LLC design questions/help/advice needed

[Sniper1]

I am designing a 240-250W LLC.
Here are my design parameters: APFC input 390V nominal 350-410Vdc range

So i a m using this transformer : https://www.we-online.com/components/products/datasheet/760895641.pdf

and for the IC i am using the: https://www.ti.com/product/UCC25660

There is a tool that is a xcel that helps calculate component values.

Here is the problem i was planning to use the built in leakage inductance BUT then xcel gave me this:   
|Maximum Switching Frequency|fSW(max)|82.9|kHz|
|Minimum Switching Frequency|fSW(min)|60.3|kHz |

As for the IC:• 50kHz to 750kHz full-load switching frequency

All fine and reasonable values but in the transformer datasheet at typical application i have this: Switching Frequency f switch 70 - 120 kHz

Should i be concerned? i am not sure if those 10Khz are a problem , or should i play with a external inductor and see if i can get some numbers in the transformer range?

This is my first LLC design and i am not so sure what i can round off / proximate and what needs to be exact.
OR Should i play with Ln and external indutor and see if i can get the min frequency to be 70kHz ? problem is i didnt manage to increase it only lower it .

Also since i am already posting this here is another thing i am not sure off:

Cr is often split in 2 since it has benefits at start up and you simply spread power dissipation in 2 components but is there any risk in this technique? ( technically i can use caps with 1/2 the voltage rating and add some protection diodes but i would rather use something safe/ stupid proof)

Also i am having a hard time finding something that can handle 1.9A RMS and is not apple sized or very expensive.
And regarding the mosfets that i should use i have many options but the problem is that i dont have a estimate of loses , i would like to use SMD but i am not sure i can keep them cool , or some THT without heatsink but again in TI s reference designs they did it both ways are 2 different power LVLs ( then again one heatsink was simple flat piece of metal shared with all other stuff from APFC AND LLC should be low loses so i am a bit split here since as of now i can only get some conduction loses estimate RMS and my MX loses would be under 2W per FET.
Any advice?
ALSO i did atack the TI xcel calculator if anyone would like to take a look or play with it.

[TimNJ]

Reduce the resonant capacitance to ~33nF? I'm not sure I would be "concerned" per se, but if the controller truly has a hard low frequency limit, it could limit the amount of "boost/peak gain" you can get if the input voltage drops considerably. On the other hand, when you reduce resonant capacitance, you reduce the peak gain anyway so it might be a wash, in this particular case. Also, consider tolerance of the nominal operating point due to tolerances of L, L, and C.

As for the Wurth datasheet range, I think they are just giving you a ballpark idea. Flux density in an appropriately sized and optimized LLC transformer usually winds up being a fairly low percentage of the Bsat of the core. (e.g. 50% of Bsat or less). So, being slightly outside of their recommended range probably does not lead to saturation or anything. On the other hand, if they specify that range, why not try to get the frequency in that range?

As for the capacitor, you can use NP0 or a properly rated metalized polypropylene. I have analyzed a bunch of metalized polypropylene. Usually you need a double-sided metalized film, but not all are created equal. I would stick to TDK B32642B, KEMET PHE450, or KEMET R76. Check the V(RMS) derating vs. frequency graphs. RMS (AC coupled) voltage across one of these caps might only be like 50-100V in most cases, but the printed spec is the DC withstand voltage, which is not the same. The V(RMS) derating graphs are not really about voltage withstand, but really talking about dissipation/loss. You usually need a ~630VDC cap for typical LLC application. Or just use NP0, whose losses are considerably lower. I am actually not sure why NP0 is not used more in the industry. Film caps must be cheaper.

[Sniper1]

Reduce the resonant capacitance to ~33nF? I'm not sure I would be "concerned" per se, but if the controller truly has a hard low frequency limit, it could limit the amount of "boost/peak gain" you can get if the input voltage drops considerably. On the other hand, when you reduce resonant capacitance, you reduce the peak gain anyway so it might be a wash, in this particular case. Also, consider tolerance of the nominal operating point due to tolerances of L, L, and C.

