Help with designing a 600V input Dcdc converter

[Pollky]

Hello,
so I need to make a dcdc converter capable of handling a 588V input and a output voltage of anywhere between 12V - 48V with a output power of around 200W. It needs to be as efficient as possible.

I have looked at a lot of designs, and i have tried to design one with the ucc28950. I downloaded the excel calculator and the problem arose with the transformer. It said i needed 20mH of mag. inductance, but there was no off the shelf one
especially with a 600V input rating so that's where i stopped with this design as I don't have that much knowledge to make one myself .

One topology that sparked my interest, was Vicors Sine Amplitude Converter found in a lot of their converters. But a suitable one would go for about 800eur so that was a no go. I then did some research and found a video of the working principle of their topology. I tried simulating it with no luck. Maybe someone here has some deeper knowledge on this topology that could help. 

Any help is appreciated and i thank to anyone in advance.

[David Hess]

I would have started with a custom wound transformer, however there might be another option.  Two (or more) standard transformers could be used with the primaries in series and the secondaries in parallel.

[uer166]

For 200W I'm not sure a PSFB would be worthwhile. Something like a 2-transistor forward would potentially do well here with a 1:4 output range/duty variation, and it has relatively benign control characteristics (I assume you are a power electronics beginner, so not sure recommending a PSFB/LLC/DAB/whatever is good here). You'll likely need a custom transformer no matter what you do here, but it's not that difficult to figure it out for this power level for a fixed frequency converter. "As efficient as possible" would also imply a sync rectifier on secondary, though you haven't quantified what this means so it may not be necessary.

[coromonadalix]

well i do understand  you don't want to pay much,   but DIY  and efficiency don't go well  unless you buy the right stuff, or develop for years

and some start at 128 euro

https://www.dwe-oss.eu/overzicht/600v-to-24v-dc-converter/




your's in time and tryouts or failures may or will cost more than that

[mtwieg]

Hard to give much info without more specs/requirements. You say "as efficient as possible" which leads to a fancier topology like PSFB or LLC, and would definitely mean doing a custom transformer. But would ~85% efficiency be sufficient? If so you could likely have a much simpler design with a forward converter and an off-the-shelf transformer.

You say this is a DC-DC converter, but does it require isolation? If so, what sort of requirements are there on the insulation (what safety standards will you need to apply)?

[Pollky]

First of all, thanks for all the replys.
Yes that was an idea i was considering aswell and i found some transformers that i could use but they had too small of an inductance.

[Pollky]

For 200W I'm not sure a PSFB would be worthwhile. Something like a 2-transistor forward would potentially do well here with a 1:4 output range/duty variation, and it has relatively benign control characteristics (I assume you are a power electronics beginner, so not sure recommending a PSFB/LLC/DAB/whatever is good here). You'll likely need a custom transformer no matter what you do here, but it's not that difficult to figure it out for this power level for a fixed frequency converter. "As efficient as possible" would also imply a sync rectifier on secondary, though you haven't quantified what this means so it may not be necessary.

Sorry for not adding that much details, but yes, i was planning on adding a synced rectifier on the secondary side, no matter the topology since i was aiming for 90% or more efficiency.
Also I was hesitant to use LLC or similar because of the complex control loops, and thats why I really liked Vicors SAC since it runs in open loop and in fixed frequency.

[Pollky]

Hard to give much info without more specs/requirements. You say "as efficient as possible" which leads to a fancier topology like PSFB or LLC, and would definitely mean doing a custom transformer. But would ~85% efficiency be sufficient? If so you could likely have a much simpler design with a forward converter and an off-the-shelf transformer.

You say this is a DC-DC converter, but does it require isolation? If so, what sort of requirements are there on the insulation (what safety standards will you need to apply)?

Yes, it does require isolation, but i forgot to write it in the post, but i could go without it as well  and just put an isolation module on the secondary side.

And yeah, from my research any flyback topology falls out as i require at least 90% efficiency.
As for the transformer, i was planning on testing a multi transformer design with the primaries in series and secondaries in parallel, as David suggested, or making a planar transformer as they are much easier to design (although with their own drawbacks) and prototype.

[David Hess]

Transformer losses can also be reduced by using a larger transformer than necessary.  Operating at lower than the maximum flux density (larger core) reduces hysteresis losses, and with larger windings reduces copper losses.

[Claude]

I would buy one meanWell WDR-240 (or similar product from similar supplier), either to look how they do it (reverse engineering can save a lot of time!), or simply to use it as the solution.

Difficult to say from your informations, but the price point of this power supply is not so easy to reach if your development and production is in small qty.


 

[jbb]

Do you need a continuously variable output from 12 to 48 V DC, or could it be switched (for example, mode A operates from 10 - 15 V, mode B operates from 40 - 58 V)?

If you need that output voltage range to be continuous, an LLC converter is probably out. They aren’t good at variable voltage ratio operation.

