3V->120V Microcontroller Boost Converter

Thread Starter

spacewrench

Joined Oct 5, 2009
58
I've wanted a nixie watch for some time, and have been collecting parts and learning how to build pieces of the circuit for a while. The next big hurdle is making a tiny boost converter that can go 3V (2xAAA)->120V@3-4mA or thereabouts. I understand the basic theory of boost converters, so now I'm playing with various hand-wound inductors and a microcontroller PWM peripheral, trying to get a comprehensive feel for what's going on before I build the real thing.

My test circuit is shown below. The GPIO/PWM is running at 8MHz/256 = 31.25kHz period, and I can set the pulse width from 1/256 to 255/256. I have a 40MHz, 2-ch analog 'scope, which I'm triggering off the GPIO pulse and looking at the inductor/FET connection. I have a cheapish LCR meter, but it doesn't seem to give reliable measurements of my inductors, so I'm not sure what my hand-wound inductors are. I also have a few commercial, marked inductors to test with.

With a 100uH inductor, I get boosted voltage between 13 and 40 with pulse widths between 1/256 and 30/256. (I think I can go higher, but my current filter cap is only 50V, and I'd rather not pop it.) On the 'scope, the FET pulls the low end of the inductor to ground and it rings while rising to ~0.5V or 1V as current increases. When the FET turns off, I get a big clipped ring between the output voltage and ground, decaying eventually to 12V before the next pulse starts.

With hand-wound inductor A, I see a similar trace, but after the inductor current rises to a certain level, the FET top voltage goes up rapidly ... I think the inductor is saturating, so I'm seeing the current through saturated-inductor / FET / 0.22R resistor. Is that guess reasonable?

With hand-wound inductor B (many turns, thin wire, possibly high-permeability toriod core) I see very little voltage rise during the "on" pulse, then the inductor output snaps up to the (low, ~13-15V) output voltage without much ring, then there's a low-frequency half-cycle dip or two towards the end of the PWM period, just before the next "on" pulse. My guess: (relatively) high inductance, not much energy stored in core during "on" time, and the ring frequency is much lower than with the 100uH inductor.

Another question: is the frequency of the ring I see when the FET switches related to the inductor value and the (parasitic) capacitance of the diode, FET, and other stuff at the scope node? Anything else affecting it?

And finally, I'm thinking of using a coupled / common-mode inductor to achieve a greater voltage boost, but I don't see how (physically) it would work. I'll add a schematic of how I think it should go together unless somebody says "won't work, don't bother!"

Thanks!
 

THE_RB

Joined Feb 11, 2008
5,438
Wrap another winding on your toroid as a low voltage winding, and use the FET to switch that. Use the high windings coil to generate the high voltage.

There are many little fluoro light driver schematics on the web that use a single transistor and a very simple 2coil inductor. You could google for "simple 12v fluoro schematic".
:)
 

Thread Starter

spacewrench

Joined Oct 5, 2009
58
D'oh! Use a transformer! Of course! Thanks!

I had gotten in my head to use a common-mode inductor, but wasn't making much progress figuring out how it would work.

I have extra PWM outputs on my micro (and code is free) so I could even set up an H-bridge on the primary side to get more swing there (since I want to run on 2xAAA cells).

Now I just have to wind a few:many transformer on one of the little ferrite toroids I have laying around...

(Here's the common-mode inductor circuit I was trying to figure out.)
 
Last edited:

Thread Starter

spacewrench

Joined Oct 5, 2009
58
My tests with hand-wound transformers have not been very successful (inadequate output voltage, even with input much higher than I will actually have in practice). Worse, I don't understand the problem enough to say which component I should attack first.

After additional Nixie testing, it looks like I need to generate 120-150VDC at (probably) 2-3mA, from 2xAAA batteries. I need to do this with as few additional components as possible (starting with a microcontroller with PWM peripheral that can clock at ~20MHz).

