Help me explain this mashup (Joule thief + multivibrator)

[carbonzit]

So I went ahead and mashed up two circuits, using LTspice for initial nondestructive testing.

I took a Joule thief circuit that I know works (one of Colin55's projects), duplicated it and connected the two circuits together as one big multivibrator. (See attached .asc file and pictures.)

I really had no idea if this would even work. But to my total surprise, it did! But not in the way I was expecting (or, more accurately, hoping) it would work.

What I'm after is two Joule thieves connected as a multivibrator so that they alternately light, say for half a second each or so. If you look at the attached waveform, you can see that they are alternating, all right--on each cycle of the flyback (blocking) oscillator! Not at all what I'm after here, but a damned interesting quirk, at least to me. (The waveforms are taken from the transistor collectors/LED anodes.)

So what's going on here? And is there some easy way to make this work as I'd like? I'm thinking I'm close here, but I have the multivibrator rigged up wrong. (R3/C4 and R4/C3 are supposed to control the multivibrator cycles.)

I think this would be a really cool way to get two flashing LEDs out of a small supply. It would be even cooler if I could somehow run them both off the same transformer.

 

[carbonzit]

Thinking on this a little bit more, I wonder if what I need is a low-pass filter for the multivibrator to remove the high-frequency flyback spikes?

The problem is that I'm having a hard time separating the multivibrator (if that's what it even is in this configuration) and the Joule thief circuits here. I wish someone could draw a nice clear line at the boundaries of each (or tell me that this whole thing is just wrong!).

 

[MrAl]

Hi,

Did you really want to connect the two 10uf caps that way? One has two resistors and the other has none.

What is this, the "Dual Thief" ha ha.

 

[carbonzit]

How do you figure that? I checked the schematic again (I've since changed it), and the caps go between the collector of one transistor and the base of the other. The circuit is perfectly symmetrical. Unless I'm really missing something here ...

**broken link removed**

I've since replaced the caps with inductors. Changed the behavior (even more interesting!), but still not what I'm looking for.

"Dual thief"; heh. I like that. Can I use it?

 

[carbonzit]

So, is such a conglomeration even possible? Could the same two transistors have two roles, one as oscillators for the Joule thief function, the other as half of a multivibrator? Seems like a long shot to me, but you gotta try, right?

 

[ericgibbs]

So, is such a conglomeration even possible? Could the same two transistors have two roles, one as oscillators for the Joule thief function, the other as half of a multivibrator? Seems like a long shot to me, but you gotta try, right?

hi cz,
Using only one battery to drive two blocking oscillators is going to be less efficient than driving a single oscillator.

 

[MrAl]

How do you figure that? I checked the schematic again (I've since changed it), and the caps go between the collector of one transistor and the base of the other. The circuit is perfectly symmetrical. Unless I'm really missing something here ...

**broken link removed**


I've since replaced the caps with inductors. Changed the behavior (even more interesting!), but still not what I'm looking for.

"Dual thief"; heh. I like that. Can I use it?

Hi,

How do i figure that? Well you ask that question about a circuit that you 'changed' later, and your original post had a different circuit and i commented about that one. You can not expect someone to predict that you are going to change the circuit later and how you are going to change it too I can only comment on what i see posted.

Anyway, it looks like the two circuits have both individual modes and combined modes, so i cant say for sure what is going to happen without looking at it in more detail, but it appears that one of the individual modes will try to dominate and that will throw off the other so that it is not oscillating as nice as it wants to naturally. In other words, one part of the circuit is forcing another part of the circuit to operate as the first part wants it to, not how the second part really wants to oscillate. I would think this would prevent the second part from operating properly in one way or another...either not on long enough or not off long enough. In the circuit simulator you might not see this anomaly until you force an imbalance between the two parts of the circuit, the way it would really work out in real life with unmatched parts.
Usually circuit that have to oscillate have one part that causes the oscillation and everything else works from that, but in a way that is known to produce good results.
Why do you want to put two together in this manner anyway?

