Fooling around w/one of Colin's circuits (3V-9V converter)

[carbonzit]

I'm currently playing around with one of our own Colin55's circuits. It's a simple but fascinating one, actually just the power supply to another project. It's a flyback step-up converter that uses 2 1.5V cells to produce 5 VDC. My idea is to tweak it to replace 9-volt "transistor" batteries for some of my own projects.

OK, so I got a simulation running in LTspice. There are some interesting quirks.

In the attached picture, the right-hand side is his "stock" circuit (but with the voltage divider values changed to produce 9 volts). Notice how jaggedy the voltage waveform is. This is something Colin points out in his description.

So I though I'd try to clean up the output with some filtering. Now, I know next to nothing about filter design, though I know basically how they work. So I played around with a pi-filter design in LTspice and came up with the other circuit. Notice the resulting waveforms, the last of which is much cleaner. Still some ripple, but no jagged spikes.

One thing about the "original" circuit behavior bothers me. What's up with that >160mS delay before the oscillator starts?

The other thing is that the whole thing depends on the size of the load resistor. By trial and error I found that for the pi-filter circuit, any value below 665Ω results in (practically) no output. I'm assuming this means that the maximum load for the circuit has been exceeded, which puts the maximum output at about 11mA (7.5V/665Ω). The "original" circuit fares a little better, at about 45mA.

My BC338 and 547s are coming with the UPS guy tomorrow, so I'll be able to test this out for real then. I really like the idea of this simple power supply; I hope the reality is somewhat close to that possibility.

Any ideas or suggestions appreciated.

LTspice sim files attached.

 

[colin55]

A lot depends on the size and type of material for the core.

 

[audioguru]

Where will the 3V input come from?
If the output is 45mA then the 3V will be 162mA and AA alkaline cells will drop to 2.4V in about 5 hours.

 

[carbonzit]

A lot depends on the size and type of material for the core.

Yes, I know. The transformers I've been winding are made from small reclaimed ferrite bobbins. Not sure exactly what kind of ferrite, but probably close to what's needed.

Next thing is to try to nail down the actual inductance, which of course has a big effect on circuit performance.

 

[Mr RB]

Put a small cap from B to E of Q1, that will slow doen the oscillation frequency a bit and probably improve efficiency. Then you can use less turns on the coils again increasing efficiency and reducing the amount of work to wind them...

 

[carbonzit]

Well, I got the converter to work, which was pretty nice: after tweaking the voltage-divider resistors a little, I can get a solid 9 volts out of it, with two good AA cells.

But it's oh so weak. I used 590Ω as a load resistor; anything smaller and the output voltage sags badly. Use a couple not-so-fresh cells and it sinks. I'm going to try to characterize it a little better, keep a constant real load (a couple LEDs on it) and see how long the AAs last.

I wonder: what could be done to improve output, efficiency or both? From Colin's description, it seems pretty well optimized.

Put a small cap from B to E of Q1, that will slow doen the oscillation frequency a bit and probably improve efficiency. Then you can use less turns on the coils again increasing efficiency and reducing the amount of work to wind them...

So what exactly would that do, Mr. RB? I tried a small cap, but it might've been too small. (47pF or so.) Will try a couple different ones.

 

[ronv]

If you are using your filter you can replace the Rs with some Ls and improve both output and efficiency.

 

[Mr RB]

It's not a great circuit to get high current out of. You need a good transistor that saturates to give a low voltage Vce when it is on, the BC337 is a good choice especially if it is a BC337-400 (beta 400).

But at high frequencies the BC337 might not have enough time to turn on hard, so there will be limit to L2 current and that means output current will suffer. Also it needs a couple of caps on the input, the big batteries won't supply the high freq current pulses very well, try a 100uF electro in parallel with a 10uF tantalum on the input.

It's probably not switching fast enough too, you can add a small cap between the collector of Q1 and the top of R3, to add a heap of positive feedback. It might also benefit from a small cap across R3 to give you about 0.2v ripple there to ensure a lower frequency and good turnoff of Q2.

Replacing the 1N4148 diode with a small schottky 1N5819 that has lower forward voltage and higher current, that will help.

It's hard to get good current and good efficiency from a 2 transistor design, you may have to go to 2 transistors as the oscillator to ensure a good tight square wave operation, and use a third transistor as the voltage regulator in feedback.

 

[colin55]

It's not a great circuit to get high current out of. You need a good transistor that saturates to give a low voltage Vce when it is on, the BC337 is a good choice especially if it is a BC337-400 (beta 400).