I'm trying to learn as much as possible on breadboard, before I commit to PCB, so any nudges would be welcome. In particular:

  • Single inductor (basic flyback) or transformer?
  • If transformer, what ratio, and about how many primary turns should I be looking for? What core material? (Keep in mind, this must be small, and can be hand-wound.)
  • What kind of diode?
  • Do I need a FET gate driver, or is GPIO pin OK?
  • If driver is necessary, is bipolar totem-pole the way to go?
  • What frequency range & duty cycle should I be looking at?
I know, for example, that higher frequency permits smaller inductor/transformer, but increases switching losses and makes greater demands on the diode. Also, at higher frequencies, my PWM resolution will not be as good. I would like to be able to adjust the output voltage with whatever PWM resolution I have, so that I can control the Nixie tube current. (Alternatively/equivalently, if the response time is fast enough, I could stutter the boost and let the output voltage cap decay to get the desired voltage.)

I just don't know which parameter will be most limiting, so I don't know where to start from. (Also, I REALLY don't understand transformers, so I wouldn't know the perfect part if it electrocuted me!)

Any [additional] help? Thanks!
 

BMorse

Joined Sep 26, 2009
2,675
try using an audio transformer (Smaller in size than most) I have used some in some of my DC to DC converter circuits where I use a 555 to pulse one side of the transformer (via a mosfet).... I have managed to get over 2k VDC with some simple voltage doubler circuits using some 1N4007 diodes and some 1KVDC capacitors.... I know its not what youre after but i just thought I would mention it....
 

SgtWookie

Joined Jul 17, 2007
22,230
Check out Ronald Dekker's "Flyback Converters for Dummies" page, here:
http://www.dos4ever.com/flyback/flyback.html

Even though you'll be better off using a transformer, Ronald's put together a pretty handy reference for things like this. And yes, when the current suddenly shot up, your inductor was saturating.

If you can wrap your toroid with some tape, it will help delay the onset of saturation.
 

Thread Starter

spacewrench

Joined Oct 5, 2009
58
Check out Ronald Dekker's "Flyback Converters for Dummies" page, here:
http://www.dos4ever.com/flyback/flyback.html
That's an excellent reference; he covers almost all the things I didn't understand. The inductor/transformer test circuit should make it much easier to figure out the characteristics of the parts I've been making.

Re: tape, would an epoxy-coated toroid be equally good? The toroids I've been using are pretty small to try to put tape, wire, more tape, and more wire onto.

Thanks!
 

SgtWookie

Joined Jul 17, 2007
22,230
Re: tape, would an epoxy-coated toroid be equally good? The toroids I've been using are pretty small to try to put tape, wire, more tape, and more wire onto.
Well, the basic idea of the tape is to provide a bit of separation between the windings and the toroid itself; so it'll add more bulk. The benefit is that the onset of saturation is delayed, and it isn't as "sharp" either.

I suppose that if you had an epoxy coating on the toroids that was several thousands of an inch thick, that may help.
 

THE_RB

Joined Feb 11, 2008
5,438
My tests with hand-wound transformers have not been very successful (inadequate output voltage, even with input much higher than I will actually have in practice). Worse, I don't understand the problem enough to say which component I should attack first.
...
Have you tried reversing the leads on the secondary winding??
 

SgtWookie

Joined Jul 17, 2007
22,230
What are you using for the MOSFET?

I'm thinking that with the low voltage supply you're dealing with, you might be better off with a high voltage bjt, like a 2N6517, using a base transistor around 1k Ohm, and a 1N4007 across the cathode & emitter to provide a current path from ground.
 

Thav

Joined Oct 13, 2009
82
My tests with hand-wound transformers have not been very successful (inadequate output voltage, even with input much higher than I will actually have in practice). Worse, I don't understand the problem enough to say which component I should attack first.

After additional Nixie testing, it looks like I need to generate 120-150VDC at (probably) 2-3mA, from 2xAAA batteries. I need to do this with as few additional components as possible (starting with a microcontroller with PWM peripheral that can clock at ~20MHz).

