Designing a 48VDC to 30VDC 3A isolated power supply

[Decee1]

Hello.
I'm working on a project where i got 48VDC and i want to isolate it and step it down to 30V @ 3A
I've been looking for a IC that can do it with opto isolation but no luck.
I'v also made a PCB based on my theory of functionality which turned out to not work. Diagram attached so i can get shamed on.. There is some additional circuit behind the opto circuit I've designed but cut out of the picture. The TPS54361 circuit is a modified version of a recommendation from TI's WEBENCH.
So now I'm asking on the forum.. Anyone had experience with isolated DC-DC designs?

The ideal way to go would of-course to just do a 230v to 30V @ 3A but i don't want to play around with mains.. I don't have a deathwish 

[Decee1]

Here is a picture of my simulation. It is just a proof of concept and not optimized at all
Inductor is 1:3 which was the value i came up with though the recommended inductor for the TPS DC-DC converter and then i just trimmed the inductor until it gave me more than 30V on the output when i had a 10 ohm load at 30V  aka over 3A load.
The generator thingie im using is just a simple ideal opamp and a comparator block i made for TI. The selected duty cycle is 40%

So in a nutshell
 32V with 10 ohm load, 33uH PRI 99uH SEC, 100KHz Switching frequency and 40% duty cycle

[TimNJ]

You should be able to use most offline flyback controllers to work with a lower input rail voltage. Just avoid controllers with baked in UVLO/brownout protection. You want one which allows you to adjust the parameters. Most offline controllers also have an integrated HV start feature. Usually these are current source driven, which means they can start-up consistently over a wide range of voltages. The only thing to consider is if there's a minimum start voltage on the HV pin. I've seen some which won't try to source current unless the voltage on HV is ~70V or so.

From a brief search, seems that UCC28740-Q1 might be an okay pick: https://www.ti.com/lit/ds/symlink/ucc28740-q1.pdf?ts=1597463696174&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FUCC28740-Q1

[NiHaoMike]

The ideal way to go would of-course to just do a 230v to 30V @ 3A but i don't want to play around with mains.. I don't have a deathwish 

Have you considered a ready made module? Meanwell is fairly cheap but good quality, just beware of EMI.

[coromonadalix]

You wont play with the main if you use a good psu

you convert a dc voltage to an lower one with isolation .... you will have some efficency losses, a 10v margin should compensate ...

and the Meanwell brand is very good, we use tons of them at my shop

[Decee1]

Have you considered a ready made module? Meanwell is fairly cheap but good quality, just beware of EMI.

First up I've already got a contact person at XP Power which provided me with the 48VDC PSU. From other CO-workers I was told that Meanwell had some EMI issues and therefor i should avoid it.
Second the output voltage will be adjustable to a certain point that's why i have the feedback circuit removed from my schematic so a readymade module wont work

[Decee1]

You wont play with the main if you use a good psu

What i meant was that if i wanted to isolate the 48V i could build my own PSU but i don't want to do that because i have no experience with designing AC to DC converters

[Decee1]

From a brief search, seems that UCC28740-Q1 might be an okay pick: https://www.ti.com/lit/ds/symlink/ucc28740-q1.pdf?ts=1597463696174&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FUCC28740-Q1

That right there is something I'll have to look into! That seems like the way to go! Also a pretty simple design! 

[Pawelr98]

A quick feedback less design would be a push-pull using the old IR2153.
Uses a transformer so with fairly constant 48V you should be able to get fairly constant 30V on the output.
No short circuit protection or regulation but if strict regulation is not an issue then it will be the simplest design out there.

If you want to make a flyback design with feedback then most popular controllers will work.
UC3845 (8.5V startup) or UC3842 (16V startup) with a low RDSon transistor should be able to do the job just fine.

[Decee1]

A quick feedback less design would be a push-pull using the old IR2153.

If the load is changing between 0-3A the output voltage would change versus the load so a open loop is not a good idea.
I have been looking into the UCC28740 as Tim recommended. EVK board is a bit expensive so i might just do a mock-up with it to try it out.

[Pawelr98]

A push pull or half-bridge topology uses a transformer so the open loop voltage is not that much different from the voltage under load.
Winding resistance and diode drops are the main causes. If only rough voltage is needed then open loop is not that bad. Especially if considered how simple the whole thing becomes.

30V at 3A is 90W.
Generally speaking flyback is less desired at such power level and conditions.
Voltage kickback which has to be clamped means lower efficiency and other problems.

Classic 494 can be used to implement a stabilised push-pull.
But I guess there are more modern chips that can be used for this as well.

