SMPS question

Hi guys.

I need a short-circuit proof, variable 100-500V .1A supply. I thought it would be a fun, educative and robust solution to build a 50 kHz flyback PWM converter running off the low-voltage (30 VDC) supply which this application needs anyway.

I've played a lot at this SMPS calculation page:

and gained quite a bit of useful information from it.

But I still have a few questions:

1) Is a 100V-500V output voltage range doable at all?

2) What are the minimum load requirements for such a supply, and how do I go about calculating them?

3) Short-circuit protection: The way I see it, if the load resistance is too low, the output current will decay so slowly that within the next couple of cycles the core will be driven into saturation, thereby tripping the primary-side overcurrent protection that prematurely cuts off the charging pulse. Can I rely 100% on this "natural" current- limiting feature of the transformer or are there issues that might come back and bite me? This supply is likely to be connected to all kinds of abusive loads, including short circuits, for indefinite amounts of time.

4) I thought about using the SG3524 PWM controller (taking note of its error amplifier's limited common-mode input voltage range of 1.8V ..

3.4V). Qualms? Thoughts?

5) Is there a more appropriate SMPS topology for this application?

Thanks for any suggestions,

robert

"Tim Wescott" a écrit dans le message de news: snipped-for-privacy@web-ster.com...

Yes, cycle by cycle current limit can be relied on. The part that can bite you is thatyour current limit circuit have a non null response time. If you have high primary voltage and low reset voltage, which you have when short circuiting the output, then current can peak really high.

Synchronous rectification is interesting, either because it:

• enhances efficiency: you have high diode losses wrt output power. This happens at high output current and low output voltage.

• allows the supply to work in more than one quadrant (switching amplifiers working on reactive loads for example).

Neither of these cases arise here, so going that route will bring much more trouble than will solve any real problem. I'd just use diodes here.

--
Thanks,
Fred.

The flyback converter looks good on paper because of its simplicity however the design of the transformer is very dificult, given the output range you want allmost impossible. The Sg3524 or 3525 chips have twin outputs making them well suited to a half bridge layout, the transformer is now easy to make. These chips have all the short protection you need just read the datasheet.

Keep in mind flyback transformers have to store energy for half a cycle, then deliver it to the load. H-bridge and half-bridge transformers don't store energy and as a result they can be much smaller, can use different core materials and have fewer turns. Speaking generally, they are more simple. Also, the issue of leakage inductance is not trivial. In a bridge circuit a little leakage inductance is often helpful to the circuit performance, whereas in a flyback it simply causes trouble. It's far easier to design good input-output isolation into a transformer if one is allowed a nice sensible leakage inductance budget. However, is input-output isolation even an issue that concerns you?

I won't address the issue of dealing with a 5:1 input, it's easily done in both cases; I don't agree with cbarn24050's remark about an impossible flyback transformer design, if that's what he meant. One thing, if you use the bridge approach, you'll need an output energy-storage inductor, but that's generally a modest issue.

Issues, issues. So many issues, so little time.

--
 Thanks,
    - Win

If you do the numbers you will see that you need a very different transformer when your supply is delivering 100v @1mA to when it's delivering 500v @100mA this means you end up with a core several times bigger than is strictly needed for a given power level. The airgap is a critical parameter so your design has to be compromised to suit a core with known gap or you have to start messing about with spacers. Leakage is hard to cancel on a flyback since the coils conduct at different times. Regulation for multiple outputs on forward converters is good if you use a common output choke, for the simple case of +/- output you can just center tap the secondary and fit a bridge rectifier if you use very fast diodes.

On 19 Sep 2005 18:03:33 -0700, snipped-for-privacy@aol.com wrote in Msg.

What is more difficult about making a transformer for a flyback converter than for a half-bridge one? Comparing the two designs I'm looking at a 1:40 turns ratio transformer plus a 1000+ turns choke vs. a

1:20 storage transformer. The only "difficulty" I find mentioned about flyback converters is that they require tight magnetic coupling between primary and secondary. Big deal. If I wind the secondary right on top of the primary, shouldn't that suffice?

I'm not disputing your expertise, I'm just asking for information.

Another thing that I find intriguing about the flyback topology is that it allows muliple output voltages while regulating only on one (somehow it automagically distributes the power correctly in the secondaries). Since a follow-up design would have a +/-500V dual output I wanted to keep that in mind.

Anyway, all this discussion is probably a good excuse to buy a book, even when I don't want to get too seriously involved in SMPS design.

Thanks, robert

Sure. But since I have to wind all magnetics myself my calculations show that transformer+choke is a lot more work than (storage) transformer alone.

Obviously, because it shoots back to the primary side. Causing great spikes. But in case you didn't read the thread from the beginning, this is a 100-500V 100mA unit running off 30V DC. So all I'm expecting is a couple hundred volts worth of spikes on the primary side that I can snub away if I need to. I also think that I can easily tweak stuff because I can safely run the low-voltage primary side from a bench supply.

