gain curve for LLC design

Posted by: mdeluca on 06/10/2014

I have a spreadsheet that compares well to a PIXLS simulation (results match) except for the gain inversion freq / Vin. PIXls gives an inversion at 172kHz @ 239Vin, while my spreadsheet gives 130kHz at 176Vin.

The PIXls sheet is attached.

I had assumed the inversion is the point at which the gain falls instead of increases as switch freq is decreasing.

---mdeluca

Files

HiperLCS Design1_1.pixls (43.49 KB)

Comments

Hi,

If you don't mind Could you please send your spreadsheet, so that I can figure out where the difference was coming from between both the files.

Regards,

Sudhakar

Thanks; I found the error. It was due to me forgetting that there are 2 output windings effectively in parallel making the load appear to double (Rac becomes half); this corrects the error on my spreadsheet and now matches PI XLS.

But I have come to another problem re Xfmr construction. See attached which shows pri winding to be constructed of 1212 strands of 44 AWG. Since Litz usually comes in 120 to 250 strands it means 5 to 10 litz wires are required! This would be very difficult to terminate onto pins.

But if you calculate the circ. mils per Amp of pri. current, 1212 strands comes to 1500cmils/A. Is this not excessive? My expectation is that 500 cmils/A is adequate which would require only 400 strands of 44AWG for pri current of 3A.

Hi,

Please find the attached document for list of litz wire.

Regarding the current design you seem to be using oversized core, the way PIxls spread sheet works is based on fill factor. It will typically fill 60%.

You can change number of strands manually such that the AC resistance/copper losses becomes minimum.

Regards,

PI-NANO