately(in ms or ns) after switching on.
Althought mosfet is a "excellent switcher" if combine with microprocessor.

It was a long shot.There was youtube video whom posted about "self starting stirling engine" using peltier at the source of heat to spin motor to start engine.
It was using mosfet 2N7002 with inductor to positive.The startup voltage was laround 0.4v.

I experimented using "multisim a virtual electronic lab".The transformer in it i'm stuck with "TS_POWER_25_TO_1" (100% efficient) 25 turns:1.(The only transformer i could use for ultra low voltage joule thief)
Tried all the transistor,mosfet,fet,jfet.Then found JFET had the self starting capability.Then spent 3 days to find the best one then slowly lowered source voltage to 0.025v.

I have "9 version of ultra low voltage joule thief" virtual lab file but only 3 is worthy of mentioning as current handling for the rest is not good enough.

JFET standalone version as shown in my youtube video, better version is combining ZTX1047A or ZTX1048A with 2SK170(Since transistor got better current capability using JFET to auto kickstart transistor as well),then there is a version
which i used JFET with 2 resistor to control the current going to transistor.The problem with last version the startup voltage is little higher
but give me "constant current draw" all the way to 2v source.

I tested with 40 over key component or components in virtual environment to derive this ideal circuit by accident.Please ignore 330Ohms resistor to diode as i placed it to allow circuit to function if provided i'm using 1000uf cap at output for 25mV input.

[gyulasun]

Hi Folks,

In the meantime I have done some patent search on this low voltage DC-DC converter topic and I think I found good info.

First of all I found the patent that includes the schematic Acmefixer showed in Reply #94. It is US4734658 (Low voltage driven oscillator circuit), see here:  http://www.google.com/patents/US4734658   (By the way here is a very similar circuit shown much earlier: http://www.google.com/patents/US3681674 ). Both are good reading for those wishing to get a deeper insight into very low voltage converters.

Here is another patent which includes what kind of features a FET should have ideally for such low voltage converters, see this: http://www.google.com/patents/US4015609 
So for a FET the drain-source ON resistance should approach zero, of course and the gate-source cut-off voltage (or threshold voltage for MOSFETs) should be also near to zero (so that a tiny mV amount of change in the gate-source voltage should cut the drain current off). In the patent you can see an attempt to improve BOTH characteristics of the off the shelf JFETs of that era.
To reduce the ON resistance, the inventor also paralleled some FETs. And to reduce the cut-off voltage, the inventor connected two such paralleled FET groups in series: the joined source pins of the paralleled group A is connected to the joined drain pins of the paralleled group B, while all the gates in groups A and B are joined to be a single gate point, the main drain pin is the drains from group A and the main source is the sources from group B. This way the resulting cut-off voltage is reduced below that of any such single JFET. (Of course the series connection of the groups increases the ON resistance BUT the advantage received from the lower cut-off voltage is more beneficial from the input DC supply voltage level point of view.)

So magpwr's finding the 2SK170 type and paralleling them was the correct first step to reduce the ON resistance, hence to be able to reduce the working supply voltage. And luckily, this type is characterised as having high transconductance (22mS) at 3mA drain current i.e. it has high enough gain to start oscillation at a low supply voltage and this transconductance is enhanced each time your paralleling the FETs. AND the cut-off gate-source voltage for this type is between  -0.2 to -1.5V only, the  -0.2V goes with the GR suffix and the  -1.5V goes with the V suffix, while the BR suffix is somewhere inbetween.

Now we know why the higher current switching type JFETs like J105 etc may not work  from a lower supply voltage than the 2SK170: because these JFET types inherently have -5 to -10V (or higher) gate source cut-off voltage ranges and this is a drawback in this respect. (The higher cutoff gate voltage inherently decreases transconductance.)

The following correct step was to turn to the ALD110800 and ALD110900 MOSFET types (from TinselKoala link) as these are advertised as zero-voltage-threshold MOSFETs. However their big drawback is the 104 kOHm drain-source ON resistance at the zero gate-source voltage (this huge drain-source resistance goes down to 500 Ohm when the gate-source voltage is increased to +4V from zero).  So it seems this zero voltage threshold feature cannot bring advantage in such low supply voltage converters. Of course these new MOSFET types can work very nicely in circuits where the high ON resistance is not neccessarily a drawback and can be matched.

Then I found  a patent which shows an interesting solution to reduce the JFETs relatively high ON resistance so that overall efficiency may be increased.  By the way, even if the JFETs are paralleled, they still can have rather high ON resistance, just consider the 2SK170BR, it surely have a drain-source ON resistance between 800 to 1600 Ohm at zero gate-source bias and if you parallel 4 of them, this range reduces to between  200 and 400 Ohm. So compare this to the primary coil's self impedance which is surely well under 1 Ohm at DC and probably represents a few times ten Ohm at the oscillator's frequency, it is a transformed impedance from the gate side of the fet, depends on the primary/secondary turns ratio too.)