...meaning, you can use a two-compartment coil former, right?

No. I'm just making 500V from 30V, with common ground. I was planning on winding the primary first and the secondary on top of it, with some kapton foil between layers. In the future I may make a +/- 500V design in which case the "self regulating" properties of multiple-output flyback converters will come in handy.

It's actually a 5:1 output. I need adjustable 100-500V @ 100mA, permanent short-circuit proof.

I've just begun dabbling with SMPS, and that's something I haven't understood about full-bridge topology: Why do they need a storage inductor at all? I'm sure it's good to have one just to reduce ripple, but it appears to be an essential part of the design and I don't know why.

So few books (that I have). Can you recommend a good SMPS book? Not that I'm planning to design enough SMPSes to warrant the purchase of a book; I'm just always looking for excuses to buy books, but they have to be good ones.

robert

I read in sci.electronics.design that Robert Latest wrote (in ) about 'SMPS question', on Tue, 20 Sep 2005:

Half primary - secondary - half primary is a lot better.

--
Regards, John Woodgate, OOO - Own Opinions Only.
If everything has been designed, a god designed evolution by natural selection.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

"John Woodgate" a écrit dans le message de news: snipped-for-privacy@jmwa.demon.co.uk...

^^^^^^^^^^^^

I knew I was better.

--
Thanks,
Fred.

If you're going to buy just one, I recommend this one: "Switching Power Supply Design & Optimization", by Sanjaya Maniktala It is a non-academic work full of practical information and real-world examples. I have Erickson, Billings, and several others, but put Maniktala at the left end of my shelf. Paul Mathews

I read in sci.electronics.design that Robert Latest wrote (in ) about 'SMPS question', on Tue, 20 Sep 2005:

Half primary - secondary - half primary is a lot better.

--
Regards, John Woodgate, OOO - Own Opinions Only.
If everything has been designed, a god designed evolution by natural
selection.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

"John Woodgate" a écrit dans le message de news: snipped-for-privacy@jmwa.demon.co.uk...

) about

^^^^^^^^^^^^

I knew I was better.

--
Thanks,
Fred.

Keep in mind flyback transformers have to store energy for half a cycle, then deliver it to the load. H-bridge and half-bridge transformers don't store energy and as a result they can be much smaller, can use different core materials and have fewer turns. Speaking generally, they are more simple. Also, the issue of leakage inductance is not trivial. In a bridge circuit a little leakage inductance is often helpful to the circuit performance, whereas in a flyback it simply causes trouble. It's far easier to design good input-output isolation into a transformer if one is allowed a nice sensible leakage inductance budget. However, is input-output isolation even an issue that concerns you?

I won't address the issue of dealing with a 5:1 input, it's easily done in both cases; I don't agree with cbarn24050's remark about an impossible flyback transformer design, if that's what he meant. One thing, if you use the bridge approach, you'll need an output energy-storage inductor, but that's generally a modest issue.

Issues, issues. So many issues, so little time.

--
 Thanks,
    - Win

vs. a

about

between

top of

that

(somehow

secondaries).

to

If you do the numbers you will see that you need a very different transformer when your supply is delivering 100v @1mA to when it's delivering 500v @100mA this means you end up with a core several times bigger than is strictly needed for a given power level. The airgap is a critical parameter so your design has to be compromised to suit a core with known gap or you have to start messing about with spacers. Leakage is hard to cancel on a flyback since the coils conduct at different times. Regulation for multiple outputs on forward converters is good if you use a common output choke, for the simple case of +/- output you can just center tap the secondary and fit a bridge rectifier if you use very fast diodes.

19 Sep 2005 18:03:33 -0700,

Msg.

converter looks good on paper because of its simplicity

the transformer is very dificult, given the

impossible.

have to store energy for half

and half-bridge

much

Sure. But since I have to wind all magnetics myself my calculations show that transformer+choke is a lot more work than (storage) transformer alone.

Speaking generally, they are more simple. Also, the issue of

inductance is not trivial. In a bridge circuit a little

often helpful to the circuit performance,

causes trouble.

Obviously, because it shoots back to the primary side. Causing great spikes. But in case you didn't read the thread from the beginning, this is a 100-500V 100mA unit running off 30V DC. So all I'm expecting is a

couple hundred volts worth of spikes on the primary side that I can snub away if I need to. I also think that I can easily tweak stuff because I can safely run the low-voltage primary side from a bench supply.

design good input-output isolation into a transformer if one

sensible leakage inductance budget.

....meaning, you can use a two-compartment coil former, right?

an issue that concerns you?

