Driving LEDs with a battery pack

Yep, almost identical.

Yep, and even longer to discharge. I can still see a tiny bit of light ten minutes after I switch it off.

It doesn't say how many volts but it's pretty big and seems OK so far (volts are good and it hasn't exploded).

Yes it is (or very close).

If volts are constant and current is varying then the only possibility is that resistance is also changing. Makes no sense to me.

The curve on your 4 ohm resistor looks about right if there is no D8 present. So is this without D8 AND without C1?

Jon

No, wait, my bad. I just had a careful look at it and the schematic was actually this:

If I connect it up properly, like this:

Then all seems well. The current across the LEDs is about 13mA according to the multimeter and the voltage drop across R2 confirms this..

So it seems all is well...

I've just been poking around the circuit with the scope ane multimeter and all the measurements make sense now so basically it's finished.

All I need to do is add the extra diode to protect the transistor (I'll get some shottkys) then use a smaller ferrite bead and fiddle with the number of windings on it 'til the current through the LEDs is 20mA.

Wonderful!

Sounds great. I think I'll play with some of the transformers I just wired up, as well. I'm using #61 material for some of them (since I happen to have it.) I also have #73 and #75B. From some curves I have, I think the #73 might be okay. And in any case, the #75B stuff I have only has a 1mm hole in it so I pretty much can't use it, anyway. So I'll go with #61 and #73 and see how it goes.

Jon

Yeah, thanks for all the help and patience with a newbie (this is by far the most complex electronics I've ever done).

I was thinking about what's been gained by all this (127 posts to figure out how to light up a LED...?).

The Joule Thief is designed to suck juice out of "dead" batteries, but that's not what I'm trying to do at all. So ... why should we use a Joule Thief instead of a plain old resistor?

a) I did a plot of volts/current to compare a JT with POR, using every combination of new/batteries I could find (I don't have a fancy bench power supply). The results are here:

It's obvious that the JT changes the slope of the line in our favour, ie. LED brightness drops off much more slowly as battery voltage drops.

We can design a circuit which has much less risk of overloading the LEDs (the original design of "20mA @ 3.75V" will put about 35mA through the LEDs with fresh batteries with a plain resistor, with a JT the same batteries would only put about 24mA through them - much safer).

b) The JT puts the LEDs in series instead of parallel so it's much easier to deliver the correct current to every LED despite differences due to manufacturing tolerances of the LEDs/resistors.

If I undertand correctly, what you're trying to say is that in the non-saturating case, you can be sure that it will *always* oscillate, because the collector current is maxed out for whatever base current there is, and any fall in base winding voltage at that point must cause a decrease in supportable collector current.

Whereas if the inductor saturates, the remaining base current with zero volts across the base winding might still be enough to support whatever collector current there is at the time.

However, I think it should be possible to make it work in that mode. You just have to arrange things so that the DC base current is less than that required to support the collector current at which the core saturates. Then a flip-over will happen in the same way as it does in the non-saturating case.

The problem is that Ipeak is directly proportional to beta. So if you choose a low estimate of beta, your output current will be far too high, and you risk cooking your LEDs. If you choose a high estimate, the ouput current will be far too low, and you won't get much light.

Like I said, for a one-off project this is okay. But you can't really *design* this circuit to have any kind of predictable performance.

On the other hand, if it's designed to work in saturation mode, it ought to be more predictable. You may have to make some measurements on the first one to find out the saturation point of your core, but having done that, the second one you build should work fairly much the same.

But then there doesn't appear to be anything to ensure that the current flow through the LEDs is smooth. You put a burst of energy into the capacitor... and then there's nothing to stop the LEDs from taking it out again in a big slurp.

Yes, but it is not R2 resistance that changes. LEDs are non-linear devices. They don't operate as if they were a fixed value of resistance. When the voltage is above the total Vf of the LEDs, they will conduct current as if they were a relatively low value resistance. When the voltage drops below the total Vf of the LEDs, they will conduct current as if they were a very high value of resistance.

Essentially, an LED acts like a zener diode, and attempts to keep the voltage across itself at Vf by varying the current through itself. So in your circuit, if the voltage from the oscillator attempts to increase above Vf, the LEDs conduct more current, dropping the "extra" voltage in R2.

Your 15V appears flat, but actually has 250 mV ripple on it. If you set your scope to AC input and look at point A with the vertical set at .1 volt per division, you'll see about 2 and 1/2 division ripple.

Well, if I remember correctly what you said in another post, that 100 mA figure is wrong. AIRC you said that supplying the LEDs with a steadfy 20 mA DC gave a brighter glow than you get from the joule thief. That means that the average current has to be lower than 20 mA.

So - something is missing or not quite right.