I'm trying to learn as much as possible on breadboard, before I commit to PCB, so any nudges would be welcome. In particular:

  • Single inductor (basic flyback) or transformer?
  • If transformer, what ratio, and about how many primary turns should I be looking for? What core material? (Keep in mind, this must be small, and can be hand-wound.)
  • What kind of diode?
  • Do I need a FET gate driver, or is GPIO pin OK?
  • If driver is necessary, is bipolar totem-pole the way to go?
  • What frequency range & duty cycle should I be looking at?
I know, for example, that higher frequency permits smaller inductor/transformer, but increases switching losses and makes greater demands on the diode. Also, at higher frequencies, my PWM resolution will not be as good. I would like to be able to adjust the output voltage with whatever PWM resolution I have, so that I can control the Nixie tube current. (Alternatively/equivalently, if the response time is fast enough, I could stutter the boost and let the output voltage cap decay to get the desired voltage.)

I just don't know which parameter will be most limiting, so I don't know where to start from. (Also, I REALLY don't understand transformers, so I wouldn't know the perfect part if it electrocuted me!)

Any [additional] help? Thanks!
I would recommend the isolated flyback for this, that way you get the transformer helping you to step up voltage. A concern I have is regarding your sense resistor. If you are triggering at 0.5-1V across that sense resistor you could be losing a lot of power in it and making your drive a little more sluggish because you'll be stealing that sense voltage away from the 3V you have available to drive.

You'll need to block ~125V with that diode. I would say go for a 200V, 1A silicon diode. Maybe an MUR series from ON Semiconductor. I think the part number would be MUR120.

Definitely build a gate driver for that FET! If you're planning to switch at 32kHz you will certainly want it. There are some cute little 20V FETs available from IR that have a miller plateau at ~2V, so you should be able to drive them to the correct VOLTAGE with your micro, but they will likely switch very slowly. A simple enough way to do this is a BJT totem pole driver with maybe 100Ω base resistor.

This would be an inverting totem pole. You can switch the NPN and PNP to get non-inverting without too much trouble. a 2n2222 and 2n2907 should be fine. Get em with the A suffixes if you can.

Ah, I see you knew about the totem pole already. Well there's the info anyway.

Leave yourself some space for an RC snubber (just a series R and C) around your FET. You may find that the leakage inductance from the transformer causes a sudden spike of voltage at the drain of your FET. The snubber can help you control the peak of that voltage so your FET doesn't avalanche. Some FETs are avalanche rated though so you might take advantage of that.

I have no reasoning behind this, but 50% duty cycle is nice to shoot for. The lower your duty cycle the easier it is to reset the transformer core.

Reviewing this post I see that this design can be quite involved!
 

Thread Starter

spacewrench

Joined Oct 5, 2009
58
THE RB said:
Have you tried reversing the leads on the secondary winding?
Yes, but no significant change in output. I see clipped ringing when the FET turns off, where the clipping appears to be at the output voltage level. Since a ring goes both directions, it looks like you can clip a little energy off either side to charge the output cap. Obviously, if you clip the first half-cycle (and then subsequent odd half-cycles) you'll do a little better, but even if you clip the even half-cycles, you get some boost.
SgtWookie said:
What are you using for the MOSFET?
It's a TN5325. (I haven't started trying to boost 3V yet, I'm still using 12V into the inductor, because I wanted to learn whatever I could about the circuit before getting wrapped up with things that change at low voltage.) But you're right: a bipolar might give me a lower effective resistance to ground, if Vce-sat is less than Ilsat * Rdson. And if it can switch off as quickly.
Thav said:
If you are triggering at 0.5-1V across that sense resistor you could be losing a lot of power in it and making your drive a little more sluggish because you'll be stealing that sense voltage away from the 3V you have available to drive.
Yes, my sense resistor is too large for use with a 3V supply, but I think it's OK with the 12V I'm using now. It also makes it easier to see the inductor current ramp up, which is what I'm looking at now.
In the final circuit, I expect I'll just measure my inductor carefully and run without a sense resistor at all, just limiting inductor current by having a maximum pulse width.
Thanks for the followups, everybody.
 
Top