Many-output LLC converter

[jbb]

Hello all

I'm considering embarking on a multi quadrant power supply design.  This will need an isolated power supply with six outputs: +-12V 100mA floating, +-12V 3A, +- 48V 100mA. 

It's theoretically possible to have multiple outputs from a single LLC converter as shown in the sketch. Does anyone have practical experience with this? I'm concerned by the large number of secondary windings...

[T3sl4co1l]

Possible, sure, go for it.

Practically?  Who knows.  What self- and cross-regulation is acceptable?  Is it okay if one low-current output is shorted and eats the full power output of the converter?

With such low levels I'd be inclined to tack on a separate DC-DC, probably flyback (off-the-shelf transformers might even exist!), and keep the LLC simple.

Tim

[Phil1977]

Why are you concerned about the nr of windings? As long as you have a large enough core there should be no reason to worry. Contrary to Flybacks you don't have to fight against stray inductance in LLCs, so magnetic coupling does not need to be that perfect.

Anyhow, even experienced SMPS developers don't always get these things right on the first try. If you need it in time and it needs to be reliable from the beginning I´d use off-the-shelf modules. If you want to spend time for advanced tinkering, then why not?

If I´d build it like you have shown in the diagram I would add some clamping on the output that protects against overvoltage. Something like a Zener together with a BJT.

[jbb]

Great questions, thanks.

I should have said first up, sorry: the desired use case is an inexpensive SMU.

I'm attracted to the LLC converter because if offers soft switching and tolerance for lots of pri-sec leakage inductance; I'd like to use a sectioned bobbin in the transformer to drive down pri:sec capacitance

... What self- and cross-regulation is acceptable?  Is it okay if one low-current output is shorted and eats the full power output of the converter?
...

I haven't properly quantified the self- and cross-regulation requirements. 
Self-regulation: I think that +-10% (including load transients) of the +12V 3A output would be sufficient
Cross-regulation of the -12V 3A output should be within 10% of the +12V
Cross-regulation of other supplies could be +- 20%.

The SMU output power amplifier will have a current limit; this limit can be used to avoid collapsing the whole LLC output.

...
If I´d build it like you have shown in the diagram I would add some clamping on the output that protects against overvoltage. Something like a Zener together with a BJT.

That's a good point - LLC converters are known to make over voltage problems at light load.

[David Hess]

Going by historic (1969) LLC designs, cross regulation can be pretty good.  CLC and common mode filtering of the secondaries will reduce noise.

[Faringdon]

With multi output LLC, you will need to custom make it so you get exactly the right leakage L associated with each output.
Then you have to rely on your winders to faithfully do it like that in production, for every unit.
What is vin tolerance?...if its wide then it doesnt suit LLC.

The cross reg will be very poor...what when the 3A goes on no load and the 48v one is on max load (or vice versa)...poor x reg.

Would go for LLC with 12V output...and spin off your other required isolated outputs from that via secondary side flybacks.

Then Presumably the 12v, -3A could just be a buckboost from the +12v rail.

...then its pretty much jobDone...as the others are pifflingly low power.
____
So to summarise......isolated offline converter to give +12v output....get the 12v +3a from that directly...then a buckboost or sepic from +12v to give -12v, 3A.

Then you just have those two piffling supplies to get done.
______
I am sorry to say it but multi coil outputs from SMPS  transformers is a myth...the Chinese  co's do it all the time, and just have loads of post prod test, or post regs...they wind a complex transformer, so the western customers think.....this is hard fiddley work...we must keep getting the Chinese to do it for us, our western assemblers could never do it..they would have nervous breakdowns trying to wind
 these transformers........when all along, it doesnt need to be done like that.....and would likely be easier and cheaper not done like that.

Go single coil output offline converter...then get the rest from that.

[David Hess]

With multi output LLC, you will need to custom make it so you get exactly the right leakage L associated with each output.
Then you have to rely on your winders to faithfully do it like that in production, for every unit.

The 1969 design example I showed has linear regulation on many of the outputs, so "pretty good" is relative.

Controlling leakage inductance explains why the schematic shows a dual primary winding.

[T3sl4co1l]

Keep in mind who you're replying to...

The above example lumps all the series inductance in an external part, so the transformer can be wound with low leakage, not super low, but probably just ordinary shell style design.

Probably they could've had reasonable regulation with some refinement, but being test equipment, just saying heck it and regulate everything resolves several concerns at once.

For my part, I've found excellent cross-regulation from the LinkSwitch-CV and relatives.  Flyback regulator that samples output voltage, primary (well, aux) side, during the flyback pulse, ignoring leading-edge overshoot.  After a modest preload, excellent cross-regulation is possible -- I've got a design that makes 5V 1A 5% and 12V 0.5A 10% output, avoiding the use of a post-reg, and the transformer has simple enough interleaving (split primary, bifilar secondaries).

