Document created by analog-archivist on Feb 23, 2016
Version 1Show Document

Q

I would like to know if the power supply rejection ratio (PSRR) of the AD797
depends on the gain of the amplifier. Assuming a power supply with 100mVrms of
100Hz ripple and the circuit of the figure 44b with a non inverting gain of
1000. According to the figure 13 of the AD797's data sheet the PSRR of the
AD797 at 100Hz is 120dB, which means that a 100Hz ripple will be attenuated by
a factor 10^6. So the 100mV 100Hz ripple should be present at the output
reduced by a factor 10^6 which results in 0.1uV of 100Hz ripple at the output.
It can be neglected. But if this ripple is amplified by the gain it becomes
0.1uV*1000=100uV which is much higher and can't be neglected. That's why my
question is : Do I need to take into account the gain of the amplifier when
computing the effects of the PSRR?

A

The short answer is yes.  For several parameters on the typical op amp
datasheet, the errors are all referred back to the input. This is sometimes
denoted as “RTI”.   Examples would be PSRR, CMRR, noise voltage and noise
current.  This has been an industry standard for many years, and does allow you
to compute all the errors and apply them all at once.  As a side note,
The electrical table specification for PSRR and CMRR is a DC measurement.  We
therefore provide curves of PSRR vs. frequency.
With the advent of switching regulators at 20 kHz and above, it is very easy to
get power supply noise into you system. Having been an audiophile in the past,
I can sympathize with the problems encountered with building an MMC pre-amp.
PSRR can be measured by moving the +supply up and the –supply down by the same
amount.  This is what is typically done in the industry.  The AD797 is unique
in that it has separate curves for PSRR+ and PSRR-.  See figure 13 of the
datasheet. This is more useful because most people either run an op amp on
single supply, or do not have +/- tracking supplies.
For your application, it would not be unusual to have separate +/- linear
regulators to power only the preamp stage.  Another thing that is common is to
have cascaded regulators, such as 78L15/79L15 feeding 78L12/79L12,  but the
regulator itself can have a noisy output, so an RC filter may also be
required.  Layout and bypassing are also critical.  Even if the highest signal
frequency is 20 kHz, the op amp can oscillate at 10-20 MHz if you have some
stray capacitance in the wrong place.

There are some "oldies but goodies" app notes on our web site that may be
helpful.  Take a look at AN-202*, AN-214, AN-257, AN-345*, and AN-581.

I would read the * app notes first.  Even if the +/- supplies are clean, noise
can sneak into your system through analog ground.  In general, “ground” is not
the same as “ground” a few inches away.