Interesting question..... and as usual, my ideas are my ideas.
SOA is particularly useful for the linear region in my humble opinion, but for pwm, there is more to it, and less to do with the SOA graphs, and a whole heap more with wave shape, gate source voltage to achieve saturation, and how fast we can do that and discharge the gate, and most importantly if your going for speed..... the rate of change of the gate, versus the on time and off time.
So can it handle the pulse... and what is the max pulse current we can stand.... for the 4008 is is a few hudred for the pulses, and 600 or more for the surge.... so we have that times 4.... a big number.... happy.
Now the SOA spends a lot of time looking at drain source voltage to calculate the power through the device, considering RDS on, and current through the device.... but when the fets are on, the Vds is zero, and the Rds on is also insignificant really, so provided I stay uder the Vds and Vgs, I take little notice of any other spec..... coz I'm evil basically.
What is more important I have found, is that the wave form first, the switching frequency second. With Rds on of .007R, resistance losses are pretty low...... but if your rise time is slow, your in real trouble, because in the linear region losses are basically of "A" class amplifier proportions.... so we want to get turned on seriously fast, and off the same.... Thats why I use diodes across the gate resistors.... faster turn off ( providing the impedance of the driver is low of course).... and we want the turn on/off transitions to be a low proportion of the time, as on is cool, off is cool, so we want to be in either of those states.
So we need to turn on fast, and the wave form needs to rise with almost no X slope, only Y... so it appears instant turn on on the scope. In reality, there is still a turn on time, but compared to the on time ( we are switching slow, so use a slow scope trace), the wave looks straight up, flat across the top, and straight down again. As we increase the frequency, I have always found the rising slope is taking proportionally more of our operating time, so we are spending proportionally more time in the linear region, and our heating losses in the die go up astronomically.
So if we turn on the fet and leave it on with say 12v Vgs, and 40 amps, the thing is the same as a .007R piece of wire, once we start pulsing it, provided we can keep the thing off or on only, it will cool further, the moment we start to spend any time in the liner region, we get hot......
So to keep within operating specs, and knowing I don't know what I'm doing, I ere on the side of caution, and switch slowly if I get the chance, that way I am assured of ignoring most of my foibles in the design and layout..... and things go pretty well.
If I tried to design for say 20khz, my circuits and layouts would need to be much more professional to get the same thermal result, and as I'm a hack, thats not going to eventuate.
So I build what works for me. Your welcome to change the freq. The Arduino can switch at any frequency if you read the specs, and use the internal timers........20khz and over is easily achieved apparently, I just have not done it, but plenty have. Poida on the back shed has used a Nano for the entire 8010 spwm, and had no difficulty with speed at all.
So beware, going high has it's pitfalls, and the fets run cool mostly..... but only if you pay full attention to the wave form.... or put simply and succinctly... every transition from on to off and vice versa will add heat somehow... less the better..... and most of the time there will be no pulse until the battery is charged, and needs regulating, the rest of the time it is hard on.
The totem is actually the opto, it is a 2 amp driver as well as the opto. You will need this isolation, and an isolated driver for a normal HF grid tie. A little transformer is just fine, as are the 2 dollar plug packs things on Ebay. They take 400VDC fine, or the AC from the inverter etc etc. On the board shown, the 180R gets lifted and 18vDC+ goes to it, and the pin 5 takes the - of the isolated supply. The opto needs at least 16v to run. (and <30v) Aerosharp and other galvanically isolated inverters don;t need the isolated pwr supply.
........oztules
ps electronic engineers the world over will be tearing their hair out if they read this.