Extremely inaccurate LTSpice simulation, rapid failure at advertised output power likely #1
Description
First, here is the corrected LTSpice simulation that actually simulates something reasonably close to what is shown in the schematics.
There are several other problems too. Here is a list of the changes:
- Fixed
L1's ESR which was the wrong order of magnitude (L1 should have an ESR of 45mΩ, but it has an ESR of 450mΩ instead) - Fixed
L2's ESR which was at 22.5mΩ instead of 45mΩ. - Fictional superconducting ceramic capacitors have been removed in favor of ones that actually exist with realistic ESRs taken from the datasheet for the chosen part number (which, at 5kHz, is 100mΩ - very different from 0Ω).
- Their capacitance values has been corrected to what it would actually be under 45-48V of DC bias, which for the selected capacitor (and this is directly from the characterization data for that part number), results in a loss of roughly 95% of it's capacitance, making them effectively 500nF capacitors rather than 10µF ones.
- The two aluminum polymer capacitors ESR was corrected from 92mΩ (their ESR values at 100Hz) to their ESR values at 5kHz (6mΩ)
With these changes, at 5kHz, both L1 and L2 are operating above their maximum RMS current, albeit just barely. However, they will not survive operation at 10kHz/600W output power for any reasonable length of time.
It appears you may have based your design off the temperature rise vs. current graph in the datasheet for the 18µH part. This graph is clearly incorrect. Yes, datasheets can (and often do) have mistakes in them. This is one of them. The RMS current specified is the current that will cause a 40°C rise in the inductor. All of these inductors are the exact same physical size. The 40°C rise occurs at roughly 2W of dissipation across the board, except, inexplicably, the 18µH part, which, if that graph is to believed, can somehow dissipate 4 times more power. If that were true, they would have specified an RMS current of over 12A, yet instead they spec'd it at 6.5A.
Additionally, the output aluminum polymer capacitor even under a 5kHz load has a ripple current of 20A. At 10kHz, it hits right up against the maximum at that frequency - 25A rms.
However, the inductor will be dumping 8W into that pcb and everything soldered to it - including those capacitors. This will cause them to get much hotter than they normally would at that given ripple current, which will have a dramatic impact on their longevity.
I would be extremely concerned about very early failure of that capacitor and the two filter inductors at anything higher than ~350W/5kHz. Have you guys done any lengthy stress testing (hours+) at 10kHz? If it survives, it survives... but who knows for how long. I don't see any active cooling either. You can get away with a lot if you have a fan blowing air over the PCB surface (and thus the components attached to it), but a passive heatsink will be far less effective (after a few minutes anyway). Are you at least heat sinking the inductors to the casing using thermal pads?