Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Attempt at a simple boost converter (3V--> 5V)

Status
Not open for further replies.

carbonzit

Active Member
Note: Long post. You might want to get a [fill in name of preferred drink here] before you start this.

OK, that's better. So this is part of my continuing quest to come up with a cheap, simple 3-volt-to-5-volt converter. Think of it as a poor man's Minty Boost. (I would actually like it to be able to supply 9 volts, but so far that seems an unattainable goal.)

So I was reading up on SMPSs, and once more came across the description of classic boost converter topology:

**broken link removed**

I thought to myself: "Self, why couldn't one drive this with a simple transistor oscillator, rather than a fancy-schmancy chip? Why not our favorite li'l oscillator, the relaxation oscillator (aka free-running multivibrator?"

Why not indeed? So that's exactly what I did. Here it is (the LTspice file is attached below):

**broken link removed**

It not only works in LTspice simulation, but also in real life. (Slightly different results, but never mind.) Of course, it comes nowhere near the Minty Boost, which can deliver a hefty 500mA at 5 volts; the best result I was able to get was 6.5 volts at 18mA (with a 390Ω load resistor). The best I could get from the LTspice simulation was 7.7V @ 19.6mA.

I think this idea has promise. Of course, this is just a starting point, if indeed there is some better ending point out there. After all, it is totally unregulated in its current form. Just raw output. (That may be sufficient for some applications that don't require good voltage regulation.) That's one of the issues I ask about below.

So comparing this to the Minty Boost, I wonder exactly why this is so much more wimpy. After all, if you strip down the Minty Boost to its essentials, you get this:

**broken link removed**

Looks familiar, no? The only difference is that the main works (oscillator and switch) are inside that little chip.

So why can't I achieve at least close to its performance with a more humble circuit?

There are several parameters that are critical to the operation of a boost converter:

  • Switching frequency (the multivibrator frequency in this case)
  • The "on time" of the switching waveform (in other words, the duty cycle). See below for more on this.
  • The inductor value

to name the main ones.

Having no way to measure frequency here, I have to rely on the simulation as my guide. LTspice tells me that the switch is being driven at about 300kHz, a respectably-high frequency. But of course, one question is whether the on-time parameter is really what it should be. Especially since the multivibrator waveform (again, the simulated waveform) looks like this:

**broken link removed**

Not very symmetrical. Now, it would be interesting to learn what effect making the multivibrator "lope" would have on the converter performance--that is, making the two times unequal (different R or C sizes). Any thoughts on this?

The inductor value also has me mystified. According to my book on power supplies, the formula for the inductor value is

L = Vin * Ton / Ipk

which gives me a value somewhere in the millihenry range, not microhenries. (Using Ton ~= 2.3µS.) (Here, we're supposed to size the inductor based on the minimum current, which makes it larger.)

So which way should I go here: where should I be frequency-wise? What's the best way to match the inductor to the switching frequency? How about using something beside a BJT for the switch: MOSFET? something else?

Thinking ahead, it occurs to me that instead of just sticking a 7805 in there to waste some electrons and give me 5 volts, maybe I could come up with a quick-and-dirty regulation scheme, like frequency-modulating the multivibrator (basically implementing some kind of VCO). Possible?

The fact that this works as well as it does is intriguing to me, and leads me to belive that there's much more potential here. Now imagine that this could be cranked up to deliver 9 volts: that would be a really nice way of eliminating 9-volt batteries in small electronic devices.
 

Attachments

  • Image3.gif
    Image3.gif
    27.6 KB · Views: 1,212
  • Boost conv 1.asc
    2.5 KB · Views: 451
Last edited:
The switch transistor is not being driven hard enough to get a decent output.

The RHS of the multivibrator is only being allowed to reach 0.6V, therefore the uneven duty cycle.

Add a buffer to the multivibrator output and you'll see an improvement - even just decreasing R3 and inserting a base resistor onto Q3 (as per attachment) will give an improvement over your reported results (better than 9V out, but I didn't bother waiting for it to stabilise).
 

Attachments

  • Boost conv 1x.asc
    2.6 KB · Views: 401
Sorry, I glossed over most of that...

Why not our favorite li'l oscillator, the relaxation oscillator (aka free-running multivibrator?"
Are you sure it's a relaxation oscillator?

There are several parameters that are critical to the operation of a boost converter:
...to name the main ones.
Peak current and switch on resistance / voltage is also important.

The inductor value also has me mystified. According to my book on power supplies, the formula for the inductor value is ...

