First, was the same thing measured?
Many differences are due to conditions and not to the quality of the instrument. A different point on the circuit, an inverted phase, a load operating at another level or at another time, or time windows that do not match are enough to produce different and legitimate results.
Differences in calculation method
- Window and synchronization: the Fourier transform requires a window with a whole number of cycles. If the instrument does not track the network frequency well, spectral leakage appears.
- Grouping: some report individual lines and others harmonic groups or subgroups that include neighboring components, which changes the value when interharmonics are present.
- Aggregation interval and statistic: an instantaneous maximum, an average over several minutes and a high percentile are not comparable.
- Percentage base: percentage of the fundamental, of nominal or of demand (for current, THD and TDD are different).
Hardware differences
- Performance class: an instrument designed to the strictest measurement requirements is expected to be more repeatable than a general-purpose one.
- Current sensors: bandwidth, linearity at low current, phase shift and inherent noise.
- Noise floor and resolution: these matter for low-level harmonics.
- Anti-aliasing filters and sampling frequency, which limit the highest order measured faithfully.
What to check first
- That both measure at the same point, with the same transformation ratios and polarity.
- That the recording period and time are synchronized.
- That the aggregation, statistic and percentage base settings are identical.
- That the current sensors have comparable characteristics.
- If they still differ, measure with both on the same signal from a reference source.
How to interpret the result
Small differences in low-value orders are normal. A large difference in a dominant harmonic usually reveals a configuration, scale or sensor error. Before attributing the discrepancy to one of the instruments, both configurations are documented and the differences in method are corrected.