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

[carbonzit]

OK, that one I already had (Vo = Vin/1 - D). So working backwards from my empirical results (3V in, ~9V out), we can see that I have a duty cycle of 2/3. Not bad (>50%) for such a crude circuit, eh?

Here are the other formulae I have:

Inductor size: L = (Vin Ton)/Ipk
Peak current: Ipk = 2 Iload (Vo/Vin)
Switching frequency: Td = (L Ipk)/(Vo - Vin)
Td = (Vin Ton)/ (Vo - Vin)
(where Ton = T - Td)

So do you have any comments about any of the circuit aspects I asked about? Voltage regulation? eliminating intermediate transistor?

 

[MrAl]

Hi,

I didnt realize you were after a min parts count circuit, so you may want to take a look at this circuit first...

To make it better, remove the LED and use a Schottky diode and capacitor with load as in your other circuit.
You can decrease R2 to increase output current.

 

[Mr RB]

Hi, as you asked; "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?"

I would build the "power multivibrator" like below with all the parts but not the regulator parts seen in grey.

It should be easy enough to get it running at the desired full power level into the 9v load.

Once that is done, add the regulator parts. The regulator works as follows;
1. if the output is too high the REG transistor turns on
2. the REG tran turns off the main switch BC337
3. when BC337 is off, the 2N2222 must be on, this is a stable condition that will continue for any amount of time
4. when output voltage drops, REG turns off, BC337 must turn back on, so oscillation is guaranteed to resume (I hope)

The result is that the regulator will increase the OFF period by as much as needed to maintain the correct output voltage. The ON period should be retty much fixed, determined by the RC at the 2N2222 base.

 

[carbonzit]

Hi, as you asked; "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?"

I would build the "power multivibrator" like below with all the parts but not the regulator parts seen in grey.

It should be easy enough to get it running at the desired full power level into the 9v load.

Hey, thanks, Mr RB. Funny, before I read your reply, I had just cobbled together almost the same thing in LTspice and tried it (.asc file posted below for those who can follow along with LTspice; I've posted a picture for those who can't).

**broken link removed**

Unfortunately, it doesn't work. Well, it sorta kinda works at first, but it poops out after about 6 milliseconds of seemingly healthy multivibrating. Can anyone tell me why? Probably not a simple answer, I'm guessing. It seems like a lot to ask a transistor to play two roles like this.

One problem seems to be that the switch transistor base is only being driven at about 0.7 volts (measured at "sw_in"). How can I get more oomph into Q3's base?

I took a slightly different tack than you and made both transistors the same. I'll have to try your circuit and see how it works.

Now I'm really intrigued ...

 

[carbonzit]

Hmm; I made a few changes (changed the left transistor to a 2N2222, reduced its collector resistor to 220Ω. The result is that the whole thing now hums happily along; the multivibrator multivibrates, and the boost regulator works. But it only gives a paltry 3.2V (it's just barely boosting).

So take a look at the attached plots for the multivibrator outputs and the switch input (base voltage). It looks like the other multivibrator output--that is, the one that's not connected to the switch (mv1)--is the one we want (much higher duty cycle). The one driving the switch (mv2) is only on a small fraction of the time. And the actual base drive waveform isn't even close to a square wave; it's kind of a sawtooth wave with the teeth ground down, or something like that.

 

[carbonzit]

Working! (@ 5 volts, anyway)

So this circuit actually works! (in simulation, at least):

**broken link removed**

Sorry for messy schematic. You'll notice I added a resistor (2.2K) across the switch's base capacitor to "push" more voltage (or is that current?) through to the base. It seems to have worked.

According to LTspice:

Output voltage: 5.5V
Output current: 55 mA
Switch base drive: 0.8V (still not enough?)

The output waveforms are very choppy and chaotic (run the simulation yourself to see). But it works!

(Note: I didn't change the filename from the previous circuit, in case that matters. I'll start doing so to reduce confusion.)

 

[Mr RB]

Why have you made R4 so high? Assuming the multivibrator makes a stable squarewave (which you want) then R4 is supplying the base drive for Q1 for the entire ON time of the square wave.

Try about 1k for R4. The about 1k for R1 (like in my schematic in post #23). Then lose R3, it is not needed. Also that diode I added on the transistor base stops it blowing up...

I would also make C1 and C2 larger, until it is running about 20 to 30kHz. Aim for about 50:50 duty and a good solid squarewave, that should be a decent starting point to get some power transfered at 9v.

