The average electric power for a sinusoidal waveform is:

Every multiplication factor contributes in its own way to the total uncertainty:

When measuring power, many small errors occur. For a total inaccuracy figure we have to sum them all together. Many small possible errors result in one big uncertainty figure, so we better pay attention to all of them.

Factors that influence the total error for power are:

• Amplitude
• Frequency of the fundamental of the input signal
• Crest Factor of the input signal (BW)
• Crest Factor setting of the Power Meter
• Power Factor
• Temperature
• Use of filters

When we measure voltage and current there is an uncertainty about the amplitude caused by the FREQUENCY CHARACTERISTICS (= BW=Bandwidth) of the measuring instrument itself.

Furthermore every range setting incorporates unique components, so every range will contribute differently to the total accuracy. This error is called the RANGE ERROR.

Within a range it makes a difference if the actual measurement value is close to 0% or 100% of the maximum range value. The instrument itself is calibrated at the 100% value, so this gives always the best result and close to 0% (entering the "noise floor") the worst. This error is called the READING ERROR. For Voltage, Current and Power measurements, the Error is specified as: (Error as % of reading) x reading value + (Error as % of range) x range value.

Basically this uncertainty is measured and calculated at the condition of

Input circuitry of the digital power meter

An unwanted phase-shift is caused by the different phase and frequency characteristics of the voltage and current inputs circuitry's.
The current shunt generates by having a few mH inductance, a little extra phase shift to the current input. The voltage input does not have this component.
Even when PF=1 this error occurs but it's effect is neglectable at 50/60Hz in relation to the reading value while the product Urms and Irms is at its max. value.
At every frequency we have a phase delay error related to the range. This error is measured in degrees, but later specified as a % of the range for ease of use.

If the freq. increases, the relative impact increases. At higher frequencies an extra error has to be added to the total power error. Even for the PF = 1 condition this error is no longer neglectable.