Ed

Yes, but as long as other things are equal, I'd expect a higher permeability core to give you a higher inductance, as long as the core doesn't saturate.

Maybe saturation is the issue, though? That is, although the inductance is higher, the current can't go very high before the core saturates, dropping the inductance to zero.

I'm studying the references you gave me now. They do seem to say fairly explicitly that the energy is stored in the gaps. I'll have to think about that some more.

It can't be as simple as "gap good, iron bad", because then the more gap you have, the more energy you could store, so you'd be best off with no core at all. That doesn't appear to be the case, so there must be some opposing effect at work.

That sounds like you follow what I wrote.

That's why I think saturation isn't good.

Yes, you are probably right. And that may provide a little bit of additional margin for operation even in the face of some saturation. But it's probably better to design for no saturation and expect that you've got some margin in case you didn't hit it perfectly.

Actually, this isn't that much of a problem. On first blush, it may seem that way. But the part you aren't including is the iterative process I mentioned, using the datasheet. I go through the equations several times and in doing so the iterative process pretty much always nails where it will hit. The assumed beta (and actual) has somewhat less impact because of that. It's not quite so "open loop" as it seems. And it appears to work reasonably well. (Certainly, though, using a different transistor with a markedly different beta performance will clearly change the Ipeak. It will do that with the equations and it will do that in practice, as well.)

In the case of powering the LEDs with the freewheeling diode and capacitor, though, the nice thing is that the current is steady and much lower so there is more room for error and less worry.

And besides, there is that base resistor. Just test the circuit and adjust the resistor. Or else design for a lower target.

For driving a stack of LEDs, a batch of the same transistor type will all perform fairly close by. Seems to, anyway. I have thousands of PN2222s in a box, so I suppose I could spend my days trying them. But so far, they all behave close enough for this use.

But follow it with an LDO and you could power a micro with it. In fact, I'm going to be testing that out soon.

I see your point here, now. Is the saturation point more predictable here than the variations in a batch of same-designated PN2222As, for example? I've not played around with cores enough to know.

Except they don't. What happens is that the BJT/transformer/resistor pretty much define the Joules per unit time. If you know the power required by your load (and that is NOT hard to estimate well with LEDs) and you know the frequency (also predictable enough even with your reticence about beta), you can size the cap well enough for whatever ripple you can accept. It will just take longer, with larger caps, to get started. And there will be wasted energy sitting on the larger cap when you turn it off, I suppose. The LEDs won't behave as radically as you seem to be suggesting. They don't in my experience, yet, anyway.

Jon

It was operator error... :-S

Yes. Higher inductance means "SLOWER." But it doesn't mean more energy per unit time. In fact, the inductance cancels out of the equations in calculating power. The great thing about higher perm is fewer windings. However... they are often wasteful of energy at lower frequencies (for example, #75 ferrite is horribly lossy even down at 1MHz with more than 20 times the loss as #64 ferrite at 1MHz, while #75 has max permeability of 5000 and #64 is only 375.) Also, the lower perm stuff seems to be non-conductive while the higher perm seems to be pretty much a semi-conductor, at least. Still, some of the higher perm stuff is okay, loss-wise. But it seems most of it is intended for RFI shielding, not transformers. But I'm still reading and learning, so there probably are some things I'm missing still.

I think I take your point here. Higher mu means the same B will be reached a earlier, I_peak wise, if keeping the inductance fixed and the magnetic loop length fixed. B = mu*H = mu*N*I/l_m, but L is proportional to N^2 and mu... so I guess B is reached by the square root of the ratio of mu's. A mu=5000 material would saturate with 10 times lower I_peak than with a mu=50 material, assuming L is constant and so is the magnetic loop length (same physical core shape?)

Yes. I had to sit and think about that, too. You mentioned just in the prior paragraph the "dropping the inductance to zero." But that doesn't actually happen. What happens is that it goes down to what amounts to an air/vacuum core. It still has some inductance, but not much. That's because when the core itself saturates, it's done with and over for the core. So if you keep trying to store more energy the field lines expand out into space around the core and use the air/vacuum there to store energy, just like an air core would. So there is still room and that means there is still inductance. Just not much, because all those windings are now just packing energy into the vast regions of air around it once the core is saturated up.

If you look closely at some of the better saturation curves, you will see that there is still a _slight_ slope after saturation. That slight slope is the emerging air core inductance.

I think it is called N. Too many windings are required. Think about the reluctance, which is inversely related to the permeability. High perm materials have low reluctance. Low perm, like air especially, have high reluctance. Net reluctance is affected by inserting an air gap -- in fact, the air gap may so completely dominate things if it is big that the other core material doesn't matter and you might as well be winding an entirely air core from the start and just not bother with the ferrite, powdered iron, etc.