[T3sl4co1l]

I'd be plenty happy with one- or two-switch forward or flyback.  Push-pull is undesirable due to the lower current draw and doubling of voltage.  I think transformers are also less available (unless that doesn't matter because it's a one-off, I don't know).  Half bridge would be okay as well, but kind goes too far the other way (dividing the voltage in half and doubling current).  And full bridge is just too much complexity for so little power output, I wouldn't bother.

So, what's been mentioned above, or an old fashioned UC3842, or a handful of other things that don't immediately come to mind that I'd have to shop for, would be likely suspects.  For the two-switch inverter, use a dual bootstrap gate driver.

Definitely a buck regulator isn't suitable here; the SW node is never rated much more than 2V below ground, and only for short transients at that.  Likewise a boost regulator won't do, but a flyback/SEPIC/Cuk capable type will.  External switch, boost or buck, non-synchronous types may also be suitable.

FWIW, active-clamp forward seems to be the most popular among fractional-brick DC-DC modules (which you might look into if you don't need to make this circuit for special purposes).  Though they also seem to have terrifyingly bad efficiency at light loads, at least the couple I've seen.  Noisy too -- that's another thing the two-switch configuration helps with, reduces EMI.

Tim

[Decee1]

I'd be plenty happy with one- or two-switch forward or flyback.  Push-pull is undesirable due to the lower current draw and doubling of voltage.  I think transformers are also less available (unless that doesn't matter because it's a one-off, I don't know).  Half bridge would be okay as well, but kind goes too far the other way (dividing the voltage in half and doubling current).  And full bridge is just too much complexity for so little power output, I wouldn't bother.

So, what's been mentioned above, or an old fashioned UC3842, or a handful of other things that don't immediately come to mind that I'd have to shop for, would be likely suspects.  For the two-switch inverter, use a dual bootstrap gate driver.

Definitely a buck regulator isn't suitable here; the SW node is never rated much more than 2V below ground, and only for short transients at that.  Likewise a boost regulator won't do, but a flyback/SEPIC/Cuk capable type will.  External switch, boost or buck, non-synchronous types may also be suitable.

FWIW, active-clamp forward seems to be the most popular among fractional-brick DC-DC modules (which you might look into if you don't need to make this circuit for special purposes).  Though they also seem to have terrifyingly bad efficiency at light loads, at least the couple I've seen.  Noisy too -- that's another thing the two-switch configuration helps with, reduces EMI.

Tim

I've never heard about two-switch forward. Old technology?
Seems like this concept I'm working on is getting more complicated than i'd imagine it would. Typical.. 

Yeah i did realize the buck converter solution doesn't work and thinking about it it also makes really good sense. If the PWM reaches 90% it would have a high voltage output but driving it as a flyback would just not energize the secondary side enough

So do i understand correctly that active-clamp forward is simply just a flyback with another FET that is activated in the switching FETs off period with a bypass capacitor to clamp the inductor leakage inductance?

[T3sl4co1l]

Maybe.  But then, what's old is new?..

2-switch forward and flyback are popular in newer computer supplies, probably in part because the higher voltage set by the active PFC input stage would be annoying to do other ways, and because EMI is lower than 1-switch (both ends of the primary move at the same time, the common mode (average) is ~zero).

The unipolar drive means the magnetic core can't be driven over a wide operating range, but probably the switching frequency has risen enough that that isn't a meaningful restriction anymore.  (Old PSUs might've ran at 30-100kHz, half bridge forward; at such low frequencies, it's advantageous to drive the core with bipolar flux, up to doubling the available operating range before hitting saturation.  Up over 100kHz, to keep core losses down, the core flux density has to be low enough that you can't run anywhere near saturation anyway, and so you have no benefit from bipolar drive.)

Tim

[NiHaoMike]

2-switch forward and flyback are popular in newer computer supplies, probably in part because the higher voltage set by the active PFC input stage would be annoying to do other ways, and because EMI is lower than 1-switch (both ends of the primary move at the same time, the common mode (average) is ~zero).

The unipolar drive means the magnetic core can't be driven over a wide operating range, but probably the switching frequency has risen enough that that isn't a meaningful restriction anymore.  (Old PSUs might've ran at 30-100kHz, half bridge forward; at such low frequencies, it's advantageous to drive the core with bipolar flux, up to doubling the available operating range before hitting saturation.  Up over 100kHz, to keep core losses down, the core flux density has to be low enough that you can't run anywhere near saturation anyway, and so you have no benefit from bipolar drive.)

Nowadays, half bridge or full bridge is the norm for PC PSUs over 400W, with full bridge most common over 700W.

[T3sl4co1l]

Hm, I wonder if that was short lived then?  My 500W PSU is this type, but it's also over 5 years old I think.

Also a good candidate for that configuration (PFC stage), resonant.  Not sure if they're using them though.  Cross regulation on multiple outputs might be a problem (that's one feature of forward converters, the transformer and multiwinding choke both serve to equalize outputs).

Tim