One will be switching fairly slowly, so only a hundred or so transitions per second.
This means the transition time versus time on is big... ie the transition time is small compared to the saturated states... so most of the time they are on or off, and not very often going in between states ( where the switching losses are highest, and so least efficient... more heat....) this is then the 50 or 60 hz switches.
The other side of the switching is the spwm signals, and they transition 20 thousand times per second or more.
As you can imagine, they spend a lot more time coming and going between on and off, than the slower 50hz switching fets.. These fets need to turn on and off 20 thousand times while the slower switches only do a hundred or so.
From memory, the high side switches are provided for by a galvanically isolated pwn supply, so high side losses are kept to a minimum, ie the high side should be as efficient as the low side with it's switching. They use pretty hefty isolated drivers for this that should have very good wave forms.
In the ozinverter, we use current pumps, and this lowers the efficiency slightly for the high sides, but it is not that significant, as the losses caused by the opto in the pj's probably is equivalent anyway. The high side spwm on that PJ has twice the coolig as the low side spwm, and that will help it seem better.
I don't know how they do it now, but either explanation will cover it I expect.
The inverter welder will power a 24v battery bank as it stands... probably only less than 6-10a on 48v pack.
6-7kw is nothing to be sneezed at, the PJ can deliver a pretty handsome house inverter from the looks of it. We are getting pretty blase now with these big transformer inverters now.... no one blinks at all if you get into the 6-10kw range for short periods now... who knew....
Interested in your findings on the newer style.
.......oztules