 

[MrAl]

So this circuit actually works! (in simulation, at least):

**broken link removed**

Sorry for messy schematic. You'll notice I added a resistor (2.2K) across the switch's base capacitor to "push" more voltage (or is that current?) through to the base. It seems to have worked.

According to LTspice:

Output voltage: 5.5V
Output current: 55 mA
Switch base drive: 0.8V (still not enough?)

The output waveforms are very choppy and chaotic (run the simulation yourself to see). But it works!

(Note: I didn't change the filename from the previous circuit, in case that matters. I'll start doing so to reduce confusion.)

Hi,

Did you try the circuit i posted in post #22? That's pretty much a min parts count circuit. You'll need the output diode and cap instead of the LED.

 

[carbonzit]

Did you try the circuit i posted in post #22? That's pretty much a min parts count circuit. You'll need the output diode and cap instead of the LED.

Sorry, Mr Al: I didn't want you to think I was ignoring you.

Yes, I did play around with your very interesting circuit. In fact, I simulated it (.asc file below). Here's what I started with:

**broken link removed**

You'll notice I made a small change (replaced Q2 with my trusty BC337). Everything else is the same, except of course for the output capacitor and load resistor.

I took the liberty of redrawing it so it made a little more sense, at which point I could see the classic boost topology. (Sometimes it makes a difference how we draw our schematics.)

**broken link removed**

When I ran it, I got rather surprising results. First of all, I couldn't use anything larger than that 01.µF cap; it just wouldn't work with anything larger. I wondered if a decent-size capacitor is just too low-impedance a load for it?

When it worked, it seemed to want to produce ~20 volts at ~18 mA. Not too shabby. But then I took a closer look at some big spikes I noticed on the voltage plot:

**broken link removed**

Yikes! Your circuit is a little scary. What's up with those super-spikes?

But seriously, do you think this could be tweaked to behave a little better? and maybe produce more current at a reasonable voltage (like, oh, I don't know, 9 volts)?

What kind of oscillator is that, exactly? I assume that Q1 is doing the oscillating and the other one is doing the switching.

Very interesting. I'm going to keep it around. I think I want to concentrate on my "power multivibrator", but this is definitely worth investigating.

 

[carbonzit]

Progress with the "power multivibrator"

Continuing with Mr RB's suggestions, I simulated the following circuit (.asc attached):

**broken link removed**

It gave me a reliable 5.5 volts @ 55 mA, so I guess if I were aiming at a pretty good 3V--> 5V converter I might feel close to done.

But I really do want 9 volts, and more current if I can get it. What's holding me back here?

You'll notice I left out your base diode. Looking at the plot of base voltage, only a very small part of it goes negative, so it doesn't seem that the transistor is in any immediate danger of getting zapped. What do you think? (See attached plot below.) I am aiming for minimal parts count too.

Regarding that base drive waveform, which seems to be all-important to the operation of the boost converter, I wonder if it's really what's wanted here. On the one hand, it looks to have an extremely high duty cycle, which is what we want, right? but on the other hand, it's not really a square wave, and the voltage rises rather precipitously after falling off that cliff.

I also reduced the inductor value, since the circuit is now operating at about ... yikes, I just checked it, thinking it should be about 300-something kHz, but it looks more like almost 500. So more questions (sorry, there's no end to them!): one can either reduce the frequency, by making the capacitors bigger, or reduce the inductor size, which seemed to work well here (the "sweet spot" seemed to be somewhere around 50-100µH). How exactly does one go about making such decisions (frequency, cap size, inductor size, etc., etc.). Seems like a very confusing multi-dimensional choice to make.

I also reduced Q1's collector resistor to 100Ω, and put in another 100Ω resistor to help isolate the oscillator from the switch. Seemed to help, anyhow.

 

[The Electrician]

But I really do want 9 volts, and more current if I can get it. What's holding me back here?

How much current do you want? Do you want the 9V output of the converter to be tightly regulated, or would the regulation you would normally get from a 9V battery anyway be good enough?

 

[carbonzit]

How much current do you want?

Good question. Can I say I want 500mA because that's what you can get out of a Minty Boost? Well, maybe not quite that much. Let's say something in the neighborhood of 100-200mA would be nice. (But of course I'll take half-an-amp if I can get it!)

Do you want the 9V output of the converter to be tightly regulated, or would the regulation you would normally get from a 9V battery anyway be good enough?

Since I see this as kind of a 9-volt "battery eliminator" (remember those?), then the latter would be OK. Seeing how simple and primitive this is, I'm not seeking overly-tight regulation. Some regulation would be nice (at least to protect light loads from overvoltage).

