Yet another micro-boost power supply! (1.5V--> x? V)

[carbonzit]

I've found another one, rooting around through Sam Goldwasser's schematic collection site. It was intended as a micro-HV supply (for the electronic flash on a Kodak disposable camera); as such, it generated 350 volts from a 1.5V AA cell. (Circuit here). But Sam G. adapted it for lower-voltage duty, apparently by rewinding the original transformer. (Interesting discussion on that page about the operation of this circuit.)

Here's my drawing of it:

**broken link removed**

The thing is so dang simple that I just had to try it. So I did. I wound a little transformer on a small (1/2" or so) ferrite bead with some of my salvaged magnet wire. Dang if I didn't get a good solid 6 volts at about 25mA out of it!

I think this is actually a good candidate to run little 5-volt powered projects from a single AA cell. Of course, one would want to add a regulator (maybe just a 5V zener to keep it simple), since the voltage will go waaaaaaay up under a smaller load (I used a 220Ω resistor as a dummy load; haven't yet measured the no-load voltage, but I'm sure it's much higher).

Looking at this circuit, it's something of a mystery how it operates. My problem is with finding the return path to the positive side of the supply: it can't be through that little 470pF cap, can it? or through that rather large resistor?

I'm going to try to model this with LTspice, but simulating that transformer is going to be tricky. Wonder what frequency this runs at?

 

[Diver300]

The return path is though the base-emitter junction of the transistor. That holds the transistor in saturation until the transformer saturates.

I don't think that the output voltage would go way up under no load as it is not a flyback design. The output voltage will be approximately the input voltage times the turns ratio.

I think that the shortcoming is that there is no way of handling the energy that is stored in the transformer as the transistor turns off. The non-dot ends of the windings will go positive, and there is nowhere for that energy to go from either winding. You might be able to increase efficiency, at the expense of complexity, and it might loose voltage regulation, if you fitted a bridge of schottky diodes to the output winding, so that the flyback power were captured. You would have to make sure that the forward current still goes though the base-emitter junction to hold the transistor in saturation during the forward part of the cycle

 

[simonbramble]

If you run into problems simulating the transformer, the following might help you:

LTSpice tutorial:
http://www.simonbramble.co.uk/lt_spice/ltspice_lt_spice_tutorial_3.htm

 

[carbonzit]

Thanks, Simon. Turns out I already knew about coupling inductors using Kn statements. The problem is that I don't know the actual inductances of my transformer. I tried measuring it today (I built an inductance adapter for a multimeter that works very well), but couldn't even get a reading, so I ASS-U-ME it's below the threshold of my test equipment, possibly in the nanohenry range.

In any case, I did model the circuit (.asc attached below):

**broken link removed**

The long and short of it is, it just plain don't work. Can't get it to start oscillating at all. Tried adding and removing "startup" and "UIC"; no difference. Played with the inductor values, tried different transistors, nothing. Now, I know this circuit works in real life. I wonder what conditions aren't being met in order for LTspice to start this? Base drive for the transistor too low? not enough feedback?

In any case, I'm going to file this one under "interesting circuits meriting more experimentation" and revisit it later.

 

[ericgibbs]

The long and short of it is, it just plain don't work. Can't get it to start oscillating at all. Tried adding and removing "startup" and "UIC"; no difference. Played with the inductor values, tried different transistors, nothing. Now, I know this circuit works in real life. I wonder what conditions aren't being met in order for LTspice to start this? Base drive for the transistor too low? not enough feedback?

cz,
The problems not with LTS, its with your circuit.
Look at the notes and changes on this image.

LTS is not a circuit designer or circuit design tutorial, its only a simulator. If you make basic design errors in your circuits, LTS will not correct your mistakes.

I would advise you to calculate the circuit parameters/values by hand, using the datasheets etc, before using the simulator to check your results.

 

[carbonzit]

cz,
The problems not with LTS, its with your circuit.
Look at the notes and changes on this image.

Well, regarding your notes (and comments here), all I can say is, it works on my test bench. Kind of the opposite of the usual situation, where something works in simulation but not in real life.

So how do you explain that? I don't doubt the calculations you made in your notes. But the fact remains that this is a working circuit that Spice can't seem to simulate.

And keep in mind I didn't design this circuit. It was actually designed by someone at Kodak, for charging a capacitor for firing electronic flash.

