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

[gyulasun]

Hi magpwr,

Here is quote from the 2SK170 data sheet:  High |Yfs|: |Yfs| = 22 mS (typ.) (VDS = 10 V, VGS = 0, IDSS = 3 mA)

and I think the key is: the IDSS is mentioned to be 3mA where the device has this high transfer admittance Yfs=22mS. I have not seen any JFET type for which as high Yfs as this would have been specified at such a low drain current as 3mA.
I also deduced the same conclusion from the data sheets of BF862 or J105 etc jfet types that they may be better candidates but your practical test for the J105 proved just the opposite, and I wonder if you have tested the BF862 too?  Because in the Design Idea in which the BF862 was shown to work from 300mV DC supply the schematic is different from your 2SK170 circuit so correct comparison can only be done by using the same circuits i.e. yours with the BF862 instead of the 2SK170. IF you have already done this, then sorry for this rant.

rgds, Gyula

Hi gyulasum,


This is a interesting finding which shows current handling capability 3 different version of  2SK170.

Just noticed the 10 pieces of (2SK170 BL  6~12mA version) which i purchased from www.futurlec.com for $0.20 ea .It's found through search "2SK170" for that site.

It's unknown if slightly higher current handling capability would have impact on "input voltage  detect sensitivity".

Unknown if the 2SK version is similar like those for transistor eg:BC547A,BC547B,BC547C which have different gains.

I noticed that base on datasheet BF862 and J105 does have better RDS/Lower on resistance than 2SK170.
But unfortunately  the startup voltage for BF862 and J105 is not really impressive at around 300mV.There maybe relationship between current capability and voltage\current detect for JFET base on current technology.

To increase current output\input at low voltage it would be more sensible to use transistor EG:ZTX1048A with 2SK170.

I have tested the J105 this afternoon which was disappointing which could not start at 0.1v but does start oscillation at 0.3.. 0.4v

[acmefixer]

I also believe it's a patent, at least it looks like ones I've seen before.  I believe it was posted in a forum, may even have been this one, and I saved it.  That's all I know.  The point I wanted to make was that in the patent schematic, they drive the gate with the same winding that is used for the output.

Thanks you for reading up on my schematics and blog.  I'm glad that someone has found it useful.  With a pot and/or resistors totaling 1 meg or more across it, the button cell's lifetime is in years, so there's no problem with it being used. I even used a 0.1 uF plastic capacitor connected to the gate, and just charged it up by touching a 1.5V cell to it.  It stayed charged for many hours, so I could use an even larger value and charge it up and it would stay charged for days or weeks, and still do the job of biasing the gate.

Today I have been experimenting with the J105 and I was able to get it to oscillate at 60 millivolts.  But I ran out of space on the core and couldn't wind any more wire on it..  I think if I wind another larger coil with several hundred turns, it should be able to go down to as low as 30 or less mV  You can read my blog about it.

I just had an idea that will save me a lot of work.  But I have to try it to see how well it works.  More about it in an update to my blog.

Best of success in your endeavors.

Hi acmefixer,

I would be interested to read the patent you show a schematic from in your attachment. Could you give the patent number?

I have read your blog, seen your schematics with the button cells to create gate bias for the TN0702.  I believe you could include in the same circuit a JFET oscillator like magpwr has shown with the 2SK170 and the only purpose for this oscillator would be to produce the 2-3V DC bias voltage.  By using a third coil next to the needed two coils for this oscillator, you could galvanically isolate the output of this oscillator so the biasing of the TN0702 (or other MOS or JFETs) should not be a problem. And then you could run the whole circuit from as low as your 80mV DC without the bias cells, and without a significant input power increase.

rgds, Gyula

[acmefixer]

This Youtube video shows one experimenter's way of doing what you're trying to do.  It may be a seventy year old low tech way, but it does the job.

I'm still experimenting with four J105 JFETs in parallel and I'm getting the LED to light at about 80 millivolts.  See my blog.

Here's another experimenter's circuit.  Lights a LED with a candle.  The circuit converts 30 to 40 millivolts.

Hi Acmefixer,

The 2SK170 absolute lowest possible voltage still remains at 25mV for oscillation to start&continue.It's not about the coil turns ratio anymore.Already tested it.
At 24mV it's no guarantee.

My latest discovery J210 N-channel JFET will work very well with ZTX1048A transistor.But lowest startup voltage is at "35mV" for JFET to kickstart transistor (Collector connected to drain,emitter&source).If ZTX transistor kick started manually on it's own lowest supported voltage is 30mV for oscillation to continue.
Secondary coil to gate of "J210" must be 330KOhms and secondary coil to ZTX1048A must be 330 Ohms at 35mV input.If resistor goes 150k or below to the gate the JFET will not work at all unless it's around 0.3volts .One more thing the J210 series JFET doesn't do oscillation on it's own at  voltage <100mV it merely produces alot of spike provided using resistor 330kOhms to it's gate...to automatic kickstart transistor.
I was surprise base on above configuration the voltage could exceed 10volts output for 35mV input(10v/0.035V =285 x voltage boost) input at around 6mA current draw ,circuit output capacitor using 100nf.Current still remains sufficient for led to be lit.

2SK170 and ZTX transistor combination still produce the best result.

[magpwr]

Hi Acmefixer,

Interesting circuits links.Noticed one of the circuit is using 10xBF245C (JFET) but using candle which produce higher temperature to light led.

I have done testing on BF245 and many other JFET series which have posted in forum base on startup voltage but noticed the for current handling 2SK series JFET is still the better choice.

It's interesting searching for a interesting article on circuits which is good and yet hard to find.It's good i  posted this circuit to share so that someone may come up with better version or design and we improve it from there.

[acmefixer]

I'm curious as to why you choose 2SK170 over others.  Toshiba says it is low noise, designed for audio preamplifiers, which is far from being used as a switch.  There are JFETs that are designed for switching, which is what the Joule Thief is really doing.  What do they do that's better and makes you choose the 2SK170 over those?

Thanks.

Hi Acmefixer,

Interesting circuits links.Noticed one of the circuit is using 10xBF245C (JFET) but using candle which produce higher temperature to light led.

I have done testing on BF245 and many other JFET series which have posted in forum base on startup voltage but noticed the for current handling 2SK series JFET is still the better choice.

It's interesting searching for a interesting article on circuits which is good and yet hard to find.It's good i  posted this circuit to share so that someone may come up with better version or design and we improve it from there.