At this point, focusing on regulation would really be putting the cart before the horse; I want to get a good robust supply first, one that can respond well to regulation. So if you have any ideas how I can coax more juice out of this (either the simpler circuit or the original one), I'm interested.

 

[carbonzit]

Further progress

I'm actually kinda glad I didn't get any responses to those latest postings. Gave me a chance to sit down and continue with simulations. With this as the latest result:

**broken link removed**

Results as posted in the text box (output 9.8V @ 98 mA).

Still not quite what I want: with an eye to some type of regulation, I'd like to be able to get a solid 12 volts out of this to have something to regulate down from. After all, in the real world the battery won't be giving 3 volts.

The other thing is the negative-going spikes on the base of Q2 (notice renumbering since last schematics). Should I be worried about this (because of wasted energy, possible damage to the transistor or both)?

The frequency of operation seems alarmingly high. However, when I tried bringing it down (by increasing the sizes of the capacitors), strange things started happening; the output started "wobbling" badly, showing signs of instability. With these values, both the voltage and current waveforms are rock-steady. And of course, at lower frequency I'd have to use a larger inductor (right?).

Speaking of wobbling, you'll notice that I made the multivibrator non-symmetrical by making one capacitor larger, with the goal of increasing the duty cycle of the switch. It had a very noticeable positive effect.

By the way, as I type this I'm running another simulation with a larger load resistor (470Ω), and so far I'm getting an output of almost 20 volts at over 40 mA.

 

[MrAl]

Sorry, Mr Al: I didn't want you to think I was ignoring you.

Yes, I did play around with your very interesting circuit. In fact, I simulated it (.asc file below). Here's what I started with:

**broken link removed**

You'll notice I made a small change (replaced Q2 with my trusty BC337). Everything else is the same, except of course for the output capacitor and load resistor.

I took the liberty of redrawing it so it made a little more sense, at which point I could see the classic boost topology. (Sometimes it makes a difference how we draw our schematics.)

**broken link removed**

When I ran it, I got rather surprising results. First of all, I couldn't use anything larger than that 01.µF cap; it just wouldn't work with anything larger. I wondered if a decent-size capacitor is just too low-impedance a load for it?

When it worked, it seemed to want to produce ~20 volts at ~18 mA. Not too shabby. But then I took a closer look at some big spikes I noticed on the voltage plot:

**broken link removed**

Yikes! Your circuit is a little scary. What's up with those super-spikes?

But seriously, do you think this could be tweaked to behave a little better? and maybe produce more current at a reasonable voltage (like, oh, I don't know, 9 volts)?

What kind of oscillator is that, exactly? I assume that Q1 is doing the oscillating and the other one is doing the switching.

Very interesting. I'm going to keep it around. I think I want to concentrate on my "power multivibrator", but this is definitely worth investigating.

Hello there Carbon,


Yes those spikes are very strange, the only possible explanation is that they are signals coming from the distant Andromeda Galaxy from 2.5 million years ago from an alien hobbyist working on this very same circuit through a process commonly known as "deep space circuit signal injection" (DSCSI).

Either that or the circuit doesnt have a ground

These circuit simulators dont work very well without a ground. Try that and see if it starts rejecting distant signals from space

Once you get it working you can lower the 2.2k resistor to get more output current, to a point. Start with say a 100 ohm resistor as is, but you may need the original transistor to get best results.
For higher current outputs you may have to go to a Zetex NPN transistor.

 

[Mr RB]

Hi Carbonzit you are doing good so far.

**broken link removed**

(Re above) The first thing to do is increase the inductor to about 470uH (500uH you had before). That will allow the lower operating frequency with the larger caps.

I also don't think R1 needs to be as low as 220, it only needs to be a similar value to R4, and a value of 1k will give you better energy efficiency later.

R2 is your friend, it controls the ON time period, so a pot here can quickly help you tune the duty and freq for testing.

D2 can be a cheap 1N4148, no need for special schottky diode there.

The -1.2v pulses at Q2 base should be ok.

I would tune the thing with a 470uH inductor for 20 to 30 kHz and about 60% ON duty. At that point you should get a pretty decent current at 9v.

Also you might want to re-think the amount of output current you need, if this is a "9v battery replacer" then most devices running off 9v battery are designed to only need a few mA, so 20 or 30mA max should be plenty! At least consider the option of making one circuit that works well (high efficiency) from 0mA to 30mA or even less, and if you need a high current one make a second set of values for 0mA to 100mA. There are very few things that will draw 100mA from a 9v battery!

I would tune it to do as high an efficiency as possible into 9v 30mA, with input from 3.3v down to 2.4v. Then it's time to add the regulator transistor...