Do you have a known minimum load? Phase Shifted Full Bridges are said to loose soft switching at very light load (but I’ve never implemented one).

I wasn’t entirely clear from what you wrote earlier, so: did you find any off the shelf transformers for 600V? Even if one has lower Lm than the ucc28950 designer requests it may still be useable.

I was going to suggest an Active Clamp Forward converter, but I think they may not be suitable for such a high input voltage.

A 600 V input is a little unfortunate, as it will push you into 900 V or 1200 V power devices and make it a bit awkward to deploy DC input capacitors (I suggest film capacitors instead of electrolytics).

[jonpaul]

600 V is lethal vlotage. NO mistakes! Beware of safety and risk to equipment.

SMPS design and debug requires certain test equipment, and magnetic design exoertise.

Suggest to read and learn about SMPS and magnetic design before you undertake this project.

Many fine app notes and books.

Jon

[Pollky]

Do you need a continuously variable output from 12 to 48 V DC, or could it be switched (for example, mode A operates from 10 - 15 V, mode B operates from 40 - 58 V)?

If you need that output voltage range to be continuous, an LLC converter is probably out. They aren’t good at variable voltage ratio operation.

Do you have a known minimum load? Phase Shifted Full Bridges are said to loose soft switching at very light load (but I’ve never implemented one).

I wasn’t entirely clear from what you wrote earlier, so: did you find any off the shelf transformers for 600V? Even if one has lower Lm than the ucc28950 designer requests it may still be useable.

I was going to suggest an Active Clamp Forward converter, but I think they may not be suitable for such a high input voltage.

A 600 V input is a little unfortunate, as it will push you into 900 V or 1200 V power devices and make it a bit awkward to deploy DC input capacitors (I suggest film capacitors instead of electrolytics).

No, the voltage is fixed, it just doesnt matter what the final output voltage is. It can be 12V, 18V or anything between 12V and 48V as I will have a different step down converter on the output of this one.

The minimum load is 10W, so around 5% of the max output.

Technically I did find a 600V transformer, but it was a 2kW one, but thats a bit overkill. But i did find 200V, so maybe i could connect 3 of them so the primaries in series and secondaries in parallel.

[mtwieg]

Sorry for not adding that much details, but yes, i was planning on adding a synced rectifier on the secondary side, no matter the topology since i was aiming for 90% or more efficiency.
Also I was hesitant to use LLC or similar because of the complex control loops, and thats why I really liked Vicors SAC since it runs in open loop and in fixed frequency.

Hard to gauge your experience level with power electronics, but I would avoid attempting synchronous rectification in an isolated supply. The increased difficulty and complexity is often not worth the efficiency gains, sspecially if you don't need Vout to go down below 24V.

Is your efficiency spec because you need to minimize the size of the converter?

Technically I did find a 600V transformer, but it was a 2kW one, but thats a bit overkill. But i did find 200V, so maybe i could connect 3 of them so the primaries in series and secondaries in parallel.

Not sure where you're searching for transformers, but when a vendor refers to a transformer with a voltage specification it's usually because it's intended to be used in a specific application where that voltage is common. For example if I go to digikey and search for SMPS transformers, I see that I can sort them by the parameter "Voltage - primary". There's also a sortable "applications" parameter, one of which is "Power Over Ethernet (PoE)". If I look just at the "POE" transformers, all of them have a max "Voltage - primary" value of around 60V, which makes sense for a POE application. However, these transformers can certainly operate with much higher voltages just fine so long as other parameters are adjusted properly to avoid core saturation and excessive core heating.

That means if you have a novel application (like a high Vin) then the vast majority of off-the-shelf transformers are not going to be suitable at a glance, even if they would actually work fine. Unfortunately it will require a lot more work going through datasheets to actually deduce which ones are suitable, though.

edit: I should emphasize that that what I said above is only for voltages applied to the windings, not between them across the isolation barrier. Dielectric strength ratings from the manufacturer should absolutely be respected.

As for the idea of combining two transformers with series primaries and parallel secondaries... I don't think I've ever tried that specific trick, but in the past I've been bitten by assuming combining magnetic components would give similar behavior to a single magnetic component (i.e. one where all windings are on the same core). For example, if the magnetizing flux of the two cores became different somehow, what would the core-reset behavior look like for the two in combination? I'd recommend doing some simulations to explore this option.

[Wolfram]

As for the idea of combining two transformers with series primaries and parallel secondaries... I don't think I've ever tried that specific trick, but in the past I've been bitten by assuming combining magnetic components would give similar behavior to a single magnetic component (i.e. one where all windings are on the same core). For example, if the magnetizing flux of the two cores became different somehow, what would the core-reset behavior look like for the two in combination? I'd recommend doing some simulations to explore this option.

Series-parallel like this is nice in that it forces both voltage sharing, through the secondary paralleling, and current sharing, through the series connected primaries.

[T3sl4co1l]

What is your budget? -- design time, BOM cost @ units/yr, etc.?

