Is There Honor Among Joule Thieves?

Flashlight Modding and DIY Parts
1

Or should it be Joule Thiefs?

I would love to find others here working on the Joule Thief circuit!

There are some sites on the Internet with some pretty extravagant claims about various magic; but I’d suspect the S/N ratio is better here.

What I’d really hope for most is the sort of discussion BLF is famous for. This works, it’s not magic, it can be modeled, and there is a horde of information that is not being touched about WHY it works and how. Contrary to popular belief, they do respond to tuning, even as rudimentary as what I can bring to bear.

I have made a few, not all successful; and can testify to the efficacy of the circuit. I have driven a single, MODERN (IDK which, maybe a “P-something”?) LED from a 3*AAA “MDXL” torch from eBay (now able to run on 1 “dead” AAA).

I still have this “belief” that I can push 1A through an XM-L LED with one. It’s a hobby, what can I say? But I want a tiny, “ball-of-light” tasklight, much smaller than anything I can find, even if it means sculpting the body out of Thermal Epoxy. I’ll show you all why when I make one.

Yes, I’ve been playing with a Circuit Simulator, but I try to double-check by dragging out a decades-old Circuit Analysis textbook. Plus when I Sim the actual circuits I made, the results seem to correspond (within the limits of the Sim). I don’t have the ability (skills, yes/tools, no) to measure the inductors, nor the spiky HF LED drive current. My Fluke 77 doesn’t seem to be able to measure the frequency — or it’s not nearly what I expected, which leads me to suspect the rest of its ‘suggestions’. This makes it very difficult to estimate, plan & test results. It’s not like I can afford to just blow LEDs if I make a mistake. Sorry, ‘WHEN’…

Is there anyone here currently working on a Joule Thief? I’d be happy to wind Toroids & pay postage, in exchange for inductance measurements. (Within reason, of course) That’s cheaper than an O-scope, and I can take it from there. The subject of Toroids and Inductance don’t get much discussion, but I think it might be possible to use a Joule Thief as an inductance meter… Still working on that one…

Meanwhile, I may be singing in the rain here, so I’ll come back after awhile and delete this if no one wants to talk about it (“ignore it & it’ll go away”). Or jump (if pointed) to another thread here (not offsite — I already found those) which I am having trouble finding.

Thanks for reading this far!

Dim

2

Check out the work of wquiles on candlepowerforums.com (He’s around here too, but I can’t remember as much info about the JT being written here.)

He’s a master machinist and seems to know something about electronics to boot.

3

I’ve found it hard to get people charged up about higher levels of investigation. This is not to say that anyone is rude about it and there are members who can and may help and if you keep plugging away and post your results you will get some encouragement but what you are interested in is well outside both the training and the interest of almost every one here. By all means keep posting here as your results have direct implications but a true electronics forum is likely where you will get the highest volume of helpfull feedback. Don’t give up. Positive results breeds more interest(success breeds success).

4

Good luck with this Dim. I am interested and will be following your thread, but unfortunately all I can offer is encouragement.

5

Hey Dimbo,

I've been playing around with them and collected my experiences her, in case you missed that thread:
https://budgetlightforum.com/t/-/11469

Unfortunately I have no advanced knowledge in electronics, what I do isn't much more than playing around.

One thing that still startles me: If I measure the output voltage of a Joule Thief across the LED, my DMM returns me only the input voltage of the battery, and yet the LED lights up. I guess it has to do with the switching frequency of the transistor.

6

My Fluke 77 is little better, at the frequencies I seem to be generating. It’s hard not having an O-Scope!!!

I add a Capacitor across the “primary” side (the turns going to the Base) in order to increase the ESR (see ‘LC Circuit Resonance’) and lower the Base Resistor. (Ib can’t exceed 20mA on my scrounged transistors, although a 2N2222 can take 200mA, if you can fit a TO-18 can). As opposed to just loading up R1, this keeps the current moving in the Toroid as well as helping protect the transistor. The hotter your transistor gets, the sooner your battery will die.

I guess what I’m trying to say is you are correct to refer to the transistor as a switch, not a booster! It hardly needs to be bothered, except to turn on and off. Tuning the frequency of all this switching is also important, albeit hard to measure.

The hard part is knowing the H of the inductor!!! The formula is quite simple then. You can “tweak and try” with a trimmer cap, but that can take a lifetime. Yes, that’s mostly Sim work for now, but I wouldn’t use the Sim until I tested it against known-value components (within reason, of course).

Thanks for pitching in! I’ve left a note on the other thread & will be keeping up with both now.

Dim

7

We concur!

BUT!

There’s YOU, and there’s ME. That’s two more than it took to invent a working telephone, which led to this Internet from which you and I now find profit and pleasure. Seems pretty encouraging to me!

Dim

8

Have at it! :beer:

9

I tried to build one, but that was just looking up the schematics, soldering it together, hooking it up to an AA,
seeing that it doesn’t work. I didn’t investigate why that was.
If it is a hobby, why don’t you try to get some cheap old oscilloscope from ebay?

