Driving LEDs with a battery pack

I was just re-reading the original web page and it says "Don't use anything other than plain cheap vanilla ferrite, since any other type will reduce efficiency."

I just ordered a bunch of them from him - the specs are way better then the ones I've got here.

I hadn't read this, but I think I've at least alluded to it in posts on this thread. The problem, if I got it right, is energy storage. Ferrites have little tiny air gaps (okay, not air, but about the same in effect) between granules with iron in them. The energy gets stored there, in the gaps, for release when the BJT turns off. Straight iron cores won't store anything much. So bad news. The other desirable effect, though, is the voltage aiding impact of the base winding. For that, iron core would be great. But there isn't that much need for power transfer to the base, just voltage mostly, so ferrite works great here for that reason AND does the main job of storing energy in the gaps. If I got it right, of course.

By the way, now you can see just how tiny it can be!

Jon

Ok, I made the "1mm" up but they're only going to be as long as the thickness of the PCB.

I'm looking at one I __just__ removed (I've got two TV set boards siting in a box waiting for harvesting.) It's long enough to solder to. And it is free!

Jon

That's the page you get if you type "joule thief" into google.

So really he's not saying "use ferrite", he's saying "don't use iron"?

I just found something about the circuit which is confusing the hell out of me. I mentioned it yesterday but nobody spotted the problem...

If I measure the current going through the LEDs it says

I assume it's not really 102mA because the LEDs aren't dead, the signal is flat DC thanks to the capacitor/diode so what's going on?

On Sat, 11 Jul 2009 12:59:42 -0700 (PDT), fungus wrote:

I think I'd read it some time back, as well.

I guess so. Iron switches alignments really easy (low reluctance.) Since it is easy to do, no energy gets used to do it and there's no energy there to be had later on. What it is really good for is passing along those flux changes rigidly so that power is transferred without losing much along the way. They make GREAT transformers -- especially at lower frequencies before their electrical conductance (low resistance along with low reluctance) allows eddy currents to become a problem. They "transfer" power efficiently. However, while you DO want a voltage to show up on the base winding and while iron would do that for you really well, it doesn't _also_ store energy for you. So when the BJT turns off and the collector winding is supposed to be delivering energy to your LEDs from all that "stored energy" somewhere... there won't be any because the iron didn't store it for you. It just tried to transfer it, had the base winding create a reverse flux change to oppose it, a tiny bit of power got transferred to the base in order to add some current there (but not much), and nothing got packed away because iron doesn't store up anything. On the other hand, ferrite has these air-like gaps in it and while it isn't as good as iron at transferring power to the base winding, you don't need it to be that good. What you need is some nooks and crannies to store up energy and the ferrite will do that for you. Also, because there isn't much of an electrical connection between granules with iron in them in the ferrite (isolated by the 'glue' so to speak), eddy currents can't really flow much so they are really good at higher frequencies where that could become a problem. For this situation, ferrite is a very nice fit on every score.

It's probably right. (Unless you are averaging.) You should be seeing high peak currents in there. But only for a short part of a total duty cycle. Are you averaging?

I think it really _is_ 102mA, but only for part of the time. That allows the LEDs to cool down in between times and your eyes perceive the average value not the peak -- at these rates, anyway.

Jon

My recollection of messing with the joule thief a few years ago was that it is transistor saturation, not core saturation, and that Vbatt is the key. In essence, the base wants to drive the transistor harder, the transistor is capable, but Vbatt prevents any further increase in Ic.

BTW - I made one using an air core (we can eliminate core saturation from that one), and several using ferrite cores where core sat was perhaps possible. But again, my recollection is that it was transistor saturation due to the low battery voltage. I suppose that may be too simplistic - you could get into the Vdrop within the battery's internal resistance (Rbatt) and Rbatt limiting the current through the coil ...

Ed

Which is ALL I've been saying here. That's why I disagree with greg on this.

Actually, Ic rises until Ic/Ib exceeds the BJT beta at that Ic.

If you read some of my longer postings here (it seems you have not), you will see a VERY DETAILED description. Not to mention the equations I've provided. So I have a pretty thorough understanding I'm applying here.

Jon

Do you remember how many windings, roughly, you used then?

Jon

Uhhh - if your LEDs are in series, that is 106mA through each of them. In the unlikely case they are in parallel, you have to measure the current through each one individually - they do not necessarily conduct the same amount of current. And if they need to be in parallel, each should have its own current limiting resistor (or some other means of controlling the current).

That aside, I'm glad to hear you got good results with your joule thief!

Ed

What about "iron powder" beads. All the ones in the PC power supply look like these:

FWIW It will work with an air core ! I tried parallel bifilar turns on a small (5mm) tube ! It oscillated, but only without a load. A pile wound bifilar pair worked much better. Putting in a ferrite core vastly improved the output.

This page seems to explain the difference but I can't decide if it's better or worse for a joule thief.

The colour is just paint !

Arn't they just ! Very very useful. Try passing coins between a pair. You might note something interesting !

I'd stay away from iron powder and stay with ferrite ('iron oxide.') Not because I know why, though. I'd like to hear someone who knows this stuff talk about the two options in this case. I'd learn something.

My mind thinks "um, how much effective air gap is there in iron powder cores?" But the permeability seems lower than ferrite and they are often used in very high frequency cases. So that makes me think the granules must be isolated and that there must be significant energy storage available in them. But I just plain don't know.

So you really do raise a good point for me. I feel unable to give any useful thoughts on this point. I'm still learning, too.

Jon

Here's a link:

It seems to pretty much say that powdered iron is fine, as I read it.

Jon

ke-design.pdf

I read it and it meant nothing... :-S

But ... there's a line on the second page which seems to say it doesn't make much difference between 25kHz and 75kHz. We're aiming at 50kHz so I guess we're OK.

I've got three of them so I hope so....

Ummm, I'm using a digital multimeter.

Does it mean the voltage is staying rock steady but the current is going crazy?