Colin's Joule Thief kicks major butt!

[carbonzit]

Just wanted to report that I've adapted one of Colin's (colin55 here) Joule Thief circuits to drive multiple LEDs, and it continues to amaze me how well it works.

I'm using his "LED torch circuit C" from this page on his "Talking Electronics" site. I've built quite a few of these, driving either one or two "high-brightness" LEDs (28,000 mcd, 20mA bright white lights). But just for laughs, I hooked it up to a 3-volt supply and started loading on LEDs in series to see what it would do. It ended up driving 6 in series, no problem, nice and bright! (That's almost 20 volts from this primitive blocking oscillator boost converter.)

The only change I made from his design was to change the resistor to 1K.

My hat is off to Colin once again.

 

[Sceadwian]

It's not his circuit.
Here's the Wikipedia entry for some history.
The efficiency is still well bellow even the simplest SMPS supplies. Not useful for much but dragging the last dredges of useful energy out of a battery and as a novelty.

 

[BrownOut]

The joule theif is a simple SMPS, and can be very efficient. It's useful for driving LED's from very low voltage batteries, which may not have sufficient voltage to forward bias the LED's alone.

 

[Sceadwian]

Sorry BrownOut, I was referring to IC based SMPS supplies. Given the capacities of the low voltage batteries are already limited for a given size and weight using a Lithium cell is cheaper and more efficient, otherwise we wouldn't see so many coin fob based LED lights that use nothing more than a Lithium cell and an LED. Use the right device for the right situation. I've seen some novel uses where Joule thieves are actually built inline in a standard incandescent bulb package, but nowdays, it's nothing more than a novelty. You can buy better for less that will last longer.

 

[BrownOut]

Even small lithum cells don't have enough output voltage for some LED's. It totally depends on the application. Certain LED's can have turn voltages of 3.6-3.8V, while Li cells have nominal voltages between 3.2V and 3.7V. Many products use small transistor based SMPS, similar to the joule theif, to effeciently step up cell votage for bright LED operation. Other applications use simple boost SMPS's for even higher operation voltages.

I have a small camping light that uses 4 white LED's (3.6V forward voltage) on a single 1.5V battery. Been using it for about 4 years now and still haven't changed the battery.

 

[carbonzit]

It's not his circuit.
Here's the Wikipedia entry for some history.

Of course he didn't invent the concept, I know that. But this particular implementation is, in fact, his creation.

The efficiency is still well bellow even the simplest SMPS supplies. Not useful for much but dragging the last dredges of useful energy out of a battery and as a novelty.

Read his page to see how his super-simple (1 transistor, 1 resistor, 1 capacitor, 1 hand-wound transformer) stacks up against some commercial implementations (chips). His circuit certainly gives them a run for their money, with respectable efficiency.

Besides, it's all about the DIY-ing.

 

[ghostman11]

Atomsoft came up with a nice implementation as well (see here https://www.electro-tech-online.com/threads/custom-boost-circuit.110264/) i tried jason's and it works great! i particularly like the smd inductor he used.
as for colin's i only have one question..... would the performance be better (even marginaly) had the windings been neatly done? or dosnt this affect the magnetic field in anyway?

 

[Sceadwian]

For the same number of turns if the windings are neat and tight together that means the magetic flux is denser, haphazardly wound coils have more air in between each turn meaning more distance for the magnetic field to be weaker (less inductance per turn) but it'd be hard to measure for as much as you need for a Joule Thief. Hardly worth considering, though unless space was at a hyper premium. That's not really the point with a Joule thief, it's doing what you cna with what you have, if you happen to have the SMD inductor around great if not hand winding is fine. Although I doubt it would have any measureable effect on performance a losely wound coil would have lower capacitance than a tightly wound one.

