[Radio Wrangler]

That's rather a good paper on excess noise created by a DC bias current, essentially fluctuations in the resistance of a resistor. Their figure 6 compares a large number of makes and types of surface mount resistors. As can be seen, the best ones give very little excess noise even down at 1Hz, while the worst on their plot are 50dB worse.

As a rough guide, the difference between good parts and bad ones is 20dB at 1kHz.

The Johnson noise is the thermal floor where a resistor carrying no current acts as a generator, converting heat into random electrical noise, and is dead flat. The other curves show fluctuations in the resistance modulating the voltage across it. This noise output is powered by the DC bias current the resistor is subjected to.

Note that this plot is limited to surface mount parts and 100 Ohm values. Carbon comps would show up as very bad.

NIST (boulder colorado) and NPL (London) maintain noise reference resistors at carefully controlled temperatures. These items are chosen for negligible excess noise at RF and are used as primary standards for calibrating solid-state noise sources for use as transfer standards for noise measurements. The fun part of these standards is calculating corrections of the noise contributions of cable losses across the thermal gradient.

Figure 6 of the paper above, shows you just which types you want to avoid in low noise audio applications and in the tuning lines of frequency synthesisers which can be fiercely sensitive.

David