But at high frequencies the BC337 might not have enough time to turn on hard, so there will be limit to L2 current and that means output current will suffer. Also it needs a couple of caps on the input, the big batteries won't supply the high freq current pulses very well, try a 100uF electro in parallel with a 10uF tantalum on the input.

It's probably not switching fast enough too, you can add a small cap between the collector of Q1 and the top of R3, to add a heap of positive feedback. It might also benefit from a small cap across R3 to give you about 0.2v ripple there to ensure a lower frequency and good turnoff of Q2.

Replacing the 1N4148 diode with a small schottky 1N5819 that has lower forward voltage and higher current, that will help.

It's hard to get good current and good efficiency from a 2 transistor design, you may have to go to 2 transistors as the oscillator to ensure a good tight square wave operation, and use a third transistor as the voltage regulator in feedback.

All these things you are saying are basically untrue.

The circuit delivers 50mA at 5v with 30mV drop and 75mV P-P noise at 80kHz.
The 2k2 reduces the impact of the gain of the transistor enormously, so an extra-high-gain transistor is not going to be of any advantage.
The circuit has been especially designed “very lightly” so that it consumes very little current. At present, it consumes 5mA.
An extra 10u cap on the input is not going to improve matters. A 100u is perfectly sufficient across the battery. However in the original design, the circuit delivers 10-30mA @5v and an electro is not needed.
Changing the diode is not going to alter the output as a 1N4148 will handle over 70mA.
And saying you cannot get good efficiency from a single transistor is simply not correct. A single transistor works much better because it responds to the requirement of the transformer much better than delivering a long-duration square wave, that maybe over-saturating the core.
The circuit self-regulates as the input voltage changes.
You should build the circuit and try the things you are saying, before espousing your beliefs as facts.
They are completely throwing newcomers “off target.”

 

[carbonzit]

I'm intrigued enough with this circuit to continue researching it (aka playing with it). I'm getting some interesting results.

One thing that would help me enormously, Colin, would be if you could give me the specs for the inductor core you used for a transformer. The version I got working used a medium-size ferrite bobbin (about 1/4" dia.) with many more turns of wire than you called for in your design (75:115). It works, but I wonder if it might not be optimal. I would really like to see just how many joules I can squeeze out of this power supply.

For what it's worth, I think it's quite an elegant design.

 

[colin55]

The inductor core was an old 10mH choke. It has a core of 1.6 to 2mm and a top and bottom of 7mm (0.3in) The wire is 0.25mm but the feedback can be any finer gauge. More than the specified number of turns reduces the output current capability.

 

[carbonzit]

So the results from the first round of testing at Carbonzit Laboratories, GmbH are in. I wound a new transformer (see below) and used two different loads for testing. Source were 2 somewhat worn-out AA cells.

Load 1: 4 super-bright white LEDs:
Vi: 2.3V
Ii: 110 mA
Pi: 250mW
Vo: 12.6V
Io: 11mA
Po: 140mW
Eff: 56%

Load2: 4 green not-so-bright LEDs:
Vi: 2.2V
Ii: 105 mA
Pi: 230mW
Vo: 8.2V
Io: 18mA
Po: 140mW
Eff: 60%

I don't really understand why the output voltage seems to track the load exactly. I used no current-limiting resistors on either set of LEDs; the output current just seems to naturally settle at the amount wanted by the LEDs (????).

Colin: I just could not get any transformer to work with the number of turns in your design (20:35). I kept that proportion; the one that drove the working circuit that gave the results above had 40 and 70 turns. That core you said you used seems awfully small; mine was 2-3 sizes larger than that. (Have to measure it and post the size later.)

 

[Mr RB]

All these things you are saying are basically untrue.

The circuit delivers 50mA at 5v with 30mV drop and 75mV P-P noise at 80kHz.
The 2k2 reduces the impact of the gain of the transistor enormously, so an extra-high-gain transistor is not going to be of any advantage.
The circuit has been especially designed “very lightly” so that it consumes very little current. At present, it consumes 5mA.
An extra 10u cap on the input is not going to improve matters. A 100u is perfectly sufficient across the battery. However in the original design, the circuit delivers 10-30mA @5v and an electro is not needed.
Changing the diode is not going to alter the output as a 1N4148 will handle over 70mA.
And saying you cannot get good efficiency from a single transistor is simply not correct. A single transistor works much better because it responds to the requirement of the transformer much better than delivering a long-duration square wave, that maybe over-saturating the core.
The circuit self-regulates as the input voltage changes.
You should build the circuit and try the things you are saying, before espousing your beliefs as facts.
They are completely throwing newcomers “off target.”

