Modbus Power Measurement Resolution
Overview
The WattNode® Modbus meter measures several power values once per second in raw units. It then scales the raw measurements based on its calibration, the nominal line voltage, and the current transformer (CT) full scale rated amps. Next, the meter averages the power readings depending on the configured Averaging register setting. The default averaging is five seconds, but it can be set to 1 second, 5 seconds, 20 seconds, or 60 seconds. The computed results are stored internally as 32 bit floating point values.
When we refer to "power values" we include the following:
| PowerSum | PowerA | PowerB | PowerC |
| PowerReacSum | PowerReacA | PowerReacB | PowerReacC |
| PowerAppSum | PowerAppA | PowerAppB | PowerAppC |
As with any instrument, the power measurements have noise and systematic errors. With the averaging set to one second, the noise will be more apparent. Longer averaging periods will reduce the noise. Systematic errors are not affected by averaging, such as errors due to imperfect calibration, drift, temperature variations, and current transformer phase angle errors.
Raw Measurement Resolution
At full-scale power (nominal line Vac, rated CT amps, and unity power factor), the raw measurement will report approximately 260,000 counts per second, resulting in the following equation for raw resolution:
= raw measurement resolution in watts
= nominal line AC voltage (120, 230, 277, etc.)
= rated CT amps
The following table shows the nominal Vac values for different WattNode Modbus models and the corresponding raw measurement resolution using the smallest (5 amp) and largest (6000 amp) current transformers we sell.
| Model | Vn Nominal VAC | 5A CT—RR Raw Resolution in Watts | 6000A CT—RR Raw Resolution in Watts |
|---|---|---|---|
| WNC-3Y-208-MB | 120 | 0.00231 | 2.769 |
| WNC-3Y-400-MB | 230 | 0.00442 | 5.308 |
| WNC-3Y-480-MB | 277 | 0.00533 | 6.392 |
| WNC-3Y-600-MB | 347 | 0.00667 | 8.008 |
| WNC-3D-240-MB | 120 | 0.00231 | 2.769 |
| WNC-3D-400-MB | 230 | 0.00442 | 5.308 |
| WNC-3D-480-MB | 277 | 0.00533 | 6.392 |
Note: The effective resolution may increase due to averaging. We estimate approximate resolution increases of 2X for 5 second averaging, 4X for 20 second averaging, and 6X for 60 second averaging.
Output Resolution
The power registers are available in two formats: integer units of PowerIntScale and floating point numbers in units of watts.
Integer Power Registers
The integer power registers normally auto-scale to one of the values in the following table so that 120% of the nominal full-scale power won't overflow. For example, if you are using 30 amp CTs with a WNC-3Y-208-MB, the full scale power is 120 V * 30 A * 3 phases = 10800 watts, so PowerIntScale would be 1, and the resolution would be one watt.
| PowerIntScale | Power Resolution | Maximum Power Reading |
|---|---|---|
| 0 | (default) | Auto-configure |
| 1 | 1 watt | ±32767 W |
| 10 | 10 watt | ±327.67 kW |
| 100 | 100 watts | ±3276.7 kW |
| 1000 | 1000 watts | ±32767 kW |
If the power reading is large, then the resolution as a percentage of the reading will be better. As the reading gets smaller, the resolution will get worse. For example, the following table shows the resolution as a percentage of reading for different power levels. In all cases, PowerIntScale = 1 watt.
| Power | Resolution (% of reading) | Resolution (watts) |
|---|---|---|
| 20000 W | 0.005% | 1 W |
| 10000 W | 0.01% | 1 W |
| 5000 W | 0.02% | 1 W |
| 2000 W | 0.05% | 1 W |
| 1000 W | 0.1% | 1 W |
| 500 W | 0.2% | 1 W |
| 200 W | 0.5% | 1 W |
| 100 W | 1.0% | 1 W |
Floating Point Power Registers
The floating point power registers have a varying resolution, depending on the power value. The WattNode meter uses a 32 bit IEEE-754 floating point number. This has a 24 bit mantissa (including the hidden bit), so the effective resolution will be between one part in 223 (eight million) and one part 224 (16 million). This supports six or seven decimal digits of resolution.
At larger power values, the resolution progressively drops (in watts). For example, the following table shows the output resolution for several different power values. The second column shows the effective seven digit decimal resolution as bold digits.
| Power | Seven Digit Resolution | Approximate Resolution (W) | Approximate Resolution (% of reading) |
|---|---|---|---|
| 100 W | 100.0000 W | 0.0001 W | 0.0001% |
| 500 W | 500.0000 W | 0.0001 W | 0.00002% |
| 1000 W (1 kW) | 1000.000 W | 0.001 W | 0.0001% |
| 5000 W (5 kW) | 5000.000 W | 0.001 W | 0.00002% |
| 10,000 W (10 kW) | 10,000.00 W | 0.01 W | 0.0001% |
| 50,000 W (50 kW) | 50,000.00 W | 0.01 W | 0.00002% |
| 100,000 W (100 kW) | 100,000.0 W | 0.1 W | 0.0001% |
| 500,000 W (500 kW) | 500,000.0 W | 0.1 W | 0.00002% |
| 1000 kW (1 MW) | 1000.000 kW | 0.001 kW (1 W) | 0.0001% |
| 5000 KW (5 MW) | 5000.000 kW | 0.001 kW (1 W) | 0.00002% |
The raw internal measurement resolution with five second averaging is about one part in 1.3 million for a nominal power; dropping in proportion to the power. So at 1% of nominal power the resolution would drop to about one part in 13,000. The floating point resolution of a 32 bit float is between one part in eight million and one part in 16 million, regardless of the power.
Summary
Depending on the WattNode model and the CT rated amps, the power resolution will normally be limited by the raw internal measurement if you are using the floating point registers. Whereas if you use the integer power registers, then the integer resolution will be much lower than the internal resolution.
For floating point values, you can somewhat increase the effective resolution increasing the averaging to 20 or 60 seconds.
In any case, the WattNode accuracy (~0.5% or one part in 200) is normally the primary limitation, and also the noise will generally be higher than the floating point resolution.