 

[ericgibbs]

Well, regarding your notes (and comments here), all I can say is, it works on my test bench. Kind of the opposite of the usual situation, where something works in simulation but not in real life.

So how do you explain that? I don't doubt the calculations you made in your notes. But the fact remains that this is a working circuit that Spice can't seem to simulate.

And keep in mind I didn't design this circuit. It was actually designed by someone at Kodak, for charging a capacitor for firing electronic flash.

This is the d/s for the transistor used in the modified version of the circuit, it has a far superior specification at low supply voltages than the 2N2222.

extract from the authors notes.

Q1, the 2SD965, is a rather special high current high gain low voltage NPN transistor. Using a 2N3904 rather than the 2SD965 will NOT work well, if at all! D1 should have a PRV rating 3 or 4 times the output voltage and the transformer needs to be insulated for the output voltage. The other components on the primary side are nothing special. Note that for higher voltage operation (at the output), Q2 in the original MAX schematic may be needed. For low voltages, performance is much better without it. I don't know what the break-even point is so you may want to leave a spot for Q2 just in case

The LTS uses typical values for its semiconductors and other components. The fact that the built version works could be any number of factors, including circuit layout, component tolerances.

Hand wound transformers on unknown core materials can vary a lot in their inductance and coupling.

I think most of us has built a one off prototype that works and later find that 'identical' production circuits will not oscillate. The design should try to cater for worst case conditions.

 

[carbonzit]

Late night thoughts

I'm done playing with my little circuit for the night, so before I go to bed, a few thoughts.

Yes, it does work, despite not being supposed to, at least according to a simulation. Not bad, either. This power supply fascinated me with the possibility of running low-powered 5-volt circuits from just a single, cheap AA cell, rather than the more expensive 9-volt battery and 5-volt regulator.

It's an odd circuit. One odd thing I noticed: I took Eric's advice and changed the base resistor. But no matter what value resistor I used, I got the same results, more or less. ?!?!?

Another strange thing: I was actually using a regular silicon diode (1N4007) instead of a Schottky as per the circuit diagram. But when I tried a Schottky, the output actually went down! Same with a 1N4148. It only seems to like the 4007 for some reason.

It turns out to be a not-very-efficient boost converter. According to my measurements, I'm getting somewhere in the 25-30% range. Even my little 3V--> 9V power multivibrator is better than that.

But then, the original design really didn't need to be all that efficient. It was for a one-use camera with a flash, and as long as it could fire off a full roll of film's worth of flash exposures without failing, perhaps after sitting on the shelf a year or so, it was fine.

It's intriguing to try to take a circuit like this, made for a very specific purpose, and try to hack it for another, similar but different purpose.

If anyone has any suggestions, you know where to leave them. I'll be reading.

 

[debe]

Not the first time ive found computer test programs wont work on real life circuits that do work! Programs arent all that reliable.

 

[alec_t]

I think LTS models such things as thermal noise but not real-life interference (too much of an unknown for consistent modelling). A sufficient level of picked-up RF might be sufficient to trigger oscillation in an actual circuit such as this, whereas the sim refuses to start up the oscillator.

 

[Mr RB]

...

If anyone has any suggestions, you know where to leave them. I'll be reading.

What seems a good start would be to separate the output from the input, and design each to do it's job as well as it can. The current design suffers as the output and input are linked in such a way as it's not really possible to refine either one to improve performance, without killing it's ability to oscillate.

So I would start with an oscillator input stage, and a forward or flyback converter on the output.

For the oscillator input you can use another transistor to make it a good oscillator, or what might be more fun would be to add a 3rd winding to the inductor and use the 3rd winding to drive the transistor base making it a good single-transistor oscillator, regardless of output.

If you google for "single transistor fluoro inverter" or similar you should find a few designs that use an overwind to drive the base. You can probably use one of those designs as it is and just change the output turns to suit your desired output voltage.

 

[carbonzit]

Your suggestions are quite reasonable. However, that's not what I'm interested in regarding this circuit. I'd first like to see what improvements, if any, are possible to the existing circuit (one-transistor flyback converter), both in terms of raw output power and efficiency. I don't forsee anything like an order-of-magnitude improvement, but I'm curious to see what ideas other people might come up with.