"John Larkin"
** Wrong as usual........I did a test a few days ago using 0.5 watt MF, 2 watt MF & finally some old
0.5 watt carbon film resistors - using two similar resistors in series with a (well filtered) variable DC supply.The mid point was cap coupled to a high input Z JFET pre-amp followed by a
1000 times gain stage ( a balanced mic pre ) and finally an audio band filter with gain of 10. This set-up showed only a dB or two more residual noise than predicted from thermal calculations when using a pair of 100k resistors ( of whatever type).I found no trouble identifying lots of "excess noise" as soon as DC bias was applied with 20 to 40 volts bringing the excess noise up to a level matching the previous thermal (ie a 3dB increase).
The oft published figure for MF types of " 0.1uV/V " per decade of frequency was found to be about right.
The old carbon film types I tried gave about 2.5 times more excess noise than the MF ones at similar DC bias voltages.
The 2 watt MF power resistors showed about 50% more noise than their lower powered cousins, plus exhibited a sharp crackling noise as well as pink noise. I did not try cermet types but from experience would expect them to be much noisier than the others.
Also, resistor excess noise is largely independent of value - so using lower values than 100kohms will shift the DC bias level *downwards* for similar results re the ratio of thermal to excess noise.
Eg: for 1 kohms, thermal noise is lower by a factor of 10 (compared to
100kohms ) so he DC bias needed will be 1/10 for the same ratio.Certainly not a trivial issue nor something one can afford to ignore.
..... Phil