I'm Givin' 'er All She's Got Cap'n!
by Darren Ashby

The third and possibly last installment regarding switching supplies covers the fly-back converter with a possible twist. Plan on discovering expert mistakes, nerdy components, and a way to get your switcher isolated. So if "inductor" is your middle name and you think that engineers who talk in terms of voltage are sissies, then this one is for you. Be aware though, it is assumed that you have read the previous two articles on this subject. That means you can expect open-ended statements and much hand waving regarding previous points. So if you seek isolation, and don't mind being alone, let's dig in.

Confessions of an ... expert?

First, by way of correction, reader Jim Delmonico pointed out a beginner mistake I made when talking about the boost converter in the previous article. I incorrectly assumed that you couldn't get better than 50% efficiency out of this design thinking that current was wasted as you pulled it into the inductor. It seems obvious that it is simply getting shorted to ground right? Well, you are shorting the inductor to ground, but I failed to apply the fact that the current on an inductor does not rise all at once, its rise is determined by the LR circuit of the inductor and the output impedance of the source.ⁱ It takes a while for it to get up there so you are not using the maximum current all the time the switch is on. What this means is you need to consider this a power storage device, using current in the same way a capacitor stores energy using a voltage. You will waste energy if you wait for the current to get to its maximum value, but the point is to switch it off before that. In fact you can get 85% efficiency or more out of one of these supplies when it is properly designed!ⁱⁱ

It's a volt eat volt world

When I look around, it appears to me that controlled voltage levels drive much of the EE world. Everywhere you look, you see how important it is to keep voltage at 'such and such' level, no real mention of the current other than this much will 'blow it up' ⁱⁱⁱ. I think it is because of this that we as engineers often get to a point where we intuitively understand a circuit that has a cap and stores energy in it, releasing it later to smooth out a voltage supply and lower efficiency losses. Then we stare at an inductor circuit for a while, sigh, then hang up our hats and go home. It's like capacitors and voltages are our friends so we get to know them well. While inductors and currents are the weird kids that pick their nose in the hall so we really never learned what makes them tick. So let's dig deep inside ourselves to find the nerd in us that thinks those weird kids could use a friend and tackle the fly-back switching topology

Switching topics now

For me, this is where switchers started getting a bit complex. Due to this I have succumbed to more of a block diagram approach, as there are many different ways to implement these circuits. Here is one fly-back topology^iv generalized to represent most cases.

The first thing you'll probably notice is that the inductor is replaced by a transformer. A transformer is really just two inductors with something in common. Like "two peas in a pod," as my mother would say. Since these inductors share a common core, whatever you do to one side affects the other. Don't assume the core has to be anything other than air though. What is important is that the magnetic field of one coil will affect the field of the other. Different core materials can affect this greatly, but that is not what links the coils, it is the magnetic field that does that.

In this case, you are pulling current through the primary winding of the inductor with a switch, then turning off said switch. Can you imagine what happens when you do that? If you have been following along in this series, the first thing you should say is, "well, obviously the magnetic field collapses." It's the collapse of this field that induces a current in the secondary winding. Given that, it isn't too much of stretch to go the next step and apply feedback. But that is where this case grows more complex. With this type of converter the isolation is often a major reason to choose this architecture. That in mind, you will easily realize you can't just pipe the signal back to the control circuit without an isolation barrier of some type. Here there are several options, some use a third coil, some rely on the effect the secondary winding has on the primary, but the most common type (in my albeit limited search) I have seen is an optic coupling. If you would like to study more, here is a link that covers the benefits and constraints of several of these feedback methods.

Since the optic method is so popular (this is only my opinion - I can't say I really know this), I thought this opto/op-amp was a pretty cool part, eliminating PCB and some of the hassle in this type of design. If you are looking for more in-depth information here are some design tips that I thought were practical and helpful. Now if you really want to get in over your head, read this article, which explains the difference between continuos and discontinuous modes in switchers. It is a real page-turner! (Obnoxious smirk! ^v )

My personal quest, or now for the twist!

This entire trip into switchers, I blame on the fact I have been trying to replace a very, very cheap transformer with a switching supply. This has been the bane of my search so far, and I hate to say it, but I haven't been successful yet. When I add up the requirements (needs isolation, low component count, and consumes very little PCB space, etc.), I find the total cost is too high. But I recently had any idea that I have not seen out there, which might just work.^vi

I figure that optos are good at transferring digital signals, so why not run the gate signal of the switch through the opto, and have the control circuit run on the secondary side. This is all fine and dandy until you ask the question what happens when you first turn the power on. Without power the control circuit can't flip the switch. If it can't flip the switch it can't run, thus we have a 'catch 22.' This is where it gets sneaky, in the application I can count on the input voltage to rise fast. This will turn the FET on (since there is no control signal to turn it off yet) and allow a transient to flow in the inductor. The change in current (hopefully if I'm right) will induce a current in secondary, powering up the control circuit just in time to shut the switch off and start the control process. I think it is even possible to use a simple bang-bang control circuit, keeping the cost low. If you can't count on the rise time of the supply, it may be possible to add a diac or other low cost switch circuit to the control signal of the FET, which would create the transient needed to start the process. As I said, this is an idea in the making, just seeing the light of day, so I expect many problems will yet crop up and many issues remain to be solved. This is where you come in. I figure there are a lot more smart people out there reading CMP/Chipcenter than there are writing it. So please let me know what you think. I welcome any insight you may have. And let the fun begin.

----------------------------

Footnotes

ⁱ Funny how the mind works, I actually pointed out the LR characteristics of the inductor earlier in the piece. I just didn't apply it to the comment that I made at the end. In my experience, this is most common mistake engineers make. That is one reason I drill the basics some much. In this case a basic LR analysis would point out the fact that energy is stored and released.

ⁱⁱ I really appreciate it when readers point out my errors. I long ago realized I can't ever know everything about everything. So I have to settle for knowing a few things, learning a little new stuff along the way and doing the best I can with that.

ⁱⁱⁱ "Such and such" and "blow it up" are real specs that I have personally documented exist on many parts. But that is the fun part of the job.

^iv As I have mentioned before, the intent in this series is to give the reader a good basic understanding of switchers, as there are reams of information out there to cover many different points in quite a lot of detail. I don't plan on going that deep, so here is a link to a pretty good place to start getting more in-depth switcher information.

^v My personal opinion is the best way to cure insomnia is to dig out a data sheet or any other typical engineering fair! I didn't get into engineering for the great literature however. Those of you who know me realize it is because it was the closest I could come to entertaining regular explosions without as much danger. So have fun with your next switcher design, you may just get a chance to see chaos at work!

^vi I did find an article from a TI engineer, touting the benefits of secondary side control (using TI parts of course, sigh, I hate how biased the industry is some times *smile*). However, it was definitely much more than my shoe string budget would allow. Besides I had my idea before I ever found this link. That should count for something right? I figure if I invent something that was already invented, I may not make a million, but I can still call myself a smart cookie! A poor cookie too, but a smart one nonetheless! Try the link if you like. If it does not work you may need to register to get access. If any of you know this has already been done please let me know, I'd sure like a jump-start on this design.

Product Engineering Archive