You almost certainly don't want "as efficient as possible".  That will run at quite low frequency (10s kHz?), with very large and pricey magnetics.  It most certainly won't be off the shelf.  I'm sure >99% is possible, but you will pay dearly for it.

Assuming more modest cost and development, I might be inclined to go with QR flyback, though it's a bit of a stretch at this power level: the high input voltage makes it more favorable, output voltage less favorable; probably fine overall.  Two-switch forward isn't bad; I'd be curious how the efficiency compares.  I'd probably not bother with LLC or whatever, but there is more possibility of saving power through resonant operation that way.

Tim

[jbb]

...
No, the voltage is fixed, it just doesnt matter what the final output voltage is. It can be 12V, 18V or anything between 12V and 48V as I will have a different step down converter on the output of this one.
...

Well that's an interesting property.

Could you get the proposed converter to deliver exactly the required voltage and remove the "different step down converter on the output of this one"? That could help improve overall efficiency.

Alternatively, could you operate the proposed converter as an unregulated fixed ratio converter?  That is to say, it always does a step down of 588V -> 24V (or whatever) and if the input voltage varies, so does the output.  This type of arrangement is quite well suited to an LLC converter operating at a fixed frequency.

[Wolfram]

Plain flyback with 1700 V (or maybe even 1200 V) SiC is a viable option here, one that I would definitely consider.

[uer166]

Plain flyback with 1700 V (or maybe even 1200 V) SiC is a viable option here, one that I would definitely consider.

+1, 90% eff definitely possible

[David Hess]

Alternatively, could you operate the proposed converter as an unregulated fixed ratio converter?  That is to say, it always does a step down of 588V -> 24V (or whatever) and if the input voltage varies, so does the output.  This type of arrangement is quite well suited to an LLC converter operating at a fixed frequency.

In that case, an inverter could be used.

[NiHaoMike]

How about a half bridge design? Then the primary voltage as seen by the transformer would be halved. Also consider using an off the shelf switching power supply module designed to run from 400V AC, rectified that would be about 566V DC.

[trtr6842]

One characteristic of Vicors SAC's are that they are fixed ratio.  You cannot achieve a variable output with just a SAC alone.  All of the regulated output VICOR modules that use a SAC also include a non-isolated ZVS buck-boost to provide regulation, but with some extra losses.

No matter what topology you choose, odds are you're going to need to design custom magnetics.  DIY'ing your transformer and inductors can be cheap and give you very good efficiency.  Here you can find a magnetics design guide I wrote, an AC winding loss spreadsheet I wrote, as well as some other magnetic core analysis tools that might help.  DigiKey and Mouser usually carry ferrites, and you should be able to find something suitable there.  If you want powder cores and are in the US, CWS Bytemark has cores available with a reasonable minimum order quantity.  At thsi power level you probably won't need LITZ wire, and standard magnet wire is easy to source, and you won't need much for a 200W converter.  I am happy to help with detailed magnetics design.

One question for your specs, do you want a constant 200W output power over the entire 12V to 48V output range?  This would be called a wide range output power supply.  Or would just 12-48V, 0-4.2A be ok?  That is a much more standard spec.  There's a big difference between those two, and it will affect your design choices.  If you want the wide-range option, you'll probably be pushed towards a powder core output inductor for the soft saturation, and it would likely rule out a single stage LLC.  If you're ok with the 4.2A limit, then a ferrite output inductor will be more efficient and smaller, and you might have some more topology options.

Also, you mentioned 588VDC as your nominal input voltage, but what is the whole voltage range you need to operate at / survive?  That can change things a lot.  Also, do you have any EMI/EMC requirements?  Or is this jut a one-off?

I think if you're ok with it not being super power dense, a 2-switch forward is probably the simplest way to go.  Super simple control, plenty of gate drive options, plenty of current sense options, and easy output voltage regulation.  You can get that up to 90% full load efficiency with SiC primary side MOSFETs and decent magnetics.  You might not even need synchronous rectification, depending if your input voltage range will let you use 200V schottky diodes for the output. 

You could always do a good old-fashioned hard-switched half bridge converter, nothing wrong with that.  Skip all the funky resonant design and control challenges.  It'll be fixed frequency, simple control, and adaptable to your wide output voltage range.  It'll help keep the transformer small and efficient, and you could do either a current doubler output or a center tapped output, both are easy to convert to synchronous rectification.  With SiC MOSFETs on the primary, 90%+ efficiency should be attainable.

If you do want to get fancy, and especially if you want a really wide range output, an LLC would be the way to go, maybe with an additional buck output stage.  The buck output stage would greatly simplify the LLC design, since it could have a fixed output voltage.  In that case LLC design and control is pretty straightforward, and there are so many all-in-one control IC's that can handle that no problem.  Then the output buck converter is pretty straightforward, you could probably use off-the-shelf power inductors.

For 200W any full-bridge topology will probably be overkill, but a PSFB with an extra beefy transformer would probably be able to tackle the whole wide range output in one stage.