mike wrote...

Long story short, I once tried to build a stable current source for a

yig oscillator. I never could get the noise level low enough to be

useful, much less worry about high stability. The TEK 492 spectrum

analyzer uses a super-beta transistor pair driving a low noise op-amp

driving a transistor with a low-noise resistor in the emitter.

Be careful, a "low-noise" resistor implies low voltage noise, which by

the Johnson-noise equation implies LOW resistance values. For example,

e_n = sqrt (4kT R) = 1.28nV/rt-Hz for a 100-ohm resistor.

But current noise in a resistor is i_n = v/R = e_n/R = sqrt (4kT / R),

which is another beast entirely and tell us that HIGH resistor values

are required for low current noise. For example, i_n = 12.8pA/rt-Hz

for a 100-ohm resistor in a 10mA current source, but i_n = 1.28pA for

a 10k resistor in the same current source! Yes, using the 10k resistor

would require a 100V voltage reference, but it would be 10x more quiet.

This effect continues for higher DC voltage drops across the resistor,

but somewhere above 100V one runs into trouble with base-current noise

and a performance limit is reached. Was that your experience?