Making sense of ADP7118/ADP7142 datasheet

[pvs]

I am looking to use this "low noise LDO" to regulate a switched output feeding a Class A amp but am struggling a bit with the ADP7118 datasheet, particularly choosing Rnr and Cnr in the "noise reduction" section (page 16).

CNR is chosen by setting the reactance of CNR equal to R1 − RNR at a frequency between 1 Hz and 50 Hz.

This might as well be written in another language. No equation to plug into Wolfram. "Capacitive reactance" google results seem similarly aimed at engineers.

I do come up with an AD application note though (that I didn't see but might have missed on the AD website), AN-1329, which includes a section on improving PSRR (page 6), but that only muddies things even more. Where in the ADP7118 datasheet you are (ostensibly) advised to keep Rnr smaller than R2,

RNR is chosen to be small with respect to R2. If RNR is 1% to 10% of the value of R2, the minimum ac gain of the error amplifier is approximately 0.1 dB to 0.8 dB

in AN-1329 they use the same value for both in their example:

For this example, RFB1 = 64 kΩ, RFB2 = 10 kΩ, RNR = 10 kΩ, and CNR = 1 μF. The zero created by RFB1 and CNR is about 2.5 Hz and is evident by the improvement in the PSRR that is above 10 Hz.

I'm just looking for Rnr and Cnr values for several configurations 16-18v. Say for 18v if I use 54.9k/3.92k R1/R2 - what should I use for Rnr and Cnr for optimal performance?

Note ADP7142 is just 40v version of ADP7118. Also note the example in the ADP7118 datasheet uses the 5v ADP7118 which is adjustable but uses 5v reference voltage for SENS - ADP7142 datasheet, LTspice, AN-1329, and I, all use the adjustable version