Official Definition of Noise Gain

I found this:

Thanks, Paul!

I also found a similar reference.

I'm hoping there's a "catch-all" definition, but maybe it doesn't exist.

...Jim Thompson

UMM, "how much the opamp's input voltage noise is amplified by the time it gets to the output"?

Or maybe

"If, in a given circuit, one added a floating AC voltage source of unit amplitude in series with an amplifier's input pin, how many units of that signal would appear at the output"?

Noise gain must of course be a function of frequency, although one could default to the midband value.

John

That's silly. Noise gain has nothing to do with noise.

John

Of course not, but did you actually READ the paper ?:-)

...Jim Thompson

You are clearly not trained in the area ;-)

...Jim Thompson

Skimmed it. Too complicated to actually read.

John

Of course nobody "trained" me to design electronics. I'm not a dog.

John

I'd say it was partial dVout/dVin, measured at the noninverting input.

Cheers,

Phil Hobbs

That's what I'm concluding, also. Is that in general?

...Jim Thompson

Definition: A gain that happens during the months before major elections. Typically affects technical newsgroups the most ;-)

Well, in real life it's a convolution, of course, but if you don't mind defining it in the transform domain, then I think that's the definition.

To do a noise analysis, you refer all the noise sources to the inputs. If you've chosen the right ones, they're all uncorrelated, so they add in RMS (in general there's a cross term, but it's almost never needed).

Since an op amp is almost perfectly differential, you can put the noise source in either input, but choosing the + input makes it trivial to analyze--it's just the closed loop noninverting gain.

Cheers,

Phil Hobbs

Sno-o-o-o-ort!

Just got back from "Bucket List" and dinner.

Great movie if you're old enough to have experienced (and appreciated) some of the representations ;-)

...Jim Thompson

The paper I found has some useful stuff about calculating both a noise gain and an equivalent noise source. They show a case in which the noise voltage is fed to the op amp non-inverting input while the signal is fed to the inverting, giving different noise and signal gains.

But I've always found it easier to calculate an equivalent noise source connected to the same input as the signal and as a result subject to the same gain as the signal. It makes S/N ratios at various points along the system easier to think about.

"Noise gain (1/B): The voltage response given by the inverse of feedback-loop attenuation" --Walter Jung p.31

and, "The term that includes error factors due to nonideal gain is the _true noise gain_ or _closed-loop gain_... " on p.30

That confused idiot doesn't know what the hell he's talking about...When it comes to linear electronics, and especially operational amplifier circuits, you need to think about catalogs and directories of standardized circuits, which can be used in various non-standard ways. To make things more useful to the cookbook engineer, it makes sense to refer the total circuit noise to a standardized input, the non-inverting input in the case of operational amplifiers, and provide a gain function from that input to the output, called the noise gain. Then the cookbook engineer need only divide that noise gain function by his signal I/O gain function to refer the noise to his signal input and thereby derive a S/N for his particular application. By this reasoning, the noise gain is by definition the ratio of total output noise to input noise referred to a standardized input for the circuit topology under consideration. If you don't like the idea of an assumed standardized input then certainly you would have to take the standardized input as the one called out in the manufacturer's specification sheet for the input voltage and current noise density functions, and this is almost always the non-inverting input. Call it the datasheet specified input if you don't like standardized input terminology.

Jim, do you have a copy of the "bible", a.k.a. Motchenbacher & Fitchen? That's where I'd look first. Mine's on loan, sorry.

steve

That's how I've been using it. I've not noticed any other definitions.

Regards, Allan

I'll pull out Walt's book here in a few minutes when I finish my coffee, but I think that is true only from the non-inverting input.

Say you have a TIA, the noise-gain is dramatically different from closed loop gain.

...Jim Thompson