Yes, i think Dual Thief is a decent name for it dont you? That's if it works out in practice however
We might have to call it "Power Thief" if it doesnt

 

[carbonzit]

hi cz,
Using only one battery to drive two blocking oscillators is going to be less efficient than driving a single oscillator.

Yes, that must be true. With that in mind, here's what I think I'm after:

**broken link removed**

Of course, the "switch" is not really a physical switch, but a way of toggling the output of the blocking oscillator (aka Joule thief) between two outputs. I proposed using a relaxation oscillator (aka free-running multivibrator) 'cause I know how to make one of those easily.

Once again, for those tuning in late, the rules of my game:

• No ICs. Discrete components only.
• Should all run off a 1.5V cell (AA).
• Simple.

As always, all ideas welcome. Go nuts.

 

[ericgibbs]

hi,
Look at this link, you may be able to get idea's from the circuit.

http://www.cappels.org/dproj/PulseBoostLED/Pulse_Boost_White_LED.html

 

[carbonzit]

The plot thickens ...

I returned to simulating my mash-up and almost got it working. This time I thought I'd try gating the base of the Joule thief with a transistor, so I simulated this in LTspice, and it worked:

**broken link removed**

(The transistor is tied turned-on for the purpose of testing here. Spice file included below.)

So I put two of these into my mash-up. Unfortunately, it didn't work: all I got were sharp spikes at each cycle of the multivibrator on the LED outputs.

Or did I? Zooming waaaaay in to these spikes, I saw that they were actually oscillations from the Joule thieves, but for some reason they were damped shortly after starting, dying away to nothing.

**broken link removed**

So why does this (the JT) work in isolation but not in my mash-up? Obviously there's enough current to turn on my oscillator: why does it poop out?

I wonder about what assumptions (LT)spice is making here about this circuit. For instance, it must be assuming a perfect voltage source. I doubt if all this stuff would be able to work off a single AA cell in real life.

 

[carbonzit]

2 designs

OK, I've come up with two designs that look promising, except that neither one works. (At least they have different failure modes!)

Design #1 attempts to provide the base drive for the Joule thief circuits directly from the multivibrator;

**broken link removed**

(For those following along with Spice, this is "JT multivibrator 2" below.)

This circuit appears to be failing because the JT oscillators are interfering with the multivibrator, resulting in chaotic MV output:

**broken link removed**

I attempted to decouple the MV with chokes (L3-L5), but obviously this didn't work. (I chose 5mH to give about 10K of reactance at the oscillation frequency, ~300kHz.)

The second design seemed more promising ("JT multivibrator 3" below). Here I added transistors to gate the base voltage of the JTs:

**broken link removed**

As you can see, the multivibrator output is nice and clean:

**broken link removed**

However, all I get from the JTs are sharp spikes (which are actually short damped oscillations):

**broken link removed**

I think I've figured out that the problem here is that the transistor diode drops (2 X 0.7V, right?) are reducing the JT base voltage too much.

Where else could I gate the JTs? How about if I somehow shunted the JT bases to ground--would that be able to start and stop them? Any other bright ideas?

 

[MrAl]

Hi Carbon,

What is the motivation behind all of this? That is, what are you expecting to gain from this as opposed to say the original circuit with one transistor and one coil?

 

[carbonzit]

What is the motivation behind all of this? That is, what are you expecting to gain from this as opposed to say the original circuit with one transistor and one coil?

Two alternately flashing Joule thieves. (Flash rate = 0.5-2 Hz.) That's why the multivibrator in the middle.

 

[ericgibbs]

hi cz,
If you want an honest opinion.
LTspice is a circuit simulator not a circuit designer, you cant expect to cobble up any old circuit and expect LTS to make it work.
The circuits you expect LTS to simulate must be designed first and fit within the parameters of the components used.

 

[carbonzit]

If you want an honest opinion.

Yes, I do. Please.

LTspice is a circuit simulator not a circuit designer, you cant expect to cobble up any old circuit and expect LTS to make it work.
The circuits you expect LTS to simulate must be designed first and fit within the parameters of the components used.

I can't argue with any of that.

