Yesterday, I rebuilt Rana's low-noise LED driver in the Bridge elab. It is based on a 2nd-order active lowpass filter (using the Sallen-Key topology). The schematic is shown below. The circuit is essentially the same as the one Rana posted a few days ago, only the R and C values are all around twice what they are in his schematic. This results in the same corner frequency of fc = 0.03 Hz.
I hooked it up and measured Vsk,out = 9.76 V. I then used it to drive a 50-ohm resistor, and measured VLED = 1.71 V, then measured the current to be ILED = 33.5 mA.
After ensuring that it was supplying the correct voltage, I hooked it up to the LED and took a spectrum of the voltage noise across it over the two frequency bands I have been using in previous posts. The following are comparison plots of the noise here and the noise with the simple RC filter used before, calibrated to displacement noise.
RA: although these plots have Displacement in the y-axis, they are NOT measurements of actual displacement noise. They are estimates for the contribution to the displacement noise made by the LED RIN based on measurements of the voltage noise across the LED.
Something is clearly wrong: not only is the new configuration worse at lower frequencies, but the rolloff seems to go as 1/f and not 1/f2. Investigating after dinner...