As for the Wurth datasheet range, I think they are just giving you a ballpark idea. Flux density in an appropriately sized and optimized LLC transformer usually winds up being a fairly low percentage of the Bsat of the core. (e.g. 50% of Bsat or less). So, being slightly outside of their recommended range probably does not lead to saturation or anything. On the other hand, if they specify that range, why not try to get the frequency in that range?

As for the capacitor, you can use NP0 or a properly rated metalized polypropylene. I have analyzed a bunch of metalized polypropylene. Usually you need a double-sided metalized film, but not all are created equal. I would stick to TDK B32642B, KEMET PHE450, or KEMET R76. Check the V(RMS) derating vs. frequency graphs. RMS (AC coupled) voltage across one of these caps might only be like 50-100V in most cases, but the printed spec is the DC withstand voltage, which is not the same. The V(RMS) derating graphs are not really about voltage withstand, but really talking about dissipation/loss. You usually need a ~630VDC cap for typical LLC application. Or just use NP0, whose losses are considerably lower. I am actually not sure why NP0 is not used more in the industry. Film caps must be cheaper.

THX
I will chek those capacitor you mentioned.
Regarding the transformer i have a LTspice model and i ran some pure sine was through it and i didnt notice a dramatic change if i went lower or higher. 100kHz was like best power out of the transformer if it is coupled with 33nF.
I did try to play with TI calculators with i didnt manage to get in that range yet. but i man not done with variations. My input is from my AFPC so in practice it will probably be more stable then my input data + i oversized a bit
As in my APFC is 500W while this LLC is 240-250W so if it operates alone i think i will be very close to nominal APFC output , and if i have a heavy load i should be at a higher frequency anyway, right?
Ill see what i cab get through experimentation in calculators and also see what cap sizes are available

[TimNJ]

In general, LLC operating frequency is the same at light load and heavy load, if the following two conditions are met:

1. Input voltage is constant
2. The input-voltage:output-voltage ratio (conversion ratio) is exactly equal to the transformer turns ratio (Np:Ns).

In this case, you always operate at M=1 on the LLC gain vs. frequency plot.

In practice, input voltage may drop during AC line/PFC dropout, and the conversion ratio may not always match the turns ratio. In this case, the operating point deviates from the resonant frequency to some degree. Some controllers may also force high frequency operation at light load/burst mode to limit energy per cycle, but that's more of an adaptation to LLC limitations.

You can design an LLC with almost no "boost gain", i.e. no available head-room above M=1, but you will have very limited holdup time capability because you need M>1 when input voltage is less than nominal, to maintain full output power capability.

Study how changing L, L, and C affects the gain vs. frequency curves. It took me a while to wrap my head around it. I've tried a bunch of the calculators available from different vendors. I like this one from On Semi the best, I think: https://www.onsemi.com/download/design-development-tools/xlsx/llc-resonant-half-bridge-converter-design-tool-for-fan7688.xlsx

[Sniper1]

In general, LLC operating frequency is the same at light load and heavy load, if the following two conditions are met:

1. Input voltage is constant
2. The input-voltage:output-voltage ratio (conversion ratio) is exactly equal to the transformer turns ratio (Np:Ns).

In this case, you always operate at M=1 on the LLC gain vs. frequency plot.

In practice, input voltage may drop during AC line/PFC dropout, and the conversion ratio may not always match the turns ratio. In this case, the operating point deviates from the resonant frequency to some degree. Some controllers may also force high frequency operation at light load/burst mode to limit energy per cycle, but that's more of an adaptation to LLC limitations.

You can design an LLC with almost no "boost gain", i.e. no available head-room above M=1, but you will have very limited holdup time capability because you need M>1 when input voltage is less than nominal, to maintain full output power capability.

