Is there a guide for how to select the size of core and what turn count/gap size to use?
I don't have a good reference off the top of my head, but the good old winding area product method is quite applicable. There are two things to watch out for:
- just because you can go above 1 Tesla doesn't mean you should, check your core losses
- the cores are laminated, and the laminating material takes up some room, so the effective iron area is somewhat smaller than simply multiplying length and width
I also read that those cores are fragile, is that really true or it that only in comparison to the conventional iron cores that are very rugged?
I would say that they are a
lot tougher than ferrite, but should still be treated with caution. Probably less rugged than big old silicon steel.
All the amorphous iron cores I could find available in small quantities are horseshoe shaped (sold in pairs) and since the bigger ones are somewhat cheaper per pound, I had the idea to buy a single large pair, wind each phase on each horseshoe, then put them together with a piece of ceramic in between to set the gap and keep the two sides isolated, as well as keep the windings away from the gap area. Would that cause undesirable interaction between the phases if they're running at the same switching frequency but the output currents are not the same?
No go. You'll need one core set (2 halves) per inductor. I also recommend using a bobbin to hold the windings down. For C cores, you can improve the DC resistance a bit by using two bobbins, and putting half the turns on each bobbin (still just 1 phase!); this reduces the average turn length while still letting you fill the winding area, and decreases resistance.
Remember to consider AC resistance at the switching frequency (i.e. skin and
proximity effect). At 18 kHz, skin depth is about 0.5mm. Using 0.5mm diameter cores for the winding would require many parallel strands (certainly more than 20!) and be a nightmare to wind, so you might need Litz wire...
Gap can be set using a shim as per usual.
I don't think it would be very easy to do 3 level switching for a split phase design. The power module I'm planning to use (3rd gen Prius inverter, goes for fairly cheap) has 2 3 phase bridges. One I will use as a 3 phase variable frequency output for a thermal storage compressor (only needs light filtering to avoid EMI and standing wave problems) and the other I will use 2 phases for interfacing to the mains (via a 240V plug to a dedicated circuit)...
OK, 2 levels it is. Great that you're looking at some thermal storage - are you doing mechanical engineering on that side too?
Also, what modulation are you planning to use on the full-bridge output? The classic 'bipolar' mode only works as 2 level, whereas 'unipolar' produces 3 output levels and might help (at the expense of more common mode filtering).
Oh, be careful here. Are you using the Prius DC link caps, or applying your own? In any case,
make sure there are bleed resistors to discharge the capacitors. Otherwise they could give you a lethal surprise.
The Prius DC link caps will likely be quite small, because the expected loads (3 phase motor/generators) are constant power types. When outputting to a single phase grid, you will get 'lumpy' power output at twice the line frequency. So you might need to add quite a bit of external bus capacitance to buffer this energy flow.
(Spitballing capacitor calculation, I could be very wrong...
- 60A output
- Allow 15V DC link ripple; 15V / 60A = 0.4 Ohms
- 120Hz ripple; Xc = 1 / (2 * pi * f * C)
- Hence C = 1/ (2*pi*120*0.4) = 3300uF
Don't forget about capacitor ageing, so you may need nearly double that!)
If adding your own caps, you will need to be really careful about stray inductance.
... and 1 phase for a 120V UPS output. The neutral is connected to the center tap of the DC bus capacitor bank, so each phase is more or less switching at +-200V or so.
Again, think hard about your caps. You could structure your output to provide a 120V - midpoint - 120V output, just like a split-phase system.
Do you really need this output? It's quite cumbersome.
(There's a control system that senses the current drawn from the grid and commands the inverter to source a current to mostly offset it without netting an export, so it's not a grid tie inverter in the conventional sense.)
Remember that you will need anti-islanding sensing. If you choose to never export to the grid you must be very confident that your control system manages this reliably.