I've got one question regarding the gain calibration of xWATTHR, xVAHR, xVARHR and would really appreciate if you could help me out.
First thing I have to mention is that I will not use the pulse outputs - I will directly read the values of the energy registers in my application using SPI. So if I'm getting anything right, the only cause to calibrate the gains of the energy registers would be to achieve constant Wh/LSB, VAh/LSB and VARh/LSB constants for the three phases, right? Am I missing accuracy if I don't perform any gain adjustment (leave xWG, xVAG and xVARG as default) but use specific constants for each phase (i.e. Wh/LSB(A), Wh/LSB(B), Wh/LSB(C)...) in my calculation instead?
I'm looking forward to your advice...
OK! Things are looking better now as the phase calibration is working well. For power offset calibration, what is the current level you are testing at? You should be taking the minimum application current level for doing offset calibration.
Anyways, I would like to clarify again that the reactive energy error that you see at PF1 and active energy error that you see at PF0 are not valid points to consider the performance of the IC. At these points, the number of bits accumulated in those energy registers would be very small, and as your expected energy value is 0, even if you get a few bits accumulated, you have a large %error at that point. If you look at the actual energy in Whr or VARhr, that result would be very small. As active energy is mainly used for billing purposes, to avoid meter creep, there is a no-load threshold feature implemented in the IC for active energy alone. If the active power falls below 0.005% of the full scale power value, then no accumulation takes place in the active energy register. You can activate this by setting Bit 7 to 1 in the COMPMODE register. You cannot do this for reactive energy though. If you still have concerns about this or if you do not completely agree with my point on why these accuracy results are not significant, please let me know and we can discuss further.
As I see it, the one bad result that you notice is the reactive energy result at PF0.5. You say that it improves with accumulation time. What is current accumulation time and what is the least error that you see if you keep improving the accumulation time? To know if you are doing this correctly, you can accumulate for twice as much time at PF0.5 (compared to what you did at PF0 at the same current level). You should see a better accuracy result this way- similar to what you got for reactive energy at PF0.
On a general note, the seriousness of this issue depends on what is the reactive energy accuracy you are looking for? Is there a spec on reactive energy that you need to comply with? Apply the worst PF condition (at which reactive energy is lowest) possible in application with minimum current level that you need the meter to measure accurately. At this worst-case condition, keep increasing the accumulation time and see when (and if ) the error falls within your spec. Then the final question would be- Can you tolerate this accumulation time?- because you would have to leave this accumulation time on, for all your measurements. Another note here is that, for you to be able to do this experiment correctly, you should have done reactive energy offset calibration at the minimum application current level before doing this.
I would advise you to wait till you get these results before deciding to move on to other ADE ICs. But just in case you want to look at other options, I would suggest that you start with ADE7858A. It has a datasheet specification for reactive energy (same as active energy). The performance would be a lot better and you can make use of a lot of other features. One feature that is relevant here is that there is a noload threshold for reactive energy as well. On top of that, you can also set the threshold yourself (programmable). Refer page 64 of Rev.B datasheet for more details.