# AW: Re: Couple EE questions...

Magnus Danielson e93_mda at drum.it.kth.se
Wed Oct 23 19:35:21 CEST 1996

```>
> > Vnoise(RMS) =3D VnR =3D (4kTRB)^(1/2)
> [...]
> > The term "Noise Figure" is a dB scale value saying how much _more_ noise =
> > the
> > system has than the "perfect" system.
>
> Thanks for looking it up - yes that's the same as "Rauschzahl" then.

Yeap.

> > A small detail, it's the SNR and not the NF that should be optimised...
> > You can optimize the NF but destroy the SNR.
>
> Which brings me back to the original question: So choosing a higher
> impedance (and selecting a fitting opamp ... fet input etc) might
> improove NF, but not SNR.

Rigth.

> SNR should *always* be best at lowest impedance, regardless what
> active dvice you choose. So impedance should be always as low
> as possible (from a noise point of view). Then, if you have chosen
> an impedance, the second step is choosing an opamp (etc) that
> won't destroy SNR too much (i.e. which has a good NF for the
> selected impedance).
> As far as I see, the whole NF concept won't work the other way round.
> It's a method to select an active device to make the best of a given
> impedance, but *not* to select the impedance.
>
> The only way I can think of to *worsen* SNR by lowering impedance,
> is when you exceed the current drive capabillity of your active devices,
> i.e maximal signal voltage (the "S" in SNR) is reduced, but "N" doesn't
> change much anymore. (But that's a rather extreme case, IMO.)

On should select Rn = en / in

For a NE5532 that will turn out to be 5000/0.7 = 7.1 kohm

Given this source impedance and full swing of the NE5532 (12.7 Vrms) you get
close to 130 dB of signal/noise ratio....

Cheers,
Magnus

```