Guru
Joined: 09/08/2007
Location: AustraliaPosts: 4406
| Posted: 07:46am 11 Jan 2022 |
Its highly likely that the gremlin is not in the software but the hardware.
Highly stressed parts just finally spit the dummy.
And when one mosfet expires, all its close mates end up a smoking ruin as well, without any clues remaining as to what let go first.
PWM is excellent at low to medium power, where you can get away with having very few mosfets per leg. Its much easier to get good balance and a symmetrical physical layout with only (say) two mosfets per leg, eight in total.
Where you have say five mosfets in parallel per leg, they may not current share particularly well, and the first one to turn on, and the last one to turn off cop most if not all of the switching stress.
Its not as simple as saying 4 mosfets in a bridge will produce 2Kw reliably, so twenty mosfets should give me 10Kw of indestructible power very easily.
The best developed software in the world cannot solve this problem.
PWM just does not scale up to very high power easily.
If you really need 8Kw reliably, high frequency PWM is going to be very difficult.
Not impossible, but a very serious challenge.
That is the Reason for the Warpverter.
The largest inverter switches at only 50Hz, and layout is far less critical. The difference between switching at 50Hz and 20Khz is 400:1
About the same as the difference as walking speed, and a fighter jet breaking the sound barrier. Everything becomes much less critical, and reliability at really high power much easier to achieve.
PWM is the obvious logical choice up to probably four to five kilowatts.
Beyond that it becomes much more diffiult.