In the case of statistics, there are two terms: "Variance" and "Standard Deviation" to - for in the latter case - have one to note the case in which the units are without squares.

Am I correct to understand, that there is no such differentiation in signal processing, and in both cases - taken root or not - it's always termed "Power Spectral Density"?

I am asking as on page 3 of Analog Devices AD8307 there's the unit of $\frac{nV}{\sqrt{Hz}}$ for the "Input Noise Spectral Density".

Is there a specific reason as to why manufacturers are taking the square-root?


Power spectral density is given in power per $Hz$. "Input Noise Spectral Density" is given in units of volts per $\sqrt{Hz}$. These are related:

$$ PSD = \frac{W}{Hz} $$

$$ W = \frac{V^{2}}{R} $$

$$ Input\ noise\ spectral\ density\ = \frac{V}{\sqrt{Hz}} $$

The $\sqrt{Hz}$ is from the conversion power to volts. For example if you want the noise ($nV$) in $1kHz$ $BW$ you must multiply $\frac{nV}{\sqrt{Hz}}\sqrt{1000}$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.