today I have an issue regarding the meter accuracy over frequency variation.
For realizing an energymeter which accuracy meets EN 50470 requirements, it is (amongst other things) necessary to keep the additional error lower than 0,3% (class C meter) for pf=1 and pf=0,5ind when varying the frequency by +/-2% (49Hz and 51Hz).
Unfortunately, our energy meter does not meet this requirement.
For PF=1, the accuracy is pretty good (additional error smaller than 0,03%).
For PF=0,5ind., the additional error is at about 0,7...0,8% (49Hz) and at about 0,3...0,4% (51Hz)!
I guess the reason is to find in the phase calibration. Since we use a dc-tolerant current transformer (to meet the dc tolerance test for EN 50470), a phase calibration of about 4,5° has to be done.
Since the phase calibration is done by applying a fixed time shift in the AFE, the compensated angle is varying when varying the line frequency, while the angle applied by the current transformers is supposed to be constant.
I read out the ANGLE1 register at various line frequencies and the delay between voltage and current actually does vary when varying the frequency.
But with this behaviour it seems to be impossible to meet EN50470 accuracy requirements for class C meter?
Is there a possibility to solve this issue which I have overlooked?
Thanks for your help!