Cuk converter PWM controllers?

[billgmdie]

Also, you (as a group) could get a generator and few batteries for collective use at the parties. This solves the shortage of space problem, and eliminates the need to run the generator in night.

 

[billgmdie]

did you try to parallel 2 batteries (1 large and another one smaller) and check the output's stability? wouldn't the small battery stabilize the larger one, or 2 big batteries stabilize each other?

 

[Schneibster]

1. Space- my truck is full with my complete kit, which comprises a large

Doesn't seem like a valid concern to me.

You don't have to pack the truck.

2. Noise and pollution- star parties are not just about the sky, they are also about experiencing the night, and the sound and smell of a generator is unlikely to make one popular. At multi-day parties, generators are often banned until noon so that people can sleep, and they are almost always banned at night. Getting your batteries charged can be a challenge, to say the least.

Hmm, I see... Never heard of star parties. Well, in this case, you could try to find an even smaller generator. Pollution is not a problem, IMHO, because those 63cm3 engines are quite efficient, and almost don't produce anything other than H2O and CO2 (and power, of course).
Noise pollution is actually a problem. This 950W generator produces > 60dB of noise at 7 meters, so you'd need to move it ~ 200m away for the noise to disappear.

First, although air pollution as a global problem is always something I'm concerned about, I was speaking of the smell- and I've never seen a generator that didn't make a smell. Talk to me again about this when you can directly breathe the output of the generator and it won't hurt you (save for perhaps lack of oxygen).

Second, we're talking about a clearing at the top of a hill out- waaaaay out- in the woods at night. I suspect that 200m would be nowhere near enough, and I base that estimate on prior experience. I've heard the "but it isn't that much noise" argument before, and I voted to ban them right along with everyone else.

You could make a hybrid system, too. During the day a few solar cells charge 2 batteries (the one you bring with you, and your truck's), and during the night the generator acts as a backup device - if voltage on the batteries drops too much, it'll kick in, and you won't loose the picture. Of course, you'll need to move the generator away, but I don't see why that would be a problem.

I can think of any numbers of reasons why it would be- from breaking an ankle wandering around in the woods at night after it fails to come on, to having something malfuction and start a fire. I don't think you have a very accurate idea of what this activity is like- I live in a fairly rural area, and I have to drive an hour to get to this place. It's out in the woods. There are four buildings within a quarter mile, and one of them is a pit toilet.

Generally, these are overnight trips- I leave a couple hours before sunset, and return the next morning after sunrise. So solar cells are of no use to me. On multi-day trips, I bring a set of solar cells that I use to charge with- but I'm drawing several amps, sometimes as much as 10, all night. 150mA for twelve hours is a help, but not much of one.

Also, you may get a gas-powered generator (if you can find one, of course). These are very quiet.

You mean propane? No, they still aren't quiet enough; we had a guy bring one and try it out, and the vote was to shut it down after ten minutes.

As far as the batteries "stabilizing" one another, I haven't tried it, but I'm cautious because of the internal resistance differences that are sure to be present. Overall, I think my plan is best, and the rest of the parts arrive today. I'll write more after I play for a few days.

 

[Glyph]

Very interesting project ya got going there. I've been trying to build power supplies myself (some work... and some don't) and find this project rather interesting for its unique requirements.

Correct me if i'm wrong but this what i believe you want:

1. low noise/ripple
2. small size/weight
3. high reliability
4. high efficiency
5. wide input range (3v-14v)
6. excellent load regulation

... good thing you didn't add CHEAP or you'd have defined the impossible.

I've been looking at your "low noise/ripple" requirement for your supply and the reasons why you choose a Cuk converter. I was wondering if you know the frequency at which your imaging system (CCD camera, tube, or whatever you're using) runs at. Because although its possible to bring the noise/ripple down to below the natural thermal noise of your imaging system you might run into a huge inductors and capacitors to do that, destroying the small size requirement of your project. There is an alternative however: if you can synchronize the frequency of your power supply to the horizontal scan frequency (or whatever system frequency) of your imaging system you can eliminate the noise problem by simply "tuning out" the noise.

I heard TVs had the problem of the power supply causing noise in the receiver which in turn appeared on the screen. By syncronizing the power supply to the horizontal scan frequency the noise was tuned to occur at the instant the scan beam was moving from the end of one line to the beginning of next, this period was essentially invisible to the viewer. When the beam finally hit the start of the next line the noise had passed and the beam could draw the next line uninterrupted.

The big part is having to find out what the frequency IS and how to synchronize with it. If there is an analogue output for your system you might be able to tap into the 'blanking' period that occurs between line changes and then synchronize to that. Alternatively, if you know exactly where this is; crack open the system and probe the frequency generator circuit (crystal or something) and then feed that (with isolation ofcourse) into your power supply.

