Noise and its weird units of V per sqrt Hz (Amplifiers #12)

Noise amplitude spectral density has a weird unit of volts per square root of bandwidth. Why does it have such a strange unit? Then let's consider that noise represents power circulating in a circuit. Can we somehow extract useful power from this noise? The answer is a little surprising. We'll conclude with some practical implications for circuit design.
Aaron Danner is a professor in the Department of Electrical and Computer Engineering at the National University of Singapore.
danner.group
Video filmed and edited by Cheryl Lim.
@randomcheryl

[@comeflywithme1694]

Outstanding work. Thank you so much.

[@epeglab1o1]

This was super helpful! Thank you

[@umerfarooqazam1399]

Professor doing wonderful job

[@juskolk5839]

Hello Professor! I really enjoy following your videos as they are very easy to understand. I would like to request if you could possibly prepare a series of lectures on the topic of EMI (common-mode and differential-mode interferences) :)

[@andymouse]

Awesome tutorial...cheers!

[@nathanfinnegan522]

Great vids!

[@EngineerAnandu]

Thank u Dear Sir.

[@SergiuCosminViorel]

Keeping it in practical terms, ok. And a natural limitting of the bandwith, it is a problem of processes inertia. Process inertia brings that tail down, and there is always something of that. Even if you intervene, placing say an RC LPF, you are modelling process inertia. That natural inertia dose, it is not contained in the formula we use to calculate noise, so we can consider that formula not to be perfect science. Being just a limitted truth, we cannot ask the circuit to satisfy that and not generate infinite power. That limitation does not come to satisfy the partial truth. What needs to be justified, it is that partial truth, and it does not impose further constraints on its root causes. Your approach, perfectly practical. Good video! My approach, looks not practical, that is until you move to quantum processing. Ther they found the power distribution function not tailed down, and it only tails down there. It means the model of resistor is more cappable than any detector, and that tailing down of the power distribution, is mainly a detector proble. Resistor noise keeps on going, untailed down, to infinity, and only some transistors tail that down. Natural limitation exists, of course, and it goes into the quantum transitions, that are so very fast! That tailing down is a matter of the quality of... matter, particle level, and the necessity to avoid some energy paradox is written inside the particles.

[@andymouse]

Indeed.

[@SergiuCosminViorel]

Am back again! Am not consulting any sources, only my understanding. Many years since i studied... So, the way i understand it, and why noise is not tailed down as you say: The contributors, that are making all the noise, are quantum transitions of electrons, and that gives the end of the noise domain. These contributors contribute perfectly random, therefore you can have, for say n contributors, with equal probability, all cases from isolated unique transition, to all contributors contributing insequence, giving uninterrupted signal, n-transitions long. It is combinatorics. The highet frequence in that noise, corresponds to the quantum transition speed. Before that, there is no taling down. How big that is? I would not calculate, but if you check, i believe you will be awed by how big is that frequency. How big is the power that can be obtained from noise? Well, big ehough! It is the power of the heat transfer we can manage to obtain, well, significant part of that power. We use to think noise in stationary regim, but if we wonder how much power we can draw from noise, well, we can get power from noise, so to speak...