AnsweredAssumed Answered

ADE9000 Calibration: issues with conversion constants

Question asked by avnrdf on May 3, 2018
Latest reply on May 13, 2018 by Gary.Kang.AnalogDevices

I'm using the ADE9000 EVAL board.


I'm not using the onboard resistor divider for the voltage channels, but a standard 230V:12V (actual output voltage is around 13.8V) voltage transformer, followed by a 2k:100k resistor divider which is placed in parallel with the dual 1k resistors (by soldering the wire directly to a pad) that the ADE9000 xVAP & xVAN channels are connected across.

The divider gives an output voltage of around 91mV (RMS measured between VxP & VxN) ) when the input of the transformer is connected to 230VAC .

The current channel consists of a 10A CT (1:1500) ratio, with a center tapped 44ohm resistor (dual 22ohm), which is connected to the IxP, AGND & IxN inputs.


I've managed to calibrate & derive the conversion constants for voltage & current using the ADE9000 evaluation software. I did this by connecting all 3 inputs to a single AC phase (to ensure that the voltage & current inputs to all inputs was the same), setting load current to exactly 10A (by setting a variable resistance to obtain 10A from a 230V source). I measured the voltage displayed by the reference meters (around 230VAC), current (which was set to 10A) and pressed the button in 'Quick Startup' to generate the constants. Note that this was done at unity PF.


The constants for voltage & current work & the readings I obtain match what my reference meters display (Fluke & Yokogawa PQ analyzers). 


I later noticed that the eval software recommends using 0.5PF to generate the constants, so I adjusted the RL load to obtain 10A at roughly 0.51PF, and pressed the button. 

The constants generated for voltage & current were similar to the ones generated using a PF of 1.0, but there is a significant difference in the power & energy constants.


I performed a load test across the input range (10A) in single phase, with the load wire passing through all 4 CTs (A,B,C & N). Using the xWATT constants generated using the original 10A current at 1.0PF, I calculated the value of power using AWATT * conversion constant = actual value, and the value was within 5% of the reference meter. However, when a RL load was used, the values of power began drifting & were incorrect.


I tried using the constants that were generated using 0.5PF 10A load, but those constants also result in incorrect values of converted power.


These are the constants generated:


233V Vin, 10A @ PF = 1.0~233Vin, 10A @ PF = 0.5

Register Name - Calibration constant - Unit

AIRMS 4.616269E-7 A/LSB
BIRMS 4.603334E-7 A/LSB
CIRMS 4.592681E-7 A/LSB
AVRMS 2.286063E-5 V/LSB
BVRMS 2.292061E-5 V/LSB
CVRMS 2.294761E-5 V/LSB
AWATT -1.372056E-3 W/LSB
BWATT -1.367890E-3 W/LSB
CWATT -1.365044E-3 W/LSB
AVA 1.416412E-3 VA/LSB
BVA 1.416147E-3 VA/LSB
CVA 1.414534E-3 VA/LSB
AVAR -2.462173E-3 VAR/LSB
BVAR -2.712263E-3 VAR/LSB
CVAR -2.788807E-3 VAR/LSB
AWATTHR -3.882052E-7 Wh/LSB
BWATTHR -3.870416E-7 Wh/LSB
CWATTHR -3.862159E-7 Wh/LSB
AVARHR -6.949110E-7 VARh/LSB
BVARHR -7.652811E-7 VARh/LSB
CVARHR -7.868438E-7 VARh/LSB
AVAHR 4.007209E-7 VAh/LSB
BVAHR 4.006669E-7 VAh/LSB
CVAHR 4.001895E-7 VAh/LSB

Register Name - Calibration constant - Unit

AIRMS 4.585139E-7 A/LSB
BIRMS 4.572149E-7 A/LSB
CIRMS 4.561535E-7 A/LSB
AVRMS 2.273238E-5 V/LSB
BVRMS 2.256897E-5 V/LSB
CVRMS 2.281691E-5 V/LSB
AWATT -3.799172E-3 W/LSB
BWATT -3.609116E-3 W/LSB
CWATT -3.596314E-3 W/LSB
AVA 1.398967E-3 VA/LSB
BVA 1.384975E-3 VA/LSB
CVA 1.396939E-3 VA/LSB
AVAR -4.566016E-4 VAR/LSB
BVAR -4.547452E-4 VAR/LSB
CVAR -4.594448E-4 VAR/LSB
AWATTHR -1.075519E-6 Wh/LSB
BWATTHR -1.021761E-6 Wh/LSB
CWATTHR -1.018180E-6 Wh/LSB
AVARHR -1.292316E-7 VARh/LSB
BVARHR -1.287065E-7 VARh/LSB
CVARHR -1.300346E-7 VARh/LSB
AVAHR 3.961866E-7 VAh/LSB
BVAHR 3.922240E-7 VAh/LSB
CVAHR 3.956129E-7 VAh/LSB

Note that although the conversion constants for I & V are the same in both cases, the values for xWATT, xVAR (and the corresponding energy values) differ significantly. The formula for conversion constants in the ADE9000 guide makes no distinction between the values of xWATT, xVAR & xVA; but that's obviously not the case here.


One more issue is that the conversion constant generated here for AWATT (even the correct one that's generated at PF=1.0) is negative for some reason. The Waveform buffer shows that there is no polarity error i.e. voltage & current are in phase, so the negative sign should not be present.


  • How do I calculate the value of VLEVEL & what does this influence? The example in the documentation states VLEVEL = x * 1144084, where x is the dynamic range. Since my voltage channel outputs around 91mV at nominal voltage, dynamic range = 0.707/0.091 = 7.76923. Therefore, VLEVEL = 7.769*1144084 = 8888652 = 0x87A14C. Is this calculation correct?
  • Why does the evaluation software different conversion constants for xWATT,xVA & xVAR? The technical reference manual's conversion constant section shows a common formula for similar values eg xWATT,xVA & xVAR (and a common value for xKWh,xVAh,xVARh)
  • What's the recommended method of calibration? I connected all voltage inputs to a single phase, and the wire that carried the load current was passed through all the CTs. This was done to ensure that I could enter the voltage & current in the eval software, since there is only one field; and I figured it's likely to work better if a common voltage & current is fed to all inputs.
  • At what PF should calibration be performed?
  • Does the energy constant generated by the eval software need to be multiplied by the AWATTHR_HI directly (instead of taking a reading from AWATTHR_LO?)