Tuning the noise so its out of phase with your system may be very difficult but it essentially cheats the "low noise/ripple" requirement. You can have a tiny power supply that generates disgusting amounts of noise but because its tuned out you don't have to worry about it. Its like having ear plugs that only tune out bad music, you only listen to what you want.

Glyph

 

[Schneibster]

Very interesting project ya got going there. I've been trying to build power supplies myself (some work... and some don't) and find this project rather interesting for its unique requirements.

Correct me if i'm wrong but this what i believe you want:

1. low noise/ripple
2. small size/weight

Well, I ain't lookin to lug a 5kg box around- but keep in mind that I use car batteries. If it weighs less than that by a fair bit, I'm happy. Certainly, it will be smaller and more portable than a generator.

3. high reliability

That's more a matter of the design than anything that costs a lot. Just taking the time to put the right kind of controller and good safety options, and make sure I understand how they work. More on that in a bit.

4. high efficiency

This is probably the most important requirement, other than being quiet.

5. wide input range (3v-14v)

No, nowhere near. More like 8-14V, and I'm actually settling for 9.7V minimum. I intend to set the trip point at 11.2V, because below that you are damaging the battery; I'll have an override for it for those situations where it's important.

6. excellent load regulation

Well, now, I'm interested in being able to plug 15A into it and have it not sag below 12V. The initial output is looking like about 13.9V, and I'm at 12.51V at 4.9A. That's about as high as I can take the current right now; I'm working out an efficiency problem I'll get into later.

... good thing you didn't add CHEAP or you'd have defined the impossible.

What's cheap mean? How's $100US grab you? It's running (though as I say, I'm working bugs out of it), and I haven't spent that yet (at least not on parts for the supply- I had a couple other projects hanging around).

I've been looking at your "low noise/ripple" requirement for your supply and the reasons why you choose a Cuk converter. I was wondering if you know the frequency at which your imaging system (CCD camera, tube, or whatever you're using) runs at.

It's a CCD camera, and it accumulates charges as the cells are struck by photons. It might be open for 10 minutes or more, collecting photons- it doesn't run at a "frequency" unless you wanna talk milliHertz.

I was familiar with the TV supply story; I've seen it before.

My real main goal in all of this is: when the battery gets to 12V, not all of its power is gone- there is considerably more power available, perhaps as much as 10% or more, but I can't use it because all my equipment needs at least 12V. Motors start sticking, drive controllers start getting crazy, and it is generally bad. So if I can get at that power (and there's plenty of amp-hours left- perhaps enough to get me through the rest of the night, and almost certainly through this imaging session), and boost it to 12V somehow, then I'll be set. I'd like to be able to use the supply all the time, to smooth out glitches I've seen in the battery power, and generally give me good line regulation; the load regulation isn't very important except I want to stay above 12V. But the efficiency is a real problem; I'll lose far more time than I will gain back if I have to put up with efficiency that low.

Where things are at:

I have built the supply, and I went with the boostbuck rather than the Cuk. The main motivation was the feedback network; the design is easier with a positive rail rather than a negative one. Thinking back, that might have been a mistake; I might be able to see now how to get at it. But in any case, that's what I did. Luckily, I can undo it if I have to.

I ran into a problem: the tantalum cap I used for my energy transfer cap's leads started heating up! Surprised heck out of me, I jumped back after hitting the panic button expecting the usual red flames of a burning cap. After it had cooled off, I pulled it, and all of the problem was at the bottom- not up top, like it is if the cap itself starts to fail. I thought about it, and realized that it had been exposed to too much current- not exactly what you expect with a cap. I thought it over, and wound up with two series pairs in parallel- it's the same value, but twice the current handling capability. I'll address it more if it becomes a problem, but for now it looks like this solution was successful.

During the breadboarding, I realized that when the controller shut down due to current limit, it turned the switch in the boost converter off- but because I am using a ground-referenced drive and a P-channel MOSFET, it left the switch in the buck converter turned on, and this provided a direct low-resistance path to the load. Obviously, this is unacceptable from a safety standpoint, and in the event of a real short, the buck switch would be destroyed. I therefore added a second driver, so that I could detect current limit (luckily the current limit builds a voltage across an external cap, so I can detect that with a comparator and use the signal to control things) and now the system shuts down both switches, so the load sees zero current and zero voltage.

While I was at it, I added a reset switch (if you drain the limit voltage from the cap, it restarts), and a disable switch (just an AND gate that either permits or does not permit the signal from the controller output to reach the switches). I also added a comparator to watch the input voltage, and a housekeeping supply for all of the logic and the controller (it's a 723, so it browns out about 3V above the needed voltage- thus, the comparator shuts the system down when the input voltage gets to 3V above the brownout voltage- which is 9.7V). I initially had bad load regulation, but when I added the 723, it improved miraculously. I think the TL5001 likes a nice steady working environment.

