Measuring RMS noise...

[Mark Hennessy]

Another problem is Crest Factor; the ratio between the mean and peak value. For a sine wave, it's obviously 1.414 (root-2), but for a pulse of 10% duty cycle, it's 3.16 (root-10). A typical multimeter might only be reasonably accurate for crest factors of 3. The specification for the venerable Fluke 87V states:

AC crest factor can be up to 3 at full scale, 6 at half scale. For non-sinusoidal wave forms add -(2 % Rdg + 2 % full scale) typical, for a crest factor up to 3.

The bandwidth of the instrument is obviously important - I mentioned that in another thread earlier today. In audio, there are many standards for measuring noise, and BW is an essential thing to state. Weighting networks (to approximate the ear) are common. In addition, RMS sensing is not desirable for audio. Using quasi-peak rectifiers give results that are more representative of how we respond to noise subjectively (which is why Dolby prefer average-responding rectifiers!).

However you look at it, AC is tricky stuff

[ppppenguin]

Craig correctly points out that the ultimate way of referencing AC measurements to the DC standards is thermal. It's horrible but the best we've got.

For most of us doing everyday things a decent meter plus awareness of its limitations is sufficient. As Mark says, for noise measurements you really need to be aware of exactly what you're measuring. For analogue video there's the additional problem that a given amount of noise becomes more visible the closer you get to black level. This is the main reason why video is sent using a low gamma. When you look into that more closely it's another nice can of rather wriggly worms.

[Nick]

My favorite DMM at home, a Tektronix DMM916, also has a crest factor of 3 (max peak 1000V). 3dB bandwidth for True RMS is 20kHz.