While some circuits use bipolar transistors connected in parallel (as a second switch) with the jfet or MOSFET (this is discussed here: http://www.google.com/patents/US4322724 ) but here in this patent a MOSFET is used in parallel with the jfet, see here Fig.3: http://www.google.com/patents/US7170762 

All in all, what I deduce from all these patents info, including the work of magpwr, acmefixer, conradelektro and scratchrobot,  I think you need to use at least 1:80 to 1:100 or even higher turns ratio, you may need to use resonance at the secondary coil of the transformer (see the tuning capacitors at the secondary coil: http://www.linear.com/product/LTC3108   and you can combine either low saturation bipolar transistors or MOSFEts in parallel with the 2SK170 jfet as a second switch, controlling this latter together with the jfet from the up-transformed input voltage. If you have a logic level MOSFET that can fully conduct at say 1V gate-source voltage, and you have (say) 1:100 turns ratio, then 10mV across the primary coil will insure operation of the MOSFET in theory BUT to have 10mV voltage drop across the primary coil, the voltage drop across your jfet defines input sensitivity because the full input voltage gets divided between the primary coil and the jfet's ON resistance. So paralleling the 2SK170 jfets is still needed I am afraid to self-start the oscillator from as low supply voltage as possible. Can you all agree with this?

rgds,  Gyula

[conradelektro]

@gyulasun

You did a very useful research. I will try to understand your conclusions by going through the many references you posted. With my bigger toroids I can do the 1:100 reatio (e.g. 5:500) with the 0.22 mm enamelled wire. Thinner wire is just to awkward to do by hand.

Just a hint, here you can look up every patent you wish to see:

http://worldwide.espacenet.com/singleLineSearch?locale=en_EP

http://worldwide.espacenet.com/advancedSearch?locale=en_EP

What I want to do, is to make a LED glow in the garden with a copper and an aluminium tube in the ground (as a basic galvanic battery). The copper and aluminium tube in the ground show a Voltage of about 400 mV, but the Voltage brakes down below 50 mV when connecting magpwr's circuit. May be at 10 mV, the tubes can supply enough current to reach the 100 µW necessary to make LED glow.

Magnesium and copper might work, but I do not like magnesium, too expensive and all sorts of side effects. I still have to do a test with a galvanised steel tube and a copper tube, but at the moment everything is frozen solid in the garden under 0.2 meters of snow.

That's why I play with snow and a Peltier Element (see the attached photos). The little saucer is tin, the big mortar (cup) underneath is brass.

Greetings, Conrad

[gyulasun]

...
With my bigger toroids I can do the 1:100 reatio (e.g. 5:500) with the 0.22 mm enamelled wire. Thinner wire is just to awkward to do by hand.

...

Hi Conrad,

Thanks and may I suggest to make taps on the secondary coil at  1:50 and 1:75  turns ratio to make matching more flexible. Also, you could consider using ferrite pot cores or normal ferrite E cores instead of the toroid core shape, these latter cores are much easier to wind.

rgds, Gyula

[magpwr]

hi conradelektro,

In theory maybe moving the rods closer to each other would reduce the resistance or internal- resistance(In the case of battery) for your garden battery.

It reminded me of a crazy experiment done around 15 years ago using zinc/carbon battery case after removing all the black stuff in battery.
I actually stuffed fertilizer in it which did not work well.Then i added little water with fertilizer to make it moist and throw mix back into battery case.
I then charged with around 7.5v to battery for 2min.I then placed the battery into a mechanical movement clock design to work on 1xAA battery.
The clock actually worked for few days then "fertilizer AA battery" died.

hi gyulasum,

Thanks for providing some interesting links.It does help us to promote creativity in each of us.I never knew people actually patent ultra low voltage voltage booster.
Good thing mine is unique and i will never patent my design especially for simple "electronics".It would be a waste of effort or time maybe money to patent.

---------


I would be planning to work on this project to compliment my 25mV joule thief.It's using radio wave to charge battery.I will be using it with ultra-caps.
http://www.youtube.com/watch?v=tjxldE8JsDw

I was active in this forum where there was discussion on the "RF diode charger" and etc.
http://laserhacker.com/forum/index.php?topic=104.0

I noticed it's very difficult to find antique items in my country like 1N34A that i have to order online from ebay.
I had problems even finding small copper tubes normally used in Air-con smaller version for copper tube\toroid transformer to experiment on resonance as i mentioned earlier in this forum.