How this calculation works
RMS is zero only when every input is zero. Opposite signs cannot cancel after squaring.
RMS = √[(a²+b²+c²)/3]A worked example
The mean of the squares of −3, 0, and 3 is 6, so the RMS is √6.
Good to know
The three observations have equal weight.
Reference: OpenStax Algebra and Trigonometry (opens in a new tab) ↗
Square first so opposite signs do not cancel
For −3, 0 and 3, the ordinary signed mean is zero. The squared values are 9, 0 and 9, whose mean is 6. Taking the square root gives RMS = √6, approximately 2.44949. The SVG traces the squared contributions and the division by the three observations.
Squaring removes the sign and gives larger magnitudes more influence. The final square root restores the original units. If the inputs measure volts, the RMS also measures volts, while the intermediate mean square measures square volts.
Relate RMS to variability and weighting
For equally weighted observations, RMS² equals the population variance plus the square of the arithmetic mean. Expanding each (x − mean)² establishes this identity. In the symmetric default example the mean is zero, so the population standard deviation and RMS coincide. With three identical values of −3, the RMS would be 3 although the standard deviation would be zero.
The Harmonic mean of three positive values calculator averages rates under equal-quantity conditions, while RMS measures a different property. These three input fields give each observation equal weight. Samples representing unequal time durations require a weighted mean square. The implementation uses a hypotenuse calculation divided by √3, which avoids some overflow and underflow problems caused by squaring the inputs directly.
A few questions, answered
Can the RMS be negative?
No. It is the nonnegative square root of a mean of squares.
Does three-sample RMS describe an entire waveform?
It describes those three equally weighted samples. A waveform RMS needs representative sampling or an integral over the intended time interval.