Think of a series resistance circuit where you have some aluminum you can run out some length of and the some carbon of the same diameter to make a combined loop. What is the total resistance? Well, if it is all aluminum and no carbon at all, pretty low. But if you start packing in much of a length of carbon instead of aluminum, it won't take much before the total resistance is pretty much based only on the carbon and the aluminum that is left doesn't really matter, anymore. It's like that.

Unless the gap is VERY small (smaller than I have ever seen), it pretty much takes over. Permeability of the rest of the core doesn't matter anymore and the calculation of inductance depends on the mu for air and the magnetic length of the air gap (no longer the magnetic loop length for the rest of the core.) You really need SMALL air gaps in order to achieve any reasonable value for the resulting L. You wrap the windings around the air gap itself to help contain its fringing into the wild world and you use the otherwise useless core with its high perm material primarily to give the flux a tight narrow conduit around from one end of the air gap to the other end of the air gap, without fringing all over hell and gone. Read chapter 5 from that site.

Jon

Yes, I realise that. I really meant it goes down to something very small compared to the pre-saturation inductance.

I've come to much the same conclusion. With very low inductance, the current shoots up to the maximum the transistor can handle in a very short time, so the circuit would have to operate at a very high frequency, and then switching losses would ruin your day.

So it helps to have a moderate amount of inductance to keep the rate of current increase down to something manageable. To get that with an air core, you would have to put on a lot of turns. Using a core with moderate permeability lets you get away with less turns. Its energy storage ability will be lower, but still enough to be useful.

Another analogy I've seen is that it's like trying to make a circuit using bare wires immersed in salt water. Really hard to keep the current from leaking all over the place!

Anyhow, from what I've seen so far, it sounds like for frequencies under 100kHz, either a metal powder ring or a gapped ferrite ring would be best. A solid ferrite ring, such as a bead, doesn't sound so good.

I don't doubt that the voltage across the capacitor will be fairly steady, but that doesn't say much about the current through the LEDs, since they have a highly nonlinear voltage-current relationship.

Have you actually observed the current waveform through the LEDs with the capacitor but *without* any resistor in series, or with only a very small one? Putting any resistor in there for the purpose of measuring the current could have quite a noticeable effect on the smoothness.

I haven't either, but there was a time in the pre-transistor and early transistor era when pulse-generating circuits involving saturating inductors were used. Apparently at the time it was a relatively inexpensive way of getting accurately-timed pulses, so the core characteristics must have been fairly predictable.

I don't doubt you can make a stable stream of pulses but I bet they had a variable resistor to fine-tune the frequency.

The issue here is energy transfer per unit time. You keep thinking in terms of current and voltage separately. Change the viewpoint.

Jon

I had been wondering out loud in this thread earlier about using a gapped arrangement. Basically an air core inductor with a ring of material linking the north and south poles, minimizing fringing that usually takes place in an all-air inductor between the poles and allowing the gap itself to be a very controlled magnetic loop length of air. I guess the secondary wound elsewhere would be linked well enough by the ring of material so that flux changes would induce the desired voltage without the kind of usual poorer linking seen in pure air cases.

My ignorance on this subject makes me feel still uncomfortable. In thinking about your above comment more, it seems to me that the effective permeability listed for any particular core (which is a max figure) may also tell us just how much effective air gap there is. If they are very high perm, there cannot be much of an effective air gap, at all. I think this couples well with the idea that high perm materials are also conductive. Low perm materials, though, as well as some comments from the site I referred to you about gaps in ferrites, suggest to me that what makes the difference here is mostly the sum of all those air gaps in the material. They both increase electrical resistance due to their presence, but also increase reluctance. So perhaps the low perm materials work better here than high perm for the reasoning line you are following above.

Still, if the permeability isn't close to 1 (and 20 seems to be the lowest I've seen), then it's NOT dominated by the air gap as it would be in anything gapped. But now I see a problem with that reasoning, too. I'm speaking about permeability. What I'm neglecting is magnetic loop length. There is a relationship between the permeability, of course. But the permeability doesn't tell me the permeance, just one part of that. I need to sit down with a specific core, look over its exact physical dimensions along with its max rated permeability and the AL rating for it, and work out how large of an effective air gap there is from all that. Not just hand wave about it.

If I had to guess beforehand, I'd say that low perm ferrites store energy pretty well, like a well-, and small-, gapped core might while still allowing some decent energy storage. But I also agree, as I wrote on the 9th, that an air gap seems to make good sense.

Jon

Maybe I missed it, but I get the sinking feeling that the discussion is overlooking coupling. You both seem focused on inductance, but I don't see where you are considering it (coupling). Jon raised the issue earlier, but we didn't go very far with it.

Ed