 

[carbonzit]

Mr Al's boost converter

Well, I played with it some more, and this is the best I could get:

**broken link removed**

Results as shown (Vo: 10.8V; Io: 49.1mA @ Rl = 220Ω).

Not bad. But I have to say I'm very skeptical of this circuit. Change one component--and not by very much--and it fails to start at all. I actually found best results with a plain Jane 2N2222. If someone can come up with better results I'm prepared to be beaten.

Yes, the absence of a ground may have allowed the circuit to pick up DSCSI, as you described. Funny that LTspice didn't warn me about having no ground as it usually does.

You never answered my question about the oscillator: how exactly does it work?

Do you have LTspice (or another Spice) so you can try my simulation?

 

[dougy83]

I'm actually kinda glad I didn't get any responses to those latest postings.

Always a pleasure

how exactly does it work?

Well, not exactly, but basically, the circuit has 2 states, On & Off.
On startup, it's off - C1 is uncharged, so Q1 is held off. As C1 charges through R2, Q1 starts to turn on. This turns Q2 on, which further turns Q1 on (through C1 - this happens very quickly and does improve the turn on speed), so now it's fully on and the inductor is passing current.
When Q2 or L1 start reaching their current limits, VCE of Q2 will rise sufficiently to start reducing the amount Q1 is on, which in turns off Q2 a little. The inductor keeps passing current and VCE of Q2 rises a lot, which holds off Q1 completely. When either C1 is fully charged in the reverse direction or the inductor current decreases sufficiently, Q1 is no longer held off and the cycle starts all over again.

 

[carbonzit]

Hi Carbonzit you are doing good so far.

Well, thanks. I do feel like I'm making progress (and having fun).

(Re above) The first thing to do is increase the inductor to about 470uH (500uH you had before). That will allow the lower operating frequency with the larger caps.

I completely disagree. Based on the results of simulation, 50µH seems to be the ideal value. Even 100µH is too high, and 500 way too much.

Are you just guessing? or have you actually run the simulation? or do you just have prior experience with these things?

By the way, notice larger caps (and bigger difference between them in an effort to boost duty cycle) and lower operating frequency (66kHz).

I also don't think R1 needs to be as low as 220, it only needs to be a similar value to R4, and a value of 1k will give you better energy efficiency later.

You'll see that I followed your advice here.

R2 is your friend, it controls the ON time period, so a pot here can quickly help you tune the duty and freq for testing.

Again, not my experience. Below a certain value, the oscillator wouldn't even start. Reducing it to 33K helped a teeny-tiny bit.

But more to the point, R2 also controls the multivibrator frequency, being one of the RC components. I've been changing the frequency by changing the capacitors (see below). I try to change only one thing at a time, and measure the effect that change has on circuit behavior.

D2 can be a cheap 1N4148, no need for special schottky diode there.

I did that, and lost a couple tenths of a volt output, so I decided to keep the Schottky. Hell, since I'm not planning on manufacturing these by the carload, who cares about a few cents more for a diode? I'm not taking the Lee Iaocca approach here ...

The -1.2v pulses at Q2 base should be ok.

Ah, thanks. They actually seem pretty benign after looking at them again. Plus some of the recent changes tamed them down quite a bit.

Also you might want to re-think the amount of output current you need, if this is a "9v battery replacer" then most devices running off 9v battery are designed to only need a few mA, so 20 or 30mA max should be plenty! At least consider the option of making one circuit that works well (high efficiency) from 0mA to 30mA or even less, and if you need a high current one make a second set of values for 0mA to 100mA. There are very few things that will draw 100mA from a 9v battery!

That is really the "big picture" question here. Look at the results shown on the latest incarnation of the circuit here:

**broken link removed**

I think it's getting close to being ready for prime time. Now I'd like to start thinking about regulation. Before asking you any more questions, I want to see what I can come up with on my own, using that circuit you posted up there as a guide. Of course, if you have any comments now on the matter, I'm all ears!

 

[carbonzit]

Regulated booster--done!

Dang, this was too easy. I can't believe it works ... first thing I tried. I figured what the hell, might as well try something, really not expecting it to work, expecting to have to come back to y'all with ten thousand more questions. Well, damnation, it worked right out of the gate!

Here it is, a regulated 3V--> 9V boost converter:

**broken link removed**

All I added was a transistor (Q3), a resistor, a zener and a regular ol' diode (that's because all I could find in LTspice's parts box was an 8.2V zener). The output is actually ~9.15 volts.

So somebody please tell me why this shouldn't work.

