Re: A Simple Lab Demo of Resistor Noise EMF.

["John S. Denker" <jsd@MONMOUTH.COM>]

At 05:34 PM 3/18/01 -0600, brian whatcott wrote:

> ... here's a simple, cheap, straight forward protocol I hope
> may be useful in the high school lab:
> ...
> 4) Connect each meter lead to one lead of a resistor of
> at least 120 kilohms.
>
> 5) Heat the resistor red hot. (If the resistor goes open circuit,
>  you will see that the DVM drifts up to 100s of millivalts.
> Replace the resistor in this case.)
>
> 6) Note the reading of 4, 5 or 6 millivolts.
>
> 7) Remove from flame and check that there is a readable resistance value.
>
> 8) Repeat from step 5)
>
> This is a reasonable classroom exercise. Just a hint of danger to keep
> the students interested.

It has certainly piqued my interest, in a picaresque sense.

The thing that particularly provocative is that my Fluke multimeter has a
100 kHz bandwidth for its AC true-RMS measurements.  Using this number, and
the numbers Brian provided, and a generous estimate for the "red hot"
temperature, I get:
        kB = 1.38e-23
        T = 2000 Kelvin
        R = 120 kOhm
        BW = 100 kHz
whence we can calculate
        V = 0.02 mV RMS (thermal noise)

So the reported 4 or 5 or 6 mV can't possibly be thermal noise.  Even in
the unlikely event that Brian's radio shack voltmeter has 10x more
bandwidth than I think it does, and even if his resistor is 10x more
resistive than advertised, and even if the temperature is 3x hotter than I
estimate.... that still can't provide a thermal-noise explanation for the
reported voltage.

I have tried to come up with a plausible explanation for the reported
observations, so far without success.

The only resistors immediately available to me, at any temperature up to
destruction, have a thermal voltage that is not measurable with my
Tektronix scope or my Fluke multimeter -- as expected.

I suggest that if you want to see thermal noise, you should try looking for
nanovolts, not millivolts.  This will require a rather better preamp than
comes with ye olde $50.00 oscilloscope or ye olde radio-shack multimeter.

For example:
  http://www.thinksrs.com/html/body_sr560.html
see in particular the noise contours at
  http://www.thinksrs.com/html/sr560technote.html


Or if you are feeling particularly wizardly and want to build something
from scratch:
  http://www.national.com/ds/CL/CLC5509.pdf