Rather than argue with you Colin I will apologise for my post which due to my hurried typing of suggestions could have could have easily been perceived as a criticism of your circuit.

I was not meaning to criticise your circuit in it's original form but was more suggesting that (in my opinion) it was not ideal for the specific higher voltage output, higher current output the OP was requiring.

I will state that 2 transistor power converters are a favorite hobby area of mine specifically high efficiency versions, so my suggestions were aimed more at that direction. Anyway please accept my apology Colin, I was not trying to say your circuit was "bad" but was suggesting modifications to try to suit the OP's specific requirements.

 

[carbonzit]

I will state that 2 transistor power converters are a favorite hobby area of mine specifically high efficiency versions, so my suggestions were aimed more at that direction. Anyway please accept my apology Colin, I was not trying to say your circuit was "bad" but was suggesting modifications to try to suit the OP's specific requirements.

Well, that being the case, I'd like to hear more from you. Given the current state of this experiment, what are your suggestions for improvment?

I should say that while I am after higher voltage (9 volts, perhaps even 12), I don't have too many unrealistic expectations so far as current goes. I'd be happy with 50-60 mA or so. Of course, more is better ...

And better regulation would be nice, if that's possible. At least as far as my version goes, it's nowhere as self-regulating as Colin claims for his (5 volt) version.

 

[colin55]

You can replace a 9v battery with this circuit.
The output is about 10.4v on no load and 9.6v @30mA output.
The advantage is the voltage stays over 9v for the life of the cells.
A normal 9v battery drops to 7v very quickly.
The output voltage is set to 9-10v by the 6k8 and 390R resistors. The 470R gives the circuit an idling current of about 20mA. The spikes are about 75mV.
**broken link removed**

 

[Mr RB]

Well, that being the case, I'd like to hear more from you. Given the current state of this experiment, what are your suggestions for improvment?
...

I won't suggest any more changes for Colin's circuit, but if you are interested in taking a little time to make a high efficiency 3v to 9v converter I can suggest a method and a basic design.

I don't like transformer driven oscillators and prefer just to have an inductor. I also like to start at the power stage. Start with a 3v PSU and a squarewave signal generator that can adjust frequency and duty cycle (even a 555 with a couple of pots will do).

Then build your power stage consisting of the main switch (power) transistor and the inductor and diode, and Vin and Vout caps of course.

Then run the power stage by itself, driven by the signal gen until you find the most efficient zone, best freq and duty etc for your desired output voltage at max output current (or at typical output current). This gives the info to tell exactly what the power circuit needs to be doing for excellent performance. Change the inductor as needed, even if just for experience to try different types and inductance values. As it is for battery use you should also check performance with a minimum Vin (flat battery) and decide on "typical" battery voltage.

Once you know the freq and duty, add a second transistor as the basic oscillator, ideally this should turn the power transistor on/off at the right freq and duty cycle. The output of the inductor gives heaps of poitive feedback to force the oscillator into a good solid squarewave.

Once it is oscillating well at full power (like the specs with the signal gen) you can add a third transistor on the output as a "regulator". When Vout gets too high this transistor starts to turn on and it is connected to reduce the ON time of the power transistor.

The 3-tran circuit should give good performance and easy tuning;
power tran - optimised with inductor for good efficiency
osc tran - optimised for good squarewave and correct freq
regulator tran - optimised for low output current drain and good regulation

 

[carbonzit]

Well, Roman, that certainly sounds like a reasonable approach with a high potential for success.

I only wish I had the skills to take that approach; at the moment, I do not. I'm starting to take baby steps in that direction.

Do you have any representative circuits posted? or could you provide one here?

Color me ... curious.

 

[Mike - K8LH]

Wasn't there an article Silicon Chip long ago, "Never buy another 9v battery", that used a little 8 pin Texas Instruments TL499A IC?

 

[carbonzit]

Yes, I know all about those chips. Plenty of them from Linear Technology and Maxim. I want to do this with discretes.

 

[Mr RB]

...
Do you have any representative circuits posted? or could you provide one here?
...

I have some stuff on my web page here;
https://www.romanblack.com/smps/smps.htm

and a boost converter here;
https://www.romanblack.com/smps/conv.htm
which is only 72% efficient but was not optimised for efficiency but for constant power.

I'm reluctant to de-rail this thread any more, as the thread title contains "Colin's 3v-9v converter" and out of respect to Colin and the fact that he is also taking time in this thread to help you out I don't think it should become a thread about my designs rather than Colin's designs.

However if you wanted to start a new thread "how to design a high performance 3v-9v converter" I would be happy to participate and you will probably get input from lots of people as this is a very useful application.