The ripple is extremely low. I am running at 216kHz with a (now four) 22uF tantalum as the energy transfer cap, and I started with a 47uF tantalum as the filter- but went with a 100uF redcap instead. It improved the ripple spec and cut some of the switching noise out. The supply puts out 4.9A at 12.51V, so far with no problem except this:

Efficiency. I expected better than 85% in my initial design, and I'm only getting about 60%. That means that I'm drawing 100W, but only getting 60- the other 40 get dissipated in the supply, and that's kind of limiting me just at the moment. I'm looking the transistors over, because they seem to be the main culprits, according to my fingers (Ow! Yep, that one's hot). Not only is it creating problems that are limiting the load, but it's a waste of battery power.

The only other complaint I have is a certain amount of transient noise from the switches. I'm looking into some beeper caps for it now. We'll see what comes out the other end.

 

[evandude]

I haven't tested the maximum current yet, but I did finally build the SEPIC converter at this site:
**broken link removed**

and right now, even with it messily soldered together on perfboard i'm getting at least 75% efficiency.

you can check out my actual setup at: **broken link removed** (click on the "power" tab)

So far I haven't had a good load to try it out at more than 22W output but it seems fine at that level... and the designer claims 8 amps output...

Not to mention, it only cost me about $20 for all the parts!

 

[Schneibster]

How did you measure the efficiency?

 

[Nigel Goodwin]

How did you measure the efficiency?

I imagine he simply measured the power it's drawing from it's input, and the power it's supplying to the load - then apply simple maths!.

 

[evandude]

exactly. I knew my load resistance, so I measured the voltage across the load, and found output power... and used a multimeter to measure input current... and measured input voltage AFTER the multimeter (since it drops a volt or two in current measure mode)

then it's just algebra.

 

[Schneibster]

OK, as it turns out the efficiency of the boostbuck converter is limited by the presence of the P-channel MOSFET in the power rail. The problem is that the design formula I am using yields a higher voltage in the boost section than in the buck section; this voltage must perforce be dropped across that MOSFET, and it therefore dissipates power in relation to the current drawn. The tradeoffs for lowering this voltage are more noise and less load regulation, which are also unacceptable. So the boostbuck design does not meet my criteria.

At low current levels, very little power is lost, so the efficiency at 1.2A is about 85%; I suspect it could be driven even higher with careful design, and lower current requirements. But at 5A (actually 4.8A), about 100W is drawn from the source, but only about 60W makes it to the load, for efficiency of only 60%. Almost all of the 40W dissipated is lost in the P-channel MOSFET. It's not a problem for the MOSFET (if it is properly heatsinked), but it is a major problem from an efficiency standpoint.

This is unacceptable in my application.

It also makes the use of this design for an offline converter questionable, since it is not more efficient at high amperage than a linear supply. It does, however, remain quieter, so that might be an advantage.

However, this problem should not exist in the true Cuk converter. The Cuk converter has both coils and the energy transfer cap in the high rail, and the diode and MOSFET go to the low rail. So the current should not be directly dissipated by the transistor (or the diode, for that matter). What is not clear is what the average voltage across the energy transfer caps will be, nor whether that voltage will vary with the current drawn. If the voltage is not high, then the caps will not have to dissipate much heat; but if it is too high, I may wind up with that SEPIC design after all.

I have a pretty good idea of the regulator and driver components, and of the correct values for the Cuk converter design; I'll post the schematic for comment now. I'm breadboarding it over the next few days to see what it does.

Two caveats: I have seen trouble that might be due to an oscillation path introduced by the undervoltage lockout comparator's input coming from the input voltage (and therefore perhaps being amplified by the MOSFET, forming a feedback loop that destroys the MOSFET and potentially the driver chip). I am not sure, and I have not yet hooked it back up to see, since I really don't want to destabilize my breadboarding. You will note the decoupling caps on both inputs to the comparator; they're "just in case," and I have no proof that they will work. So watch out if you breadboard this. Second, I have not completed breadboarding and test, so this is not a confirmed design yet, nor is it fully back-annotated. If you build it, you do so at your own risk.

One last note for any newbies out there who might be thinking about messing with switchers: watch out. Switching power supply breadboards have a nasty habit of blowing up various creative ways, and I have had one tantalum cap burn and one aluminum electrolytic canister blow (and I do mean blow, it was pretty impressive) just during the breadboard phase on this project. These are not toys; they are serious stuff, and there are high voltages and large amperages floating around in them. If you aren't fairly well trained, you are taking your health or your life in your hands messing with them. Always, always, safety first. Personally, I wear shooting goggles and a heavy sweatshirt when I mess with them, and I protect my hands when I first power up a new breadboard. I also keep an extinguisher handy. Should I screw up and have something blow or burn, I don't wanna get hurt, and you should take precautions to make sure you don't either.

 

[Schneibster]

Here is the schematic.