No. I'm just making 500V from 30V, with common ground. I was planning on winding the primary first and the secondary on top of it, with some kapton foil between layers. In the future I may make a +/- 500V design in which case the "self regulating" properties of multiple-output flyback converters will come in handy.

5:1 input, it's easily

It's actually a 5:1 output. I need adjustable 100-500V @ 100mA, permanent short-circuit proof.

if you use the bridge approach, you'll need an output

but that's generally a modest issue.

I've just begun dabbling with SMPS, and that's something I haven't understood about full-bridge topology: Why do they need a storage inductor at all? I'm sure it's good to have one just to reduce ripple,

but it appears to be an essential part of the design and I don't know why.

So few books (that I have). Can you recommend a good SMPS book? Not that I'm planning to design enough SMPSes to warrant the purchase of a book; I'm just always looking for excuses to buy books, but they have to be good ones.

robert

at such low voltages I would use a diagonal half-bridge primary - a switch at either end of the winding, one to +30V, one to 0V. And a pair of (schottky) diodes, one to 0V (at the +30V end), the other to +30V.

drive both switches simultaneously. Both on = energy storage. both off = energy transfer. the primary winding voltage is now clamped to the +30V bus, by the two diodes. Voila, no spike and almost no losses.

sandwich the smallest winding (fewest turns) between the two halves of the largest winding (most turns). This'll give you a significant reduction in leakage inductance, and allow you to use thicker copper (the number of effective layers halves)

all first-order switchers have an L and a C. in a flyback, L=Lmag.

otherwise, the primary switches see a capacitive load, with either dV/dt or R setting the peak current (ignoring leakage....)

Keith Billings Switch Mode Power Supply Handbook, followed by Abraham Pressmans Switching Power Supply Design. both McGraw-Hill

Cheers Terry

On Wed, 21 Sep 2005 08:56:58 +1200, Terry Given wrote in Msg.

Sounds good; I'll consider it.

I must confess that this doesn't make sense to me. The way I understand it is: Leakage inductance is created by flux that goes through one winding but not fully through the other (which makes it synonymous with poosr magnetic coupling). In other words, the portion of flux that the primary created but that doesn't go through the secondary will obviously not be reset by the secondary (and be put to good use on the load side) but will cause voltage spikes on the primary side that need to be dealt with. In the "diagonal" setup that you described, it will be fed back into the primary-side storage cap.

But before you go into explaining what "effective layers" means I really should read a book.

Ooops .. I should have thought of that myself.

A lot of money for a lot of knowledge.

Thanks, robert

all very true.

its easiest to think of this stuff in terms of H=NI/l (l = path length).

consider 2 simple windings, one layer each. "look" into the end of the transformer, down thru the center leg. You "see" the two winding layers, edge-on. One close to the core, the other far away (ish). H ramps from 0 to Np*Ip/l across the thickness of the primary winding (lets pretend H is linearly distributed). H is then constant across the inter-winding insulation, and (remember, Is flows the opposite direction, and Np*Ip = Ns*Is) ramps back down to zero thru the secondary winding. All well and good.

Hpeak = Np*Ip/l

If you split the primary winding, this happens:

H ramps up to 0.5*Np*Ip/l thru the first half primary

H ramps down Ns*Is/l = Np*Ip/l thru the secondary. So H ends up at

-0.5*Np*Ip/l

then H ramps back up thru 0.5*Np*Ip/l thru the 2nd half-primary, returning to zero.

Hpeak is now 0.5*Np*Ip/l

H is also zero at the centre of the 2ndary winding.

I cant recall if leakage is proportional to peak MMF or peak MMF squared, its one of them.

the "effective layers" are the number of layers between a plane of zero MMF and a plane of peak MMF. for a non-interleaved winding, its the number of layers. for a 1/2P-S-1/2P, its half the number of layers.

try Soft Ferrites, E.C. Snelling, its a *lot* more expensive.

Cheers Terry

In article , Robert Latest wrote: [....]

Yes and then some. The parts get a lot bigger as the range increases. Loop stability issues are also harder.

The power transistor grows in size as something like the square of the max output voltage. If we assume that the transformer has a 1:7 turns ratio, the transistor must be rated for over 30 + 500/7 = 101V. This is likely to force you to use a 200V part.

If you drop the idea of going with a constant frequency, there is no min load[1]. If not, you need to figure out how much energy gets put into the core with the shortest on time you can have.

[1] Including cycle skipping in with variable frequency.

Near enough:

Ipk(min) = Vin * Ton(Min) /L

E = 1/2 * I^2 * L

P = F * E

If you have a minimum on time, with a dead short on the output, the current will ratchet up to destruction since it doesn't decrease much during the off time. An extra comparitor can detect this (in at least 2 ways) and drop the running frequency enought to protect the parts.

--
--
kensmith@rahul.net   forging knowledge