I am trying to understand how noise in electrical circuits can be modeled via stochastic differential equations (SDE). For example, the current through a resistor can be described as the sum of a noiseless, deterministic current and a current due to a Gaussian white noise process with spectral density $S(f)=\frac{2k_BT}{R}$, where $K_B$ is the Boltzmann constant, $T$ the temperature and $R$ the resistance of the resistor.

The additional current is then modelled as \begin{equation} I(t)=\sqrt{\frac{2k_BT}{R}}\xi(t), \end{equation} where $\xi(t)$ is a white noise process. This would enter the SDE as $\sqrt{\frac{2k_BT}{R}}dW(t)$, where $W(t)$ is a Wiener process.

Why do you take the square root of the spectral density here? Why does this white noise process have spectral density $\frac{2k_BT}{R}$? I know that the spectral density of a stochastic process is the Fourier transform of its auto-correlation function. But it seems I am missing something here.


The square root is a normalization. Note that $$\mathbb{E}(\xi(t) \xi(t + s)) = \delta(s)$$ and the Fourier transform of $\delta$ is just the constant $1$ everywhere. Thus, if you want white noise with spectral density $C$ instead of $1$, you use $\xi_{C} = \sqrt{C} \xi$ instead of $\xi$, which gives you $$\mathbb{E}(\xi_{C}(t) \xi_{C}(t + s)) = C \delta(s)$$ which has spectral density $C$.

A physicist might say you want the energy dissipation of the "random current" to be $C = \frac{k_{B} T}{R}$ so the random current should be $\sqrt{C} \xi$. After all, energy dissipates like current squared.

The constant $\frac{k_{B}T}{R}$ is from physical considerations. I remember seeing it in EE classes, but that's as much as I can say. I have a hunch there's a derivation in Kittel's Thermal Physics; if you like, I could report back when I get home. At any rate, Boltzmann's constant is something fundamental about statistical physics, often it's multiplied by $T$, and, well, you know what $R$ is.

Edit: I didn't find $\frac{k_{B}T}{R}$ in Kittel's book. I'll try EE textbooks at some point and get back to you.

  • $\begingroup$ Thank you for the explanation! It would be very interesting to understand the physical derivation of this constant. But there's no rush. Only if you have time for this! $\endgroup$ – Spike Jul 29 '17 at 21:51
  • $\begingroup$ This seems like the place to start: en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise#cite_ref-8 $\endgroup$ – fourierwho Jul 30 '17 at 18:26

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.