I'd like to know if I'm even in the ballpark with any of this. (I think I am, at least partially.) Even just pointing out whatever gross errors there may be would be helpful.

I guess this is as good a place as any to insert my

Disclaimer:

This project has little or no practical value. It certainly has no commercial value. It is being undertaken for the purposes of education (and amusement) only.

If this offends your sensibilities, I suggest you find other threads here to comment on.
​

 

[ericgibbs]

hi cz,
Neither you, or the circuit you posted offends me.

For a gross error report: the circuit as designed will not work, so LTS cannot simulate it.! I would call that a gross error, wouldn't you.?

It doesn't matter it if its of no practical value, I will always try to help anyone with a project, but when asked for for feedback, I try to give a considered reply.

If the directness of my comments upsets you, I will in future ignore your requests for help..

 

[carbonzit]

Neither you, or the circuit you posted offends me.

Good to know!

For a gross error report: the circuit as designed will not work, so LTS cannot simulate it.! I would call that a gross error, wouldn't you.?

Yep. So could you possibly give me a clue or three as to why? (Grossest errors first, please.)

Specifically, in the 2nd design, am I correct about why I get no JT output (starved base drive)?

It doesn't matter it if its of no practical value, I will always try to help anyone with a project, but when asked for for feedback, I try to give a considered reply.

If the directness of my comments upsets you, I will in future ignore your requests for help..

Sorry, that "disclaimer" wasn't really directed at you. I know you're pretty game about answering questions, which I appreciate.

 

[carbonzit]

Getting closer

So here we are again. This time I'm trying to gate the LED drive directly with series pass transistors (Q5-Q6). It's sorta working, sorta not. Here's the schematic:

**broken link removed**

Sorry it seems such a mess. See the attached LTspice result display below. This shows at least three major problems:

• 1. A "filling-in" of the end of the multivibrator off cycle with increasing JT oscillations (more about this below).
• 2. An incomplete turning off of the JT output.
• 3. And what's with those high voltages?!?!? The peaks of the JT outputs appear to be at more than 60 volts! That's probably a little bit more than one wants to put across their LEDs.

I suspect that most, if not all, of these problems are caused by the RF from the JT (Joule thief) oscillators leaking into and interfering with the multivibrator, which runs at a much lower frequency. Actually, until the end of the "off" cycle, the multivibrator outputs (the 2 lower traces) look remarkable clean, the best I've been able to get out of this mash-up so far. The JTs appear to be running at somewhere in the neighborhood of 300kHz, so all those solid areas are 300kHz HF.

I've tried to block the RF from getting back into the multivibrator (MV) by using 100pF bypass caps. They helped, but as you can see they haven't eliminated the problem. In other attempts I put in series chokes to try to keep the HF out of the multivibrator, but couldn't get better results than here. Is this something I should do? or does my problem lie elsewhere?

The high voltages make absolutely no sense to me. The JTs, when run as stand-alone modules, generate spikes of maybe up to 10 or 12 volts; the average output is closer to 6 volts. How could a series transistor increase the voltage so severely? (Hmm, maybe this is a "feature" that could be exploited for another circuit?) Now I'm kinda afraid to try to actually breadboard this circuit!

By the way, this took several hours to simulate in LTspice. (.asc file attached below for your viewing pleasure.) The resulting .raw file was 3.3 gigabytes. The reason it runs so slow and took so long is no doubt because of all the HF oscillations it has to track, so no fault of LTspice here.

So if you have any ideas, I'd really appreciate any light you could shed on this situation. I feel I'm very close to getting this working.

Specifically, what I want to do is use the low-frequency output from the MV (1-5 Hz) to alternately enable/disable the two JTs. This could be done either by enabling the JT oscillator (as I thought I was doing here), or by disabling or "quenching" it on the off-cycle. (I also hope not to waste too much energy when the JT is just sitting there idle.) Several possible methods have occurred to me:

• Gate the LED output as I've tried to do here
• Gate the JT collector
• Gate the base drive of the JT
• Inhibit the JT by injecting a voltage somewhere that'll "quench" it (???)
• ???