24V to 325V Flyback converter

[Electro Dodo]

Hello

I tried to design a 24V to 325V Flyback converter and PCB for it.
I used the UCC287502 controller.
For most of my calculations I just used the formulas from the datasheet.
It would be great if I could get feedback for it, both for the PCB and the design itself.

[Phil1977]

My two cents after a quick look onto the schematics:
- Do you want to run the converter without a coupling capacitor between primary and secondary? It may work, but the EMI will be bad.
- Do you want to use just one output cap? Output noise probably will be better with a set of different caps.
- Do you think you need no input filter? The snubber network seems okay, but anyhow flyback converters tend to have quite some EMI backload onto the power source.
- Is galvanic isolation security critical? Then do not forget to use a qualified transformer.

[Electro Dodo]

1.How would you find out what kind of capacitance you need, or if you need it at all. I know It should be a Y type.
2.Is there some rule of thumb or formulas how you should split up the total output capacitance you need in multiple capacitors?
3.The Flyback will be powered from my bench power supply. So I thin it should be fine.
4.The Transformer has a isolation voltage of 3000Vrms. Is there anything else I need to check for the Transformer?

[youngda9]

Drain and Source connections are reversed on the FET.

Add a cap on the CS pin to filter the sensed current.  Maybe 470pF to start.  Change R2 to maybe 1K to start.  Need to filter out the leading edge (leakage pulse) and have a smooth current sense feedback signal.

Add a 20-100 ohm resistor between Vout and R1FB, along with a test point on either side and a nearby GND test point, in order to make a Bode plot measurement to dial in stability.

Ensure snubber resistor size can handle the power dissipation required.

I didn't look into the PWM, but the optocoupler may need a pull-up resistor on the FB net.

[Electro Dodo]

1.You're right. That would have been bad.
2.R2 was calculated with the formula from the datasheet, so I think it's ok. Any good way to calculate what capacitor is needed on the CS pin? Or is it just try and measure.
3.Good idea
4.If I calculated that right it should be able to.
5.The pin has a build-in pull-up resistor.

[youngda9]

I didn't realize that the CS pin also had slope compensation, neat chip.  Per page 15 of the datasheet "A small capacitor with a value of 10pF to 50pF is recommended on CCS."

[Electro Dodo]

You'r right. I missed that.
Thank you for your help.

[youngda9]

I think the emitter and collector connections are reversed on the optocoupler as well.   

[Phil1977]

1.How would you find out what kind of capacitance you need, or if you need it at all. I know It should be a Y type.

It´s long ago I personally worked on this topic and from there I only remember to make it as large as possible to reduce EMI and as small as possible to reduce the longitudinal AC current. Since you are working with DC input you should have no problems there, so I would just try the largest Y1 cap that is available from the catalouge. With DC input this cap has no obvious disadvantages.

2.Is there some rule of thumb or formulas how you should split up the total output capacitance you need in multiple capacitors?

I think there was one EEVBlog video on Youtube about this question. Practically you should look at the impedance curves and you should try to spread the impedance minima over a large frequency range. The electrolytic cap usually has its resonant frequency below 1MHz and e.g. the diode of a flyback transformer easily creates EMI around 30MHz. So you definitely should add some good ceramics from 100pf to 100nf or similar.

3.The Flyback will be powered from my bench power supply. So I thin it should be fine.

Anyhow, if I already spend the money for a custom PCB I´d also spend the money for a small common mode choke in the input line. May some radio amateur say thank you for this.

4.The Transformer has a isolation voltage of 3000Vrms. Is there anything else I need to check for the Transformer?

If the transformer is intended to isolate a SELV-circuit from mains, then there are some more requirements for its construction. Like using triple insulated wire and/or additional insulation sheets. If insulation failure would create a hazard for the operator, I would apply the same standards.

[Electro Dodo]

You' right.

[Phil1977]

One more thing: Did I get it wrong or do you load the poor little TL431 with full 320V switching voltage?

[Electro Dodo]

I don't think so.
I used it in a resistor network, which should reduce the voltage to low enough levels.
I used the application notes from ti for the TL431.
I do agree thou, it feels sketchy putting the TL431 in the high voltage side.
But that's what the application note says.
So it should be fine.

[Phil1977]

I doubt it...

Just imagine the output voltage is 300V. Then the resistor network makes the "REF"-voltage of the TL431 smaller than 2.5V and it completely switches off. Then all voltage drops between A and K and its much more than the TL431 may survive.

[ArdWar]

Some quick thought mostly on layout. Use wider traces. At least use wider spokes for your pad thermal reliefs, some of them are too thin on high current path.
Maybe even direct connect the vias, I don't think they need thermal relief.

Tune your polygon settings/parameters so you won't get those tiny skinny sliver of traces going below random components. That's just asking for manufacturing problems.

Pick your feedback (R1FB "upper" trace) from nearer the output connector if possible, or at least at or after output bulk caps.


Also what's with the huge current sense, and absolutely humongous snubber resistor?

[Electro Dodo]

So I checked the application notes again, and it says nowhere, that the Vout of the Flyback matters.
But I think that is simply because ti just assumes that flybacks are only uses to produce small voltages of under 36V.
I just assumes that it doesn't matter.

Now thinking more about it, I realize I fucked the feedback loop up completely, and misunderstood the TL431 and its functions.

[Electro Dodo]

I did not think about changing the spokes width.
Now that I think about it, why do vias have thermal relive? I didn't really question it cause my EDA software just added them automatically.
Isn't thermal relive is only needed when I wanted to solder something to something?

The current sense resistor is so big, because the cheapest current sense resistor with the right value is this size.
The snubber resister is so big so it can dissipate enough power.

Thanks for the feedback.

[NiHaoMike]

The usual solution to powering the feedback circuit is a small auxiliary winding to get on the order of 10-20V. You can also add some zener diodes so that they drop most of the voltage with the 431 acting as "trim" for an accurate output voltage.
I suggest adding spots for EMI filtering if there's the slightest possibility you'll need it, you can always just not populate those if it turns out unnecessary. A resistor between the controller supply and input would be a good idea.
I also suggest removing thermal relief on all the power components to minimize stray resistance.

[Martinn]

Minimize the copper area of the switching node.
It is noisy as hell and will emit that in all directions. Use a thick enough trace, not a polygon fill.