Joule Thief 101

[hoptoad]

Also another good old Lidmotor video.


https://www.youtube.com/watch?v=vSpVuxD1hY4

[MileHigh]

Rusty:

MH:
  Your 'outside the box' JT is interesting but I really don't understand it.  It did remind me of one design that was shown years ago where the oscillator had a second button cell power source that was dedicated strictly for transistor switching.  The drain on the button cell was tiny and the cell could run a long time.  I never built one but it looked interesting.  Maybe Bill remembers it and can link to it.  To me that was a good example of 'outside the box thinking' that most experimenters don't do.  I know a guy who built a JT using dollar store green steel twist tie wire. How weird is that?  I think he called it 'Penny'.

---Rusty

It's all based on the inertia of the current flowing through the coil.  Please see the attached stripped down version of the design.  Let's assume that the normal current rating for the LED is 20 milliamperes.  I will assume that you have a variable bench power supply so you could adjust the voltage so that you get the desired 20 miliamps of current flow.

There are coil design issues to consider.  There is not too much current flow so there is a concern about how much energy you can store in the toroidal core.  But you also don't want too many turns wrapped around the toroid because that will mean too much wire resistance.  For the sake of argument suppose you settle on five turns of thick wire around the toroidal core.

When you manually make contact and connect the positive of the power supply to the coil/diode junction, the LED lights fully say after a half second of contact, and when you disconnect the power supply the LED remains lit for say half a second.

That means that by making intermittent contact with an alligator clip in your hand, you will see the LED remains permanently lit.  So then you are just a short step away from using a signal generator and a transistor in an emitter-follower configuration to control the switching of the power supply in and out of the circuit.

The net result is this:  Let's say you connect the battery into the circuit with a 10 Hz pulse train that has a 15% duty cycle.  That's the "gate" that controls the "pushing" of the inductor.  This will give you near-DC through the LED.

MileHigh

[Lidmotor]

MH:
  That make a lot more sense to me.   Thanks.  I could probably lash up something using a CMOS 555 and give it a go.  I think you are right about it not beating a JT but it is something very new and different to try.  The guys with the good testing tools could have some fun testing it.

Hoptoad:
  Thanks for posting that old video of the two Penny oscillators having a little chat.  I actually had forgotten about that one.  Here is the video that has the circuit diagram and build description of the device if anyone is interested:
https://www.youtube.com/watch?v=mLeC9bHMeiY

[MileHigh]

Rusty:

I did a proto-schematic this time, see attached.  I show the potentials in the loop in red when the LED is illuminated and S1 is OFF, and the potentials in brackets are for when S1 is ON and the LED is illuminated.

Note you could use one CMOS 555 to drive S1 and S2, I would suggest it be running off 4.5 or 5 volt battery source.  You would drive S1 with a low-duty-cycle pulse and connect the complimentary output to S2.  So when S1 is ON then S2 is OFF and vice-versa.  You could use a second CMOS 555 to drive S3.  S3 is your "LED flasher" that allows you to take advantage of the persistence of human vision.  The two timers would be completely independent of each other.  More importantly, you want S2 and S3 to be ON (CLOSED) most of the time to reduce the power drain on the main coil when the LED is not lit.

Note I show this with NPN transistors.  Since they are switching only about 20 miliamperes, the base resistors can be quite high in value.  However, presumably most of the time S2 and S3 will be ON, which represents a power drain to the battery powering the timing circuit.  If you assume that one or more FETs or MOSFETs in parallel will give you as good or better an ON resistance, then that would probably be the better choice because there is no power drain on the timing circuit to keep a FET ON.

Note that with a big coil when the LED is flashing ON with a low duty cycle and a 30 Hz or greater repetition rate, the current through the LED will be essentially flat.  By adjusting the pulse with of the CMOS 555 controlling S1 you can set the current through the LED to an optimized value.  Contrast that with the attached scope shot showing the potential inefficiencies with a typical Joule Thief LED waveform.

In essence, you can play with the perceived brightness of the LED by setting the current level with the S1 duty cycle, and by playing with the S3 duty cycle and frequency.

MileHigh

[MileHigh]

1.  Not exactly correct.  At first, I thought that when you "tuned" (using a vr to the base) and the led got a lot brighter, that it was a type of resonance.  I was told this back then and, it seemed to fit because you could adjust the resistance and get the led to be very bright...move the vr a tiny bit up, or down and the led dimmed.  Later, we just called this the sweet spot.  What I am saying is that we found that you could go past the sweet spot and have to adjust back to it.  Also, about this time I learned how to properly measure the amp draw (mA draw, ha ha) in these circuits and found that you could tune the base vr to achieve a very low mA draw.  You could watch the led while doing this and tune so you got a decent brightness with as low an amp draw as was possible given that circuit's parameters.  Thus, you could tune for max brightness, or tune for max running time with still usable light.  So, I don't think we still called it resonance, but it is possible that we did after that time.

2.  Close.  It was those led Christmas lights (man, what a bargain for $3/ string of 100 leds)  I found that I could light 100 of them with the Fuji AA battery JT and they were very bright.  Using the same circuit, I added another 100 and could not see any difference in the light output.  (This is where a light meter would have been very useful back then) Stefan told me that those lights must have been in series as that is the most efficient way to hook them up like that with many leds.  300 leds made it noticeably dimmer and 400 leds brought it down to about 1/2 brightness of what the 100/200 leds were.

I know this is going back a few years (7) and my memory is not 100%.  Before I did the 100 leds strings I think the most leds I lit from a JT had been 4 or 5 and that was using a "standard" JT circuit.  Once I learned how to mod the Fuji, I was off to the races, ha ha.  That was also the time I learned how to zap myself many times even if I was being careful.  Some of those really hurt, ha ha.

Bill

Well, like I stated before, the base resistor in a Joule Thief is not supposed to be varied at all.  Rather, it is supposed to be based on the EMF coming from the feedback coil and the amount of base current required to switch the transistor hard ON, which is based on the maximum current that will flow through the main coil.  The real way to change the way the Joule Thief operates is to play with the size of the core, and the number of turns in the primary and the secondary.  The value of the base resistor "falls out" from those and related parameters.  Any astute electronics experimenter should be able to show exactly how the value of the base resistor is determined for a given standard Joule Thief configuration.

For the issue of the LEDs in series, I am close enough.  If you build a Joule Thief and you are driving a single LED, and you like the illumination level, and then you try 10 LEDs in series and get the same illumination level in each individual LED, then that is telling you that you can reduce the power going into the single LED by a factor of 10.  So if you are a keener and you are up to it, you can challenge yourself to figure out how to do that.

MileHigh