 

[ghostman11]

thanks for that Sceadwian it's kind of what i assumed. but always nice to get a good explanation

 

[carbonzit]

Atomsoft came up with a nice implementation as well (see here https://www.electro-tech-online.com/threads/custom-boost-circuit.110264/)

I think that implementation is waaaaaaay overkill for lighting a few lousy LEDs. I mean, using a PIC? Give me a break.

Look at how simple Colin's JT circuit is by comparison. It's self-oscillating, has only 5 components including the LED, and probably works about as well as the PIC-based boost converter does. And I just proved that by running it at 3 volts, I can easily light 6 high-brightness white LEDs in series at full brightness.

But I guess if you really like PIC programming ...

By the way, regarding your question about coil-winding neatness, I don't actually know how much difference it makes (I suspect not that much), but I strive to wind my JT transformers as neatly and tightly as possible. For what it's worth, they've all worked great from the get-go.

I have a collection of broken ferrite pieces that I use for the cores, scavenged from various pieces of electronic equipment left on the curb in my neighborhood. I cut small pieces, about 1/8" square by about 1/2" long, smooth them down with a file, then wind the transformers on them, using magnet wire also scavenged. 40 turns of fine wire for the primary first, with a heavier 60-turn secondary over it, and it's done. A little epoxy keeps them from unwinding.

 

[ghostman11]

knowing jason he used leds to just prove a concept, like a great deal of what he does he does it to learn a concept or try an idea. learning for the shear pleasure of learning for a guy with no formal training he has aquaired a huge amount of knowledge that he's always happy to pass on. as for a pic being overkill......... i guess that depends on the situation, i came accross the thread i referenced because it was sugested i take a look when i posted regarding a rs231 related question. i was using a pic for serial communication but the chip i was using for level shifting didnt have an internal boost and i needed a 12v supply. i didnt much fancy adding a power supply so hence why jason's was mentioned to me.
so in the situation i was in colins would have been the wrong choice.as for the question on windindings it was a curiosity question, my thoughts were if it relies on magnetic flux then surely efficiency is affected if the field isnt 'aligned' not sure alinged is correct phrase but i am sure you get the picture, again i was thinking of size and how much smaller colins untidy coil could be made if it was wound neatly. i might have a go at trying it out by doing a 'scruffy' one with x numbers of coils and one with same number of coils but very neatly wound, then i guess it should be as simple as measureing the uH with a meter to determind what diff it makes. i will deff keep a look out for ferite rods when i take stuff apart as they are a handy circuit to have. as for magnectic wire i have no idea what this is, i had assumed it was just laquered copper wire from something like a old transformer.
anyway time to visit the workshop and have a play around with it
jason

 

[carbonzit]

Well, keep in mine that Colin's (and others') Joule Thieves are totally unregulated (and very spiky!) power supplies, pretty much unsuitable for powering PICs or anything much besides LEDs. Although Colin does have some simple boost power supply circuits on his site that he uses to power PICs. They're a little more complicated than the JT, but not much (two transistors).

But they're a lot of fun to experiment around with.

Yes, "magnet wire" is the generic term for any varnished/lacquered wire used in transformers, inductors, solenoids, motors and other electromagnetic devices. There are tons of it out there for the scavenging.

 

[acmefixer]

Jt ps

A JT can be use for a power supply by replacing the LED with a 1N4148 diode to rectify the pulses, and a filter capacitor of 100 uF. If the voltage must be regulated 5V, a 1N5231B 5.1V zener can be put across the filter cap. It is a simple circuit, but this wastes a lot of power in the zener.

I've **broken link removed** about what JT is, and what it can and can't do.

 

[ghostman11]

interesting read thanks acmefixer, nice blog

 

[acmefixer]

Schematic

I would not use D1 or C1, I would connect the LEDs directly across the MOSFET with nothing else. It is not really necessary to rectify and filter the output of the MOSFET. If you really need to use D1, then it should be a 1N5819 and C1 should be a tantalum, or else it should be a 100uF low ESR electrolytic. The 2N7000 has a resistance of several ohms and it may get warm from the current it has to handle. I would use a BD135 or similar regular BJT, or else a logic level power MOSFET.

