[time-nuts] re low noise regulators

[Bruce Griffiths]

A more accurate summary of the noise characteristics of voltage
regulators is something like (neglecting the effects of excessive noise
on the regulator input voltage):

1) For all regulators when the reference isn't low pass filtered the
reference noise dominates the output noise. Even when the reference
noise is low pass filtered it dominates the regulator output noise at
low frequencies.

2) When the reference is low pass filtered the error amplifier noise
dominates the midband regulator output noise.

3) The regulator output noise scales with output voltage. However a
capacitor bypassing one of the output voltage setting resistors in a
conventional voltage regulator circuit can reduce the midband output noise.

Using the 723 regulator as an example:

>From the datasheet:
Vout = 5V
output noise (100Hz - 10kHz) 86uV rms (unfiltered reference)
Output noise (100Hz - 10kHz) 5uV rms (filtered reference)

If we assume that in both cases the noise spectrum is flat (i.e white)
in the 100Hz - 10KHz band
If we also assume that the noise bandwidth is 10KHz (difficult to know
more precisely as the details of the band limiting filters used arent given)

Then crude approximations of the noise of the reference and error
amplifier are:

Reference noise (when attenuated to 5V)     860nV/rtHz     (in the 100Hz
to 10KHz band)
Error amplifier input noise                            50nV/rtHz      
(in the 100Hz to 10KHz band)


With an error of less than 30% (if high frequency band limiting filter
has a single pole noise bandwidth is actually ~16kHz) due to possible
error in noise bandwidth.

The reference noise (~1.2uV/rtHz at the 7V reference output) is somewhat
higher than that of a modern buried zener reference (50-100nV/rtHz).

The error amplifier input noise (50nV/rtHz) is also significantly higher
than that of most modern opamps.

Thus it is possible to design and build a voltage regulator using
discrete components, a low noise (buried zener) reference and a low
noise opamp that has lower output noise than a 723 based regulator.

To facilitate comparisons between regulators the estimates for the noise
of the 723 reference is perhaps best expressed in ppm/rtHz.

i.e. 723 reference noise is around 0.12ppm/rtHz in the (100Hz-10kHz) region.

A similar analysis is possible for other conventional regulators.
However if one requires more accurate characterisation of regulator
output noise there is no substitute for measuring the output noise of a
real regulator.

Comparing regulator output noise characteristics is perhaps easiest if
the relevant data (and assumptions made) are tabulated in the form of a
spreadsheet.
I am happy to do this (in both Excel and OO formats), just let me know
what regulators you would like to be included.
If anyone has actual measurements of regulator output noise noise please
post them together with actual circuit schematic and measurement method
details.

Bruce