How does noise "work"?

[fryingpan]

This is one of those topics that have always been sort of a mystery to me.

Semiconductors add noise, high resistances add noise (and some technologies more so than others).

Despite this, oftentimes you can read things like "this 1M resistor doesn't contribute to noise much because it is shunted to ground". I do get that currents naturally flow towards the paths of least resistance, but when we are "sampling" voltage (say, at the input of an opamp, which for the most part has a relatively high input impedances and tends to be used with impedance matching) the opamp's input should still see the disturbance caused by the chaotic movement of electrons within the resistor, even though no or very little current is passed onto the opamp's input. So how does one go about at least guessing approximately the impact of a component on total noise?

Also, a gain element may have a so-and-so noise figure (which in isolation means very little, and needs to be integrated across the passband). Is that noise figure absolute regardless of gain or must also be multiplied together with all the rest?

[ElectricDruid]

As they love to say here in Portugal, "It's complicated".

I'd recommend getting a copy of Douglas Self's "Small Signal Audio Design" and reading what he has to say about it. That's about the best-informed reference I know. Anything I could tell you will have been gleaned from there anyway.

It's a great book anyway for those of us who are trying to learn about audio circuits and how to design them.

HTH

[antonis]

I'd recommend getting a copy of Douglas Self's "Small Signal Audio Design" and reading what he has to say about it. That's about the best-informed reference I know. Anything I could tell you will have been gleaned from there anyway.

+1  ..!! 

For a more detailed interpretation (with many practical implementations) : "The Art of Electronics" , 3^rd edition, P. Horowitz, W. Hill , pages 473 to 590

[PRR]

This is good, but a very heavy chew, and assumes you can reason well. Particularly series and parallel resistances.

Low-Noise Electronic Design by Motchenbacher and Fitchen

I bet Self has it on his shelf.

[antonis]

Low-Noise Electronic Design by Motchenbacher and Fitchen

Thanx Paul !!
(didn't happen to come to my attention till now..)

[ElectricDruid]

a very heavy chew

You make it sound like one of those leathery dog-bone things. Still, I bet if you chucked it to your Alsatian, that's exactly what they'd do to it...

[merlinb]

Resistors in series with the signal add their noise directly to the signal.
Resistors in shunt with the signal do not, rather their internal noise gets shunted by the source impedance.
In other words, you want series resistances to be as small as possible, shunt resistances to be as large as possible. This is also another way of saying 'attenuate the signal as little as possible'.

[Mark Hammer]

Here's the question I'd like answered.  Assuming we have a metal oxide and carbon comp resistor of identical low series resistance, will the one have more noise than the other, and if so, why?

[Rob Strand]

Here's the question I'd like answered.  Assuming we have a metal oxide and carbon comp resistor of identical low series resistance, will the one have more noise than the other, and if so, why?

In theory the resistance alone determines the thermal/Johnson noise.

In practice there's always more to it.  Resistors tend to have more noise than theory predicts.   That effect is put under the umbrella of "excess noise" - noise above the (simple) theoretical noise.    Carbon comp resistors have more excess noise than metal film.

https://www.mdpi.com/1424-8220/23/3/1107

The why is a question for research but in general materials that are less homogeneous, more pitted, have more surfaces tend to have higher excess noise.  In general weird stuff happens on surfaces and that's true for semiconductors.

I should add that Johnson noise is a physical effect not a flaw in the material or manufacturing.

Semiconductor junctions have Shot noise which is a different physical process.  It relates to individual charges passing through a barrier.  Think of raindrops vs a continuous stream.   You can pass the same amount of water but one is "noisy".   I posted more details on this in the past.   Again this isn't a flaw and there is an "excess noise" in semiconductors as well.

[amptramp]

One of the components of noise is 1/f noise, a phenomenon where the noise goes up below a certain frequency and it is usually attributed to surface effects.  It is one reason why a buried zener diode is not as noisy as a surface junction zener diode and in fact, you can make a low-noise regulator using the LM399 which is a subsurface zener diode with internal temperature compensation and current compensation circuitry added around it.

Popcorn noise is an effect that sounds in audio equipment like popcorn being popped.  This is due to the presence of defects in a crystal lattice of a semiconductor that cause the output to switch randomly from one fixed level to another fixed level.

All of this is in addition to the Johnson noise which varies with the square root of the resistance.

[ElectricDruid]

Is popcorn noise the same thing as shot noise, or are those different?

Thanks.

[antonis]

Just to add the fact that, even if there aren't "physical" resistors (items) in a semiconductor circuit, noise is still generated on invisible resistances (like r_bb' & r_e for BJTs and 1/g_m for FETs) either in the form of equivalent Johnson noise or in the form of Shot noise (current through Collector-Emitter junction or Drain-Source channel)..

FWIW, a comparison chart between BJTs, JFETs & MosFets..

[antonis]

Is popcorn noise the same thing as shot noise, or are those different?

Different..

Shot noise is like "rain on a tin roof" where Burst noise is like " the birth throes (thanx Paul) of popcorn"..

(additionally, Shot noise exhibits Gausian distribution where Burst noise looks like random jumps between two (or more) voltage levels at unpredictable frequency..)

[PRR]

...Burst noise is like " the birth throws of popcorn"..

Or maybe "throes" :

throes /THrōz/ noun plural noun: throes; noun: throe
intense or violent pain and struggle, especially accompanying birth, death, or great change.
    "he convulsed in his death throes"

Yes, English is a bastard language. (I even like "throws" better, but it bothers my eyes.)

Sound clip (ignore the screaming after 2:41)

[antonis]