Here, we're supposed to size the inductor based on the minimum current, which makes it larger.)
Why are you using the minimum current? Personally, I'd use the maximum inductor current. That's what Ipk is, surely.

How about using something beside a BJT for the switch: MOSFET? something else?
A mosfet would be good. Get one with a low enough threshold voltage that you can turn it on from your 3V power source. Also make sure it has an appropriately low RDSon.

Thinking ahead, it occurs to me that instead of just sticking a 7805 in there to waste some electrons and give me 5 volts, maybe I could come up with a quick-and-dirty regulation scheme, like frequency-modulating the multivibrator (basically implementing some kind of VCO). Possible?
yuk. If you actually did use a relaxation oscillator, for example, you could hold it in reset while the output was too high and inadvertently implement PFM or 'fixed On time' control.

Now imagine that this could be cranked up to deliver 9 volts: that would be a really nice way of eliminating 9-volt batteries in small electronic devices.
The minty boost can provide 9V (in fact, it should be able to go up around 24V) if a resistor divider is added before the LT1302 sense pin.
 
Are you sure it's a relaxation oscillator?
[/quote]

I guess that's my bad. For some reason I had it in my head that the 2-transistor multivib. was a relaxation osc. So a relaxation oscillator produces a sawtooth wave, right? More questions on this below ...

There are several parameters that are critical to the operation of a boost converter:
...to name the main ones.
Peak current and switch on resistance / voltage is also important.

Yes, on resistance and voltage: does this mean the voltage at which the switching device turns on? I need to read more about this.
And I'm confused about "peak current". I thought this was a design goal, not a circuit parameter, but I guess not.

Why are you using the minimum current? Personally, I'd use the maximum inductor current. That's what Ipk is, surely.

My book says:

Knowing the peak current, t(on), and the voltage(s), the inductance can be calculated. The inductance value for continuous current in the inductor is chosen based on the minimum current that the switching regulator must handle--the lower the current, the larger the inductance.

Thinking ahead, it occurs to me that instead of just sticking a 7805 in there to waste some electrons and give me 5 volts, maybe I could come up with a quick-and-dirty regulation scheme, like frequency-modulating the multivibrator (basically implementing some kind of VCO). Possible?
yuk. If you actually did use a relaxation oscillator, for example, you could hold it in reset while the output was too high and inadvertently implement PFM or 'fixed On time' control.

Questions about the oscillator, which (surprise, surprise) actually seems to be a crucial part of this whole deal:

1. Is there a better/different way of implementing the oscillator? We need square waves (or at least square-ish), so it seems a relaxation oscillator wouldn't work; or would it? maybe followed by a squaring stage? (Remember, discretes only, no chips. If I wanted chips I'd just build the damn Minty Boost!)

2. Look at the attached simulator output below and check out just how ugly the multivibrator output gets after running for a while. The "hash" consists of large negative-going spikes, as well as some ringing around 2 mHz. What's up with that? While I suppose I could live with it, I get the sneaking suspicion that this is robbing power that could otherwise be sent to the output. Not to mention the chance of a very rich source of RFI ... anyhow, I think this would be worth fixing before proceeding too much further. Not good behavior.

3. I still think it's worthwhile playing around with the duty cycle of the MV. Any suggestions on this?

4. And while PFM, or something similar, is kind of pie in the sky at this point, it's still something I'd like to see if it can be implemented, even if in a crude way.

The minty boost can provide 9V (in fact, it should be able to go up around 24V) if a resistor divider is added before the LT1302 sense pin.

So I implemented your suggestions (see new attached .asc file), and was pleasantly surprised at the improvement. Not an order of magnitude, but very noticeable.

I changed the MV collector resistor to 470 as well as the drive base resistor. I also changed the MV caps to 220 pF, bringing the MV frequency to 75 kHz. In the simulation, I get 11.7 volts at almost 30 mA. (My breadboarded circuit isn't quite so good, since I only have a 100uH choke instead of a 50, but it's definitely better than before too.) So I still think there's more power to be coaxed out of this little device.

But it's getting too late now, so I'll have to pick this up later ...
 

Attachments

  • Boost conv 1A.asc
    2.6 KB · Views: 341
  • Image5.gif
    Image5.gif
    19.5 KB · Views: 583
Last edited:
Yes, on resistance and voltage: does this mean the voltage at which the switching device turns on?
I meant on resistance of the [mosfet] switch or on/saturation voltage of the [bjt] switch. You will of course need sufficient voltage to operate the switch, but that's not what I was meaning.