Study how changing L, L, and C affects the gain vs. frequency curves. It took me a while to wrap my head around it. I've tried a bunch of the calculators available from different vendors. I like this one from On Semi the best, I think: https://www.onsemi.com/download/design-development-tools/xlsx/llc-resonant-half-bridge-converter-design-tool-for-fan7688.xlsx

I did fiddle with numbers a bit but if i only use the leakage inductance i get nearly exactly the lower range of the recomanded frequencies. now only transistors and caps are a problem to be solved
I noticed that capacitor value vill change a lot my frequency and ill have to adjust at the very end witrh real standard values but not a great fear is removed.
regarding transistors it is power dissipation that i am concerned about since i have no idea about what are my FET loses i only have RMS current and conducted loses....
As for the caps i must check what u posted

[TimNJ]

What kind of cooling? Natural convection? Fan? Conduction to something?

MOSFET losses should be pretty low, when everything is working right. For 250W, 2 x TO-220 is probably appropriate. Heatsink something like this ought to be OK: https://www.digikey.com/en/products/detail/ohmite/CSM221-28AE/6138415

[Sniper1]

What kind of cooling? Natural convection? Fan? Conduction to something?

MOSFET losses should be pretty low, when everything is working right. For 250W, 2 x TO-220 is probably appropriate. Heatsink something like this ought to be OK: https://www.digikey.com/en/products/detail/ohmite/CSM221-28AE/6138415

Regarding loses: Ti has 2 reference designs and in both the FETs have a RDSon about 100m ohm they are still available one is dpack and the other is isolated to-220.
From a TI took i know my RMS current is like max 1.33A  I^2 is 1.77*RDSon that is 0.177 so about 0.2W.....  and i forgot a 0 when i said 2W   
Then assuming neglect able switching loses i can use SMD with no heatsink .....
TI s DPACK is 50 C/W so that is 5C hotter assuming natural convection   that is super cool
There is also a small period when there is reverse current through a built in diode and that will be 1.2V*Irms so about 1-1.5W worth of loses for a small power but then again is such a small period i am not sure if it really matters
Then again i make it so that i can also mount those isolated TO-220 , like through the PCB
OR i am missing something major.
I will also look at something a bit better since the reference designs are a bit old and i am sure i can get better ones.

[Sniper1]

Reduce the resonant capacitance to ~33nF? I'm not sure I would be "concerned" per se, but if the controller truly has a hard low frequency limit, it could limit the amount of "boost/peak gain" you can get if the input voltage drops considerably. On the other hand, when you reduce resonant capacitance, you reduce the peak gain anyway so it might be a wash, in this particular case. Also, consider tolerance of the nominal operating point due to tolerances of L, L, and C.

As for the Wurth datasheet range, I think they are just giving you a ballpark idea. Flux density in an appropriately sized and optimized LLC transformer usually winds up being a fairly low percentage of the Bsat of the core. (e.g. 50% of Bsat or less). So, being slightly outside of their recommended range probably does not lead to saturation or anything. On the other hand, if they specify that range, why not try to get the frequency in that range?

As for the capacitor, you can use NP0 or a properly rated metalized polypropylene. I have analyzed a bunch of metalized polypropylene. Usually you need a double-sided metalized film, but not all are created equal. I would stick to TDK B32642B, KEMET PHE450, or KEMET R76. Check the V(RMS) derating vs. frequency graphs. RMS (AC coupled) voltage across one of these caps might only be like 50-100V in most cases, but the printed spec is the DC withstand voltage, which is not the same. The V(RMS) derating graphs are not really about voltage withstand, but really talking about dissipation/loss. You usually need a ~630VDC cap for typical LLC application. Or just use NP0, whose losses are considerably lower. I am actually not sure why NP0 is not used more in the industry. Film caps must be cheaper.

Those cap series made my life easyer https://www.mouser.de/ProductDetail/KEMET/R76QI21504040J?qs=pfd5qewlna7ymMYyi0JNjQ%3D%3D is the winner i will use 2 in split Cr configuration
Kemet and TDK have nice datasheets while Panasonic made my life hard figuring out reratings and basically made ma calculate graphs .
As the output rectifier i made a synchronous designs and free room for diodes in case it does not work well.
Now i am almost done with the schematic. I only need to make the FB and decide on a way to measure my output voltage and current digitally ( probably with some HALL sensor or large shunt.