Tim

[jbb]

Thanks all

I guess my next step is to do a bit of simulation, see how sensitive the cross-regulation is. If it doesn’t work in simulation it won’t work in reality :-|

I’m interested in low primary-secondary capacitance, so I might try a split-bobbin coil former. Fortunately LLC needs some leakage inductance.

It’s for a personal project, so rather than trying to integrate everything I’ll plan to have an external resonant inductor Lr.

[jbb]

Well, a crude set of simulations suggests that if I can achieve a coupling factor of 0.98 between the secondaries then things should be roughly OK.

[trtr6842]

Check out coilcraft's HTX7045 LLC transformers, they are purpose built for little isolated supplies.  Also check out TI's UCC25800, the datasheet has some good info on how to optimize open-loop LLC supplies.  The concepts discussed in that datasheet could help you design your own circuit, even if you don't use their driver.  I would say the main takeaway would be to move the resonant capacitors to the secondary side, and use one capacitor per secondary.  That way the entire system will have better cross regulation.

[julian1]

There may be a limitation with moving the resonant capacitor to the secondary side, in that the ordinary centre-tap bridge rectifier circuit can't be used to derive a bipolar output.

[trtr6842]

With a secondary side resonant cap you can use a voltage doubler rectifier to double the output voltage and then use a rail splitter to achieve a bipolar supply. The extra hardware might be worth the benefits of better load regulation.

[julian1]

It is an interesting point.
for a smu output, the bipolar analog eg. +-15V supply rails are referenced/centred around the floating force-hi output, rather than LO.
So one doesn't even need extra isolated bipolar supplies.
In a sense, the whole floating AB output section driving force-hi, acts like a kind of rail splitter.

[trtr6842]

To me, it makes sense to combine the ±12V 3A and ±48V 100mA on the same group of secondaries and run that in closed loop control.  The little floating ±12V supply might make more sense as its own converter, but I'm sure it could work combined.

After thinking about it a bit more, primary side resonance does make more sense, but the resonant inductor must be a discrete inductor on the primary side.  This is for two reasons:

• If you don't use a discrete inductor, then you are relying on the leakage inductance to store all your resonant energy, which will be an EMI mess in a sensitive instrument.
• You then use standard low-leakage winding practices for all the secondaries so their cross regulation will be as good as any other converter.  High isolation voltage or low coupling capacitance from the high current to floating ±12V secondary shouldn't matter much, so this approach makes sense.

Attached is a AC winding loss calculator.  LLC's can really bite you on AC losses, especially with the resonant inductor since there's no DC current component, which is not the case for power inductors in most other DC/DC applications.  This tool will also tell you if it's better to do your high current winding as a single layer of thinner wire vs a double layer of thicker wire.  I would recommend splitting the primary in half as two series windings and sandwiching all the secondaries between them.  That will help with cross-conduction and reduce AC winding loss from the interleaving.  The spreadsheet can't compensate for interleaving, but the improvements will be significant.

[jbb]

I could be dead wrong, but I think interleaving pri-sec-sec-pri should help with proximity losses in the secondary. Is that right?

(For comparison, the other way to wind it would be sec-sec-pri-pri.)

I ask because I’d like to design a transformer using standard magnet wire because it’s so much easier for a DIY builder to source. But standard magnet wire has comparatively thick strands which could cause trouble due to proximity effect losses.

Edit: especially if using more than one layer per winding

[T3sl4co1l]

Yeah.

FYI, Litz is readily available on eBay.

Tim

[trtr6842]

I'm pretty sure pri-sec-sec-pri will help with both primary and secondar ndary losses.  I'll have to check. It will definitely be better than no interleaving at all!  And it is a easy winding structure to keep the secondary leakages low.

[Phil1977]

Just don't forget: If it´s to be SELF-insulated, you need

- reinforced insulated winding wire: It may directly touch dangerous voltage, but it´s difficult to get reinforced insulated AND stranded wire. It´s available, many of the small wall-wart flybacks use it for their secondary, but I´ve rarely seen it at ebay or distributors. For one professional project I had to get it directly from the japanese supplier (X-Tech I think) who was absolutely not happy about quantities <1km.

or
- additional insulation sheets between the windings. Seems easier to do, but can get complicated with bifilar or very close coupling designs.

After winding we put our prototype transformers to a 15min-2500V insulation tester.

Of course, my knowledge about this is 15years old but risks of electricity didn't change very fundamentaly since then...

[jbb]

I was planning to start with 48 V DC; that way I don’t need reinforced insulation for safety.

Edit: if this were for volume production then I’d do an offline switcher design (380 V DC in), including reinforced insulation etc.

[trtr6842]

Adding insulating layers isn't too difficult but it definitely affects leakage.  The best method I've found is to use calipers to mark a precise width of some thin kapton tape, then cut it to the precise width with a ruler and knife.  Then it is pretty easy to cleanly wrap over a coil.

Starting with an already isolated 48v supply is a great idea.  Your resonant caps might get physically large since you'll need a lot of capacitance.  You also might end up needing a very small resonant inductance, and your transformer leakage might exceed that value.  You might need to prototype your transformer earlier than usually in the design process.