My book says: ... for continuous conduction mode...
The minimum current they're referring is for operation in continuous conduction mode. It's OK to operate in discontinuous mode, especially if there's not much load on the output. **broken link removed**

1. Is there a better/different way of implementing the oscillator? We need square waves (or at least square-ish), so it seems a relaxation oscillator wouldn't work; or would it? maybe followed by a squaring stage?
Different: You can use the joule-thief type blocking oscillator you've already been playing with. It has the advantage that it sets its own frequency based on the saturation limit of the coil or transistor. Just add a means of regulation and you're done.

As for the spikes at the base of the switch, I don't know what trouble they'd cause. You could clamp them with a diode I guess...

The ringing on the collector can be suppressed by using an RC snubber across the diode (e.g. a starting value might be 1n/10R)..

PFM is not that fancy - just set up the oscillator to go for the maximum expected load, and whenever the output is too high, force the oscillator to the idle state.
 
One problem with your oscillator is that good power transfer requires the right duty cycle to suit Vin and Vout. You have a fixed duty oscillator and no feedback.

The best bet with a simple design is to run the oscillator duty set for high power transfer, then regulate it by reducing duty when the output voltage is reached (which is pretty much what Dougy83 said).

Or an even simpler system regulate at a fixed inductor current like this 2-transistor boost converter;
https://www.romanblack.com/smps/conv.htm
which gives good power transfer as inductor current is well controlled, but runs at a fixed input power so it's not ideal for battery use.
 
Last edited:
Hi,

Another little factor is the ability of the transistor to remain in saturation for the entire duration of the 'on' pulse. If it starts to pull out of sat a simple fix might be to increase the inductor value to 200uH. That lowers the peak current and the transistor may then make it just the way it is driven now.
Another factor is the inductor series resistance. 2 ohms is very high for these kinds of circuits unless you have a low output current. Try 0.1 ohms instead, but make sure the transistor stays in sat for the full 'on' time duration.
 
.....Different: You can use the joule-thief type blocking oscillator you've already been playing with. It has the advantage that it sets its own frequency based on the saturation limit of the coil or transistor. Just add a means of regulation and you're done.
Aye, there's the rub. There's no easy way to regulate the output voltage of a joule-thief other than using a lossy linear regulator.
 
Another factor is the inductor series resistance. 2 ohms is very high for these kinds of circuits unless you have a low output current. Try 0.1 ohms instead, but make sure the transistor stays in sat for the full 'on' time duration.

That's a good point, about the inductor resistance. I actually made some attempts to make this circuit conform to reality by checking available inductors, using Digi-Key's parameterized search facility. Here's what I found, in the range of interest for this project:

50µH: min. resistance: 12 mΩ

100µH: min. resistance: 84 mΩ

The problem is that the low-resistance inductors get pretty expensive, which violates another of the parameters of this experiment (those damned parameters again!). Turns out the inductor I'm testing with (100 µH) has a resistance of 8Ω, which is certainly limiting the results I'm getting.

Aye, there's the rub. There's no easy way to regulate the output voltage of a joule-thief other than using a lossy linear regulator.

Yes, I think for the time being I'll stay away from the blocking oscillator.

I added a buffer/amplifier stage before the switch, as suggested by Dougy (.asc file attached) and saw another fairly big jump in performance. Maximum output was now 12.6V @ 32mA. Not bad; now we're talking some real usable power here. (In simulation, of course.)

The buffer changes the shape of the driving pulses (pic attached), which is now both better and more problematic. Better because I now see something approaching a clean 50% duty-cycle square wave on the tops of the signal, but the bottoms are still spiky and ringy. So as Dougy suggested, there is the solution of simply clamping away these faults, but that seems like a brute-force method. Is there anything else I can do to my setup to minimize this stuff? But more importantly, should I really be concerned about the "ugly" look of this waveform? does it really matter to the operation of the switch? or is it good enough? (And then there's still the unresolved question of duty cycle. Yes, I know, this is going to take more research on my part. But hey, if you have any suggestions ...)

The base of the switch is now being driven pretty hard (~1.5V). Is this hard enough?

Also, I had a hell of a time setting up the buffer. The only configuration I could get to work was this emitter follower. Anyone care to give me a just-above-noob-level critique on this? is there a better configuration? different resistor values?

Ideally I'd be able to drive the switch directly from the output of the MV. Is there any way to do this by rejiggering things here, without adding another transistor? (I get the sinking feeling that every transistor I add is just drawing more precious current from my battery source!)
 