Or that it's the stupidest way to regulate a power supply. (Well, I know it ain't the most elegant solution ... kinda brute force, the way it throttles Q2's base like that, but hell--IT WORKS!)

I need to give credit to Mr RB, who posted practically the same idea in this reply.

 

[MrAl]

Well, I played with it some more, and this is the best I could get:

**broken link removed**

Results as shown (Vo: 10.8V; Io: 49.1mA @ Rl = 220Ω).

Not bad. But I have to say I'm very skeptical of this circuit. Change one component--and not by very much--and it fails to start at all. I actually found best results with a plain Jane 2N2222. If someone can come up with better results I'm prepared to be beaten.

Yes, the absence of a ground may have allowed the circuit to pick up DSCSI, as you described. Funny that LTspice didn't warn me about having no ground as it usually does.

You never answered my question about the oscillator: how exactly does it work?

Do you have LTspice (or another Spice) so you can try my simulation?

Hi Carbon,


I also think you are doing quite well with these circuits BTW.
Earlier when i quoted "100 ohms" i meant the output resistor, not R1, sorry about that.

This circuit is quite reliable. There are a few requirements but that goes for any circuit really. For example, the inductor ESR has to be low enough to allow enough current to flow to pull Q2 out of saturation. More about this later.

Try this:
Q2=2N4401
L1=200uH, Resr=0.1 ohms
C2=10uf
RL=33
R1=50
C1=300pf

and note the output current level.

Oh yeah before i forget, if you are generating output voltages that are greater than V1+5 volts (battery voltage plus 5 volts) then you need to add a 1N4148 diode across the base emitter of Q1, so i would suggest adding that for your 9v output.

One of the important things about this circuit is the gain of Q2. If the gain isnt high enough for a particular part number then we wont be able to get enough output current.
To help illustrate this, try connecting another transistor 2N4401 in parallel with Q2, except for the base, which is fed from the collector of Q1 through a second 50 ohm resistor. Change RL to 20 ohms, note the output current level !


This circuit is a little like a multivibrator but it uses not only the capacitor C1 as timing element but also the inductor itself.
When the circuit is first turned on, Q1 turns on via R2 so the collector goes high, turning Q2 on. Q2 turning on pulls the right side of the inductor low and the right side of C1 low. The left side of C1 charges quickly so C1 gets a charge across it with polarity +C- (left side positive right side negative). The inductor has approximately the battery voltage across it, so it begins to conduct current. The current keeps rising and rising, and that means the current through Q2 keeps rising too. Eventually the beta of Q2 and base drive through R1 can no longer keep Q2 in saturation, so the collector voltage starts to rise. As the collector voltage rises, that voltage across the capacitor (which acts like a small battery) adds to the collector voltage so Q1 base voltage rises. Q1 base voltage rises up and cuts off Q1, which means no current to the base of Q2 any longer, which means Q2 shuts off completely. The current through L1 that has built up no longer has any place to go, so the voltage at the right side of L1 shoots up suddenly in an attempt to find a path to discharge through. The voltage goes up high enough to turn diode D1 on which then charges capacitor C2.
As the inductor is dumping its energy through D1, the cap C1 now finds a discharge path through R2. With the right side of C1 quite high now the left side starts to charge toward ground making the charge across C1 now -C+ instead of +C-. Eventually the left side of C1 goes low enough to turn Q1 back on, and the whole process repeats.
Eventually after several of these cycles the charge in C2 builds up to some nominal voltage level so RL gets plenty of current.
The extra diode has to be added for voltages more than 5 volts greater than the battery voltage because as the Q2 collector goes high it pulls the right side of C1 high which pulls the left side even higher, which would reverse bias the Q1 base emitter by more than 5 volts which could destroy it.

One limiting factor is the gain of Q2 and of course the base drive. If Q2 can not stay in saturation long enough the current in L1 can not build up high enough to supply the output with enough energy to drive the load RL. This means lowering R1 increases the output energy and using a high gain Q2 also helps. Using two Q2's in parallel of course increases the output energy two fold.

Since Q2 does have to be pulled out of saturation however because that's part of the way this circuit oscillates, that means that the inductor ESR has to be low enough to allow enough current to build up through the collector. The max current is Vbatt/ESR so the gain has to be low enough to allow that current to pull Q2 out of sat.

Oscillation occurs due to two mechanisms: C1 left side charging low causes turn on, and L1 charging to a high level of current causes turn off. These cycles repeat and that's how it oscillates.


BTW many times you can use a max simulation time of 20ms unless you use a large output cap.