You really don't need the boost circuit. You could put all of the LEDs in parallel and use a 100 ohm current limiting resistor for each LED. It's a lot simpler and it would not be that much less efficient than a JT type circuit.

 

[carbonzit]

I've **broken link removed** about what JT is, and what it can and can't do.

Nice. Can't have too many web pages out there about JTs. Couple comments:

1. Your site unfortunately doesn't lend itself to actually finding stuff within it. Any chance of adding some index pages, or some navigation aids, so one can find posts by topic? As it is, the only links are to archives by date.

2. Heh; I like your little tag line: "My favorite programming language is solder."

By the way, regarding DIY transformers used in JTs, I notice that you seem to favor toroidal windings. I'd just like to point out that all my JTs use a simple transformer wound on a single small bar of ferrite, as I described above. The bar is about 1/8" square by about 3/8 to 1/2" long, with the primary wound first, the secondary wound on top of it. They seem to work as well as any toroid, and are a lot easier to wind (no threading wire in and out of the doughnut). Just a note that there are several ways to skin this particular cat.

 

[BrownOut]

Toroids typically make for better inductors.

 

[carbonzit]

Because they have a closed magnetic circuit, right?

Any real difference in efficiency?

 

[acmefixer]

@carbonzit

Nice. Can't have too many web pages out there about JTs. Couple comments:

1. Your site unfortunately doesn't lend itself to actually finding stuff within it. Any chance of adding some index pages, or some navigation aids, so one can find posts by topic? As it is, the only links are to archives by date.​

Google is your friend. If you google watsonseblog and then the joule thief topic you're looking for, google will generally sort out the blogs and give a line or two from the beginning, which hopefully will give you what you're looking for.

2. Heh; I like your little tag line: "My favorite programming language is solder."​

Yeah, I stole that from Steve Ciarcia.

By the way, regarding DIY transformers used in JTs, I notice that you seem to favor toroidal windings. I'd just like to point out that all my JTs use a simple transformer wound on a single small bar of ferrite, as I described above. The bar is about 1/8" square by about 3/8 to 1/2" long, with the primary wound first, the secondary wound on top of it. They seem to work as well as any toroid, and are a lot easier to wind (no threading wire in and out of the doughnut). Just a note that there are several ways to skin this particular cat.
​

Yeah, in my blog you'll find some JTs made with air core coils, ferrite bars and bobbins, etc. The Joule Thief started out with the pioneers using toroids, but there is nothing that says you have to use them. The 'right' toroid has the big advantage of a high permeability, which means you can get enough inductance with as few as 10 or a dozen turns. Winding one is easier than a bobbin. but the wrong toroid, a low permeability one, will require a lot more turns and a lot more time winding it. Of course the same goes for ferrite bars and bobbins. I took some cylindrical bars from the board of a dead PC power supply, the ones that have a few turns of very heavy wire wrapped around it. I unwound the heavy wire and wound 28 AWG magnet wire on it, and measured it with my LC meter. The inductance was still only a few microhenrys, so those bars don't make a good coil for a JT.

 

[carbonzit]

Nice. Can't have too many web pages out there about JTs. Couple comments:

1. Your site unfortunately doesn't lend itself to actually finding stuff within it. Any chance of adding some index pages, or some navigation aids, so one can find posts by topic? As it is, the only links are to archives by date.

Google is your friend. If you google watsonseblog and then the joule thief topic you're looking for, google will generally sort out the blogs and give a line or two from the beginning, which hopefully will give you what you're looking for.

BZZZZZT! Wrong answer!

My friend, if it's your web site, and you actually want people to come there and read stuff, then it's up to you to provide adequate navigation tools to let them find that stuff. Don't expect anyone to go out of their way to unearth whatever gems of knowledge may be buried in that sequential list of archives. I certainly won't.