Attachments

  • Image2.gif
    Image2.gif
    15 KB · Views: 576
  • Boost conv 1B.asc
    3.1 KB · Views: 352
3V--> 9V converter

By the way, I'm officially changing the topic to 3V--> 9V converter, since that was my real goal all along, and the project seems well capable of that.
 
Hi again,

Change your 470 ohm resistors to 270 and use a 200uH inductor and you'll get at least 150ma output.

What does your collector voltage waveform look like? Lets see if it stays in sat or inductor runs out of juice :)
 
Thanks, Mr Al. I'll do just that. Just wanted to report the latest results from Carbonzit Laboratories, GmbH, where changing the current inductor (50uH) resistance to 0.1 ohm gave an output of 15.6V @ 40mA.

Onward and upward!
 
Change your 470 ohm resistors to 270 and use a 200uH inductor and you'll get at least 150ma output.

I don't think so. I made those changes, and these are the results I got:

R-------L-------V-------I
270----200----12.4----31.98
470----200----12.1----31.2
270----50-----15.7----40

So the 200µH actually makes it worse; 50µH seems to be the optimal value. 270Ω seems gives slightly more voltage and less current. (This is all according to the simulation, of course.) Remember, my previous "best" was 15.6V @ 40.2mA.

So why did you suggest those changes? You didn't simulate them, I take it?

What does your collector voltage waveform look like? Lets see if it stays in sat or inductor runs out of juice :)

See attached pic of collector voltage. As you can see, it's fairly narrow pulses. I don't know how to interpret this: is this good or bad? how can you tell if the transistor is in our out of saturation here? (The small purple waveform at the bottom is the switch base drive.)

I'd still like to get some answers about this part of the circuit:

**broken link removed**

I admit I'm still pretty clueless about how to design such stuff. Questions:

  • Should this be an emitter follower? If so, I'm guessing that's because Q3 presents a fairly low impedance looking into its base?
  • What about the resistor placements and values? Close? No cigar?
  • Why can't I just use the multivibrator to drive the switch directly? Not enough "oomph"? Need some wave-shaping action?
  • Different transistor here?

I think I might have a MOSFET lying around here somewhere; that's on the to-do list to try instead of the BC337.
 

Attachments

  • Collector voltage.gif
    Collector voltage.gif
    20.9 KB · Views: 455
Hi,

Your inductor is running out of steam. That's why the pulse looks so narrow. If it runs out of energy it cant produce current anymore.
You can use the multivib to drive the transistor as you did, but you can only expect so much out of that unless you lower impedances or use another drive transistor.
The main idea is to make sure the inductor gets the full battery voltage for the required time that the transistor is on, and that the transistor stays on for the whole time and does not pull itself out of saturation because of high collector current vs base drive, and also to make sure the inductor has enough energy to produce the required output voltage. There are some formulas if you are interested. With 39 ohm load you can get 150ma, but im not sure what output you want to get there.

To get more current you need to reduce your load. Try 39 ohms. If you still have a problem, i'll post the whole circuit with only the change of 270 ohms, 200uH, and 39 ohms and a little series resistance for the real world battery simulation.
 
Last edited:
The main idea is to make sure the inductor gets the full battery voltage for the required time that the transistor is on, and that the transistor stays on for the whole time and does not pull itself out of saturation because of high collector current vs base drive, and also to make sure the inductor has enough energy to produce the required output voltage. There are some formulas if you are interested.

Yes! I'd love to see some formulae! Please, bring them on!

With 39 ohm load you can get 150ma, but im not sure what output you want to get there.

To get more current you need to reduce your load. Try 39 ohms. If you still have a problem, i'll post the whole circuit with only the change of 270 ohms, 200uH, and 39 ohms and a little series resistance for the real world battery simulation.

Already done that. Here are the results:

Rl-------V------I
40Ω----7.7----194
50Ω----8.1----163
75Ω----8.7----116
100Ω---9.2-----92

So we're starting to see this thing's V-I profile.

Looks like a pretty respectable 9-volt converter with a 100Ω load ...

************************************************************

Hmm, with your 200µH, I get 9.7V @ 97mA. That's nearly 1 watt.
 
Last edited:
Aye, there's the rub. There's no easy way to regulate the output voltage of a joule-thief other than using a lossy linear regulator.

I'm not sure that's true.

You could a regulator transistor that turns the joule thief off when the output is over X volts, so assuming the JT (joule thief) has good reliable startup it will regulate fine.

Or the base of the JT could be decoupled somewhat from the transformer by a resistor, so the transformer becomes part of the base drive (not all of the base drive) then a regulator transistor can be used to bias the base "more off" when output voltage is >X volts. That should keep the JT oscillating but reduce duty cycle.

Personally I still like the multivibrator design as it can give good square wave switching and will always oscillate well regardless of what the output is doing. I would try to build a 2 transistor "power multivibrator" with the BC337 as one of the 2 transistors, and get the desired 9v at full current output. Then a 3rd transistor as the output voltage regulator.
 
Personally I still like the multivibrator design as it can give good square wave switching and will always oscillate well regardless of what the output is doing. I would try to build a 2 transistor "power multivibrator" with the BC337 as one of the 2 transistors, and get the desired 9v at full current output. Then a 3rd transistor as the output voltage regulator.

I'm intrigued by that idea (I like the multivibrator too). Could you (or someone else) maybe just sketch such a circuit so we could see what it might look like?

Regarding Regulation

It seems to be coming to that part of the project where I should start to think seriously about regulation.

Since the converter is supposed to be as simple (some might say primitive!) as possible, the following scheme occurs to me:

**broken link removed**

Should the output voltage rise over the setpoint, the zener conducts, turning on the transistor, which then ... [does something, not sure what] to slow/stop/delay the multivibrator.

Precision regulation is not even a design goal here. Close enough is good enough.
 
Aye, there's the rub. There's no easy way to regulate the output voltage of a joule-thief other than using a lossy linear regulator.
There's a bunch of ways to regulate a joule-thief type converter. A couple might be:

1) Colin has posted one sometime back that shunts the oscillator transistor's base current as a means of regulation, something along the lines of this:
**broken link removed**
It's not the exact the circuit I was thinking of). It's not my favourite method, due to the fact that the shunting wastes energy. By inverting the logic to control the sourcing of the base current to the oscillator rather than diverting it that problem is removed.

EDIT: actually, the attachment is the one I was thinking of (is this circuit in carbonzit's posting anyway?)


2) adding an extra winding and rectifier which connects from GND to VIN, which will regulate the output based on the input voltage. The inductor energy during regulation is returned to the input capacitor so it's not completely wasted.

EDIT: 2nd attachment shows this basic concept
 

Attachments

  • 3-9 converter.png
    3-9 converter.png
    6.5 KB · Views: 766
  • resetreg.png
    resetreg.png
    3.4 KB · Views: 914
Last edited:
The circuit attached below stops my multivibrator dead in its tracks, with 1 volt to the base of a controlling transistor. So maybe this could be the basis of a regulator. Now, this is a pretty gross way of controlling things: the MV either runs full throttle or comes to a screeching halt. Don't know how this would work out in a running boost converter.

I'm thinking I could couple this to the circuit snippet in post # 17 above. I was hoping for a more nuanced way of controlling the MV, like some way to alter the frequency (FM/VCO), wave amplitude or something else with some ... gradation, instead of just ON/OFF.

Comments? Am I nuts? (Don't answer that!)
 

Attachments

  • MV reg.asc
    2.2 KB · Views: 325
Yes! I'd love to see some formulae! Please, bring them on!



Already done that. Here are the results:

Rl-------V------I
40Ω----7.7----194
50Ω----8.1----163
75Ω----8.7----116
100Ω---9.2-----92

So we're starting to see this thing's V-I profile.

Looks like a pretty respectable 9-volt converter with a 100Ω load ...

************************************************************

Hmm, with your 200µH, I get 9.7V @ 97mA. That's nearly 1 watt.


Hi again,

The best place to start with any converter is to decide what input you are going to be working with and what output you need in terms of voltage and current.

The first formula would be the basic input/output relationship between voltages. The output voltage comes from the input voltage and the duty cycle:

Vout=Vin/(1-D)

where D is the transistor 'on' duty cycle expressed as a fraction (0.1 is 10 percent for example).

From this we can see that as the transistor stays on longer and longer the output voltage rises higher and higher. There's a limit to this, but for now we dont think about that.

With say a 50 percent duty cycle this would make D=0.5, so the output voltage would be twice the input voltage:
Vout=Vin/(1-0.5)=Vin/0.5=2*Vin

For now we ignore certain smaller quantities such as the diode drop and transistor sat voltage, but they will come in later.


There are various ways to introduce voltage or current regulation. What makes or breaks these converters though is their efficiency.
 
Last edited:
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top