Anyway it’s nice seeing that you really put some thought & research into this project,
i always love a scientific approach :slight_smile:

Good luck! I’d offer to measure your toroids, but my spare time is very limited and you’d have to ship to europe.
Also i think i would need a signal generator for that.

10

Like I posted in Steve_the_Chief’s thread, I’ve played with the circuit a bit. Here’s what I found:

- Seems like a minimal of 5-6 windings on the toroid for the circuit to work, probably has to do with the inductance (didn’t measure).

- More winding on the feedback coil lets the circuit work down to a lower voltage, but startup voltage is the same.

- Toroid size doesn’t matter, though I won’t be surprised if I’ve hit the saturation current on some of my smaller coils (~5mm diameter toroids)

- Jfet joule thiefs do work, but are inefficient. They do work to a very low voltage (I’ve had one work under 200 mV, could be lower if I had even more windings)

- Can’t get too much current from the joule thief. Probably has something to do with discontinuous mode operation. Haven’t tried the super joule thief.

  • BJT does matter, I like the 2N3904.

I should probably check things with an oscilliscope and power with a power supply, but I’ve been too lazy to do that.

Also, I have’t tried any of the efficiency boosting methods, haven’t gotten a MOSFET joule thief to work, and would like to figure out how to make the jt into a synchronous boost driver.

Been also trying to figure out how to make a really low power joule thief that isn’t too particular on input voltage. Maybe a diode in reverse bias would work instead of a resistor.

I did make a tiny light that fits into any AAA light. It’s pure flood with a small SMD white LED and uses a single LR41/AG3 battery. Should probably grab a few images and post them up.

Quantstuff has some pretty interesting notes on the jt circuit.

Sooo, that’s what I have.

11

As a note, this utility could be helpful (I haven’t used it myself)

http://zelscope.com/

(2 week free trial, there should be freeware versions of something similar)

12

Did someone ever try the version with the third winding like this one here?

http://djzeba.altervista.org/joule%20thief.jpeg

13

I’ve only played with LEDs, so I haven’t tried that before.

Look interesting though.

14

Over the years I’ve looked at DC:DC converter circuits a few times, which is what you are describing. It sounds like you are leaning to discrete transistors instead of ICs. This might allow you a lower input voltage. Have you considered using prewound coils like Mouser sells?

I haven’t done much prototyping with these but would be interested in helping. I do have access to a scope. My background includes an electrical engineering degree, moderate prototyping and pcb layout.

15

I have tried some prewound coils and they seem to work well. I’d like to find some with thicker wires for the main coil, and thin wires for the feedback coil though. Thicker for lower DC resistance, thinner for more windings and lower voltage operation after startup.

Yeah, the jt is a blocking oscillator dc/dc converter and is pretty simple to make. Not the most efficient but runs down to a low voltage. The lowest I’ve seen for a boost IC is 0.7-0.8v for startup/shutdown voltage and I’ve made jts start at ~0.55v. The ICs are usually pretty efficient though, usually have a higher operation frequency (smaller inductor), and can require far less work (just solder, not winding of inductors).

I have taken basic electrical engineering classes (I’m a BME) and like to experiment on the breadboard or LTSpice on occasion.

There are some ICs for thermal generators that work down to 20 mV, but those top out at 500 mV and are pricey.

16

Would it make sense to buy a prewound inductor with the desired main coil inductance, then add the feedback turns by hand? You should be able to roughly calculate the number of turns on the original coil w/o taking it apart.

The lowest starting voltages that I’ve seen with boost converter ICs is also about 0.8 v. I checked the LM3909 LED flasher, but it needs 1.0 v. The LMC555 needs 1.5 v.

I just looked at a few jt circuits online and can see why its hard to improve on their simplicity and size!

17

I haven’t tried adding windings, but I do have a spool of 38 gauge magnet wire and some 40-42 gauge stuff salvaged from a motor. I’ll try that when I get time and see how well that works. The inductor with decently high inductance rating and small size have higher DC resistance than I’d like though (~50-100 µH), and they already use pretty thin wires to reach that impedance.

I did try a jt with a 1 mH coupled inductor, and it produced a very audible wine when it was working. As expected, the winding got higher pitch as the battery drained.

18

Have you tried a framistan set on discombobulator mode with the UV-noise generator level pushed to ludicrous? :bigsmile:

Seriously…I don’t understand a thing y’all are saying…but somehow…I’m intrigued. :slight_smile:

19

38 ga copper looks to have 0.66 ohm/ft. A few feet at 25 mA would drop 0.033 volt which isn’t too bad, but gets multiplied since the duty cycle is less than 1. I would think that this would be tolerable, but definitely contributing to losses.

Even that 38 ga must be kind of delicate to wind at 0.47 lb breaking strength:

20

I was going to use the thin wires for the inductor going to the base of the BJT, so there shouldn’t be too much current going through it, making the resistance less important. The wires going to the LED will be 32 gauge or better to decrease resistant wile still keeping it small.

Not too many issues yet with winding the things, only snapped one wire, and that was because it had a kink in it (~40 gauge).

kronological, I completely forgot about that! Would you also recommend setting the Crankafier with the demodulater attachment to Semi-transient? :stuck_out_tongue: