It works! (JT/multivibrator mash-up) - comments welcome

[carbonzit]

For those who have followed my grueling progress here, I finally got my Joule thief/multivibrator mash-up working! I feel vindicated.

I first got this circuit working in LTspice (.asc file below):

**broken link removed**

Here's the block diagram:

**broken link removed**

I was very pleased to see nice clean waveforms on my multivibrator outputs, and even more pleased to see the Joule thief oscillators being gated on and off. (The simulation is still running as I type; I'll post the plots later when it's done grinding away.)

Then I built (breadboarded) the circuit for real. It worked great! LEDs flash very brightly and cleanly.

First of all, I'd like to thank those who helped me, especially Eric Gibbs and Colin, Eric especially for his gracious help with LTspice, which was indispensible to this project.

Now, it works, but like they say, there's always room for improvement. And since the main reason for this in the first place is my continuing quest to learn electronics properly, I'd like to better understand what's going on here and how I could improve and simplify this circuit.

First of all, the method I'm using here of "quenching" or pinching off the Joule thief oscillators seems a little crude to me. It does work, and seems to work reliably, but I'm wondering if there are other, better ways to do this. Basically I'm using the high output from the multivibrator to turn on the intermediate transistors (Q5 & Q6), which then _______ (in other words, I don't really know what they're doing to stop the blocking oscillators). How exactly is this working here? Does the C-E of that transistor alternate between hi-Z and low-Z? All I know is that it does work, but I'd really like to know why it works.

Would it be possible to eliminate these in-between transistors and somehow inject the multivibrator output directly into the blocking oscillator to gate it on and off? I tried this (in LTspice) but couldn't get it to work.

Another aspect, in response to a comment by Eric G., is that there are two separate blocking oscillators. Would it be possible to implement this with a single oscillator, and somehow alternately direct its output to one of two LEDs driven by the multivibrator?

There's also the concern about battery drain, since the whole shooting match is powered by one little AA cell. (The LEDs flash very brightly, by the way. Thanks to Colin for his "LED torch" circuit used here.) In particular, I'm wondering about the time the Joule thieves are "off". There's still ~1.5 volts across the (inactive) LED. But since it's below the Vf of the diode, does this mean that there's very little current flowing through the LED (on the order of µA)? That's my understanding, anyhow. And any other ways to reduce consumption would be nice to know about.

So any comments welcome. Keep in mind the rules of my game here: discrete, easily-available components only (no ICs allowed!). Simpler is better.

 

[MrAl]

Hello,

Many congrats to you! Its nice to see you got it working to your liking, but now that you have it is mandatory that you provide some pictures of the board and how about a video (if you feel up to it)?

 

[carbonzit]

Well, let's see: I'm expecting my first eval boards Thursday, hope to go into production early next week ...

So far as video goes, sorry, no video capability at this end yet. But thanks anyhow!

So any comments on the circuit?

(Just to show how impractical this circuit is, this weekend I went to a dollar store and bought--for a buck--a bike light with 5 LEDs. Opened it up, of course, and all the "magic smoke" is contained in one of those round black blobs, the only other component being a momentary rubber switch. Popped 2 AAAs in it and fired it up: not only does it flash, it has 5 different chase modes, and lights up the LEDs really brightly. Don't know how they get all that magic into a little blob, no external caps, inductors, no nothing.)

 

[MrAl]

Hi again,

Sounds like you want to move into microcontroller land? You can do lots of that stuff with one of those things for the cost of a single electrolytic capacitor. External parts? Just a resistor and LED, a couple batteries and a battery holder.

 

[carbonzit]

I do (want to get into microcontroller land), but not here. I like both worlds. I've done a fair bit of µp programming (SX-28, not PIC) and enjoy it. But I also like the more analog, discrete side of things. Silly, perhaps, because my transistors here just have a bit less of the invisible magic than do ICs, but still, I like messing with discrete circuits like this.

In fact, one of the things I really like about the "Joule thief" part of this circuit is how a simple li'l circuit cobbled together with a hand-wound transformer can actually out-perform a commercial 1-chip product (the Prema PR4401), as explained on Colin's LED torch page. (Take that, highly advanced Silicon Valley!)

So how does my oscillator-quenching thingy work here?

 

[dougy83]

So how does my oscillator-quenching thingy work here?

When on, the transistor (e.g. Q6) pulls the base of Q4 low and stops it from being able to turn on.

You said you were concerned about the LED being across the 1.5V supply; if you put the LED cathode to the 1.5V rail instead of gnd, you won't have that problem.

 

[carbonzit]

When on, the transistor (e.g. Q6) pulls the base of Q4 low and stops it from being able to turn on.

Aha. I suspected that's what might be happening. Thanks for that.

You said you were concerned about the LED being across the 1.5V supply; if you put the LED cathode to the 1.5V rail instead of gnd, you won't have that problem.

Except that the - rail is ground. Or do you mean something else?

 

[dougy83]

Except that the - rail is ground. Or do you mean something else?

No, I mean what I said: "1.5V rail". It certainly is not ground (0V reference).

The current through the LED will be less as the inductor will be delivering its energy at the forward voltage of the led, Vf, instead of using the additional 1.5V rail in the original configuration (in which case it delivered its stored energy at a voltage of (Vf-1.5V) ).

 

[carbonzit]

No, I mean what I said: "1.5V rail". It certainly is not ground (0V reference).

The current through the LED will be less as the inductor will be delivering its energy at the forward voltage of the led, Vf, instead of using the additional 1.5V rail in the original configuration (in which case it delivered its stored energy at a voltage of (Vf-1.5V) ).

Hmm, now I'm really confused. You mean the 1.5V negative rail? but isn't that the same as ground? (the battery negative goes directly to ground)

You can't mean the +1.5V rail, can you? Where would I connect the LED cathode in my circuit?

 

[dougy83]

Ground is ground is the 0V reference, as depicted in your schematics. There is no -1.5V rail. There is a 1.5V net that you've labelled 'V+', which is what I'm talking about. I was saying that you can connect the LED cathode to 'V+' instead of ground, because you said you didn't want the LED to have 1.5V across it always.

 

[carbonzit]

So you're suggesting connecting the LEDs across the transformer primary?
I don't see how that could possibly work. But maybe I'll try simulating that in LTspice and see what happens.

By the way, I removed the two decoupling caps (100nF) from my breadboarded circuit and it still works fine. They were what I called my "paranoia filters", which apparently aren't needed.

 

[dougy83]

So you're suggesting connecting the LEDs across the transformer primary?

yes

I don't see how that could possibly work.

Well maybe you'll learn something new then. If you come to a dead end & need to read why, see https://en.wikipedia.org/wiki/Flyback_diode

 

[carbonzit]

Here's the LTspice plot of my mash-up. Only took all day (literally) and 6 gigabytes of disk space. (When I went to close the .raw file window, it took the computer 10 minutes to repaint the screen so I could close it!)

Oh, and regarding what happens when one of the JTs is off and there's 1.5V across the LED, no problemo: there's pretty close to zero current flowing through them, as found both by simulation and actual measurement. (IOW, if you're fairly far below the diode's Vf, its If will be ~0.)

 

[carbonzit]

So doesn't anyone want to come out to play and give me some hints, suggestions, anything? [sniff]

I still have a couple of questions that I think are legitimate and would really like answers to:

1. Can I somehow eliminate the transistors (Q5 & Q6) between the multivibrator and the Joule thieves? How?
2. Is there a way to use just a single Joule thief oscillator and use the multivibrator to switch its output between 2 LEDs?

 

[MrAl]

Hi again carbon,

I think what you may want to do next is dig into the reason behind how even one of these things works, that is, one "Joule Thief" circuit alone. That will give you ideas how to proceed with two circuits if you still want to do that later.

Also, there is most definitely a way to switch the output of a power supply to pulse two LEDs one at a time. The "Joule Thief" would just be a power supply in this case. Since only one LED is on at any given time, the "Joule Thief" still only has to power one LED so it should work almost the same as with one single LED.
Perhaps have one JT power one multivibrator with LEDs in the collectors? Im sure there are other ways too.
Might end up needing a Schottky and capacitor to smooth out the JT output a little, but maybe not.

I think maybe now you are entering that "I want to use the minimum parts necessary to do this" phase of the design. That's good too because this often leads to many insights in circuit design. You might think of it this way too: if you only have to use one transformer for one circuit instead of two transformers for one circuit, that means you can built twice as many circuits with the transformer stock supply you have on hand without buying any more cores and winding more transformers

Another little interesting thing here, i have a flasher circuit i got from some store display a long time ago, but it only has two red LEDs that flash, and they can not be changed out with white LEDs because the voltage is only 2xAA batteries and is driven with one of those epoxied over chip circuits. With your circuit you can flash white LEDs or blue LEDs, which is nicer i think. Red is too outdated

 

[ericgibbs]

hi Al,
I understood that the JT got its name from its ability to utilise most of the remaining energy from 1.5V battery cells.

I have seen them operate from discharged batteries as low as 0.7V, it would be interesting to see what the minimum voltage the OP's circuit will still start/run.

Eric

 

[MrAl]

Hi Eric,

Oh yes, the JT tries to use up all of the energy in a battery that may already be mostly depleted. I agree with that
I just found it useful for other purposes too because it only uses one transistor rather than two like some other similar circuits. It does require a dual winding transformer though, but that's not too hard to wind by hand i guess.

Down to 0.7v? Yes, i would think so as the main limiting factor here is the base emitter voltage. I ran into this problem back around year 2001-ish when i was working a lot with low voltage LED circuits. I had considered using a germanium transistor rather than a silicon one, but it was hard to find any of these. As you know their base emitter voltage is much lower meaning they would run even lower yet. I thought about this a lot and wondered why germaniums are so hard to find these days considering so many things run on batteries these days, then i read an article that said that researchers are trying to revive the germanium transistor, although it would be a sort of hybrid or advanced design in some other way. It would also have a very low band gap with base emitter voltage roughly around 0.2 volts, which would be nice to see again, plus higher speed than the older germaniums.
We'll have to wait for them to get to market i guess.

There is another problem that comes up though, and that is the 'power vs voltage' problem. As the input voltage goes lower, it requires more current on the input to get the same voltage output into a given LED. This means as the energy in the battery depletes, it drains down faster and faster. This is something i hated in regular 'bulb' type flashlights, and LED lights improved on this by dimming very gradually. That's with output regulation though, and the JT doesnt regulate so it's a little better anyway.

 

[ericgibbs]

hi Al.
Any particular Ge type transistor you are looking for.?

BTW: a useful little 'beeper sounder, continuity/high/low level tester can be made with a Ge front end.
Just enough voltage to check continuity at 0.2V , but not enough to turn on digital gates.

 

[MrAl]

hi Al.
Any particular Ge type transistor you are looking for.?

BTW: a useful little 'beeper sounder, continuity/high/low level tester can be made with a Ge front end.
Just enough voltage to check continuity at 0.2V , but not enough to turn on digital gates.

Hi Eric,

Why do you know of a place to get these things these days are a reasonable price? Of course i dont want to pay 5 bucks a piece for them
A type for mostly switching applications.

Oh yeah, good idea too. That's the kind of thing i was talking about where the voltage of the BE is low enough to allow some unusual applications, unlike the 0.7v silicon requirement which i never quite liked. Can not run that off of one solar cell even, but a germanium can!
It's also interesting that with a boost converter there's no way to get down blow 0.6v with si, but with a germanium a boost converter can boost that 0.5v up to 5v and run some other important circuit perhaps with bootstrapped power to keep it going.

 

[carbonzit]

Germanium transistors?

Hmm; couldn't find them anywhere else (so far), but NTE, of all places, has replacements for the venerable 2N107/109 (datasheet).

Wouldn't be surprised to see this available other places as well.