1. When you say 6266W is the full scale power level, do you mean that you apply +500mV to both current and voltage channel inputs of the IC at that point?
Yes. At a Vrms of 295V V will be swinging +/- 500mV and Irms of 21.23A I will be doing the same. That's electrically, external to the chip. Internally I also use VGAIN , IGAIN and WGAIN to give me real-life values for the readings. Those values get determined at calibration time. So perhaps that needs to be taken into consideration too, when I'm calculating my PMAX, since those gains are applied early in the logic?
2. So, try applying reverse polarity signals such that the amplitude stays about the same even after the switch (eg. 160W on one side and 320W on the other) and try to see what difference it makes.
I found in general the smaller the reversed load, the longer it took for the interrupt to fire, suggesting it was to do with energy accumulating somewhere and needing to be backed out.
3. Also, how long do you apply the 160W load before turning on the reverse load?
Indefinitely.. i.e. a long time. But even if we assume the accumulator was just about to hit WATHR in the forward direction when I caused the reversal, with the reverse load being roughly 1/3rd of full-scale deflections, once things have settled, the accumulator should then feed into the AWATTHR register at roughly 8000/3 Hz right? That's roughly every 375usecs, or have I misunderstood that part? I use the recommended setting for WATHR (2 in WTHR1 and 0 in WATHR0).
4. Try repeating this experiment multiple times to see how repeatable the measurements are, particularly to observe if the point of reversal ( close to ZX, close to peak of the signal, etc..) affects the time taken to get the interrupt
It seems pretty repeatable. The position in the cycle at which the reversal happens doesn't seem to matter. The only thing I've noticed is the lower the reverse load, the longer the interrupt takes (as mentioned in #2 above).
If you let me know what is your application requirement for a wait time period to get an interrupt and how you plan on using this feature,
To be honest it's more out of curiosity then a hard need. Currently the only use I have for the direction is to determine LED colours, so latency is not a problem (so much so that I actually just poll the CHSIGN register to get it). I enabled the interrupt to see it in action (just using channel A), and was a bit surprised by the latency, but it's not holding us back. I guess I'm a little concerned that things mightn't be set up the way I think I've set it up. It may be all my gain register settings that is making PMAX a lot different from what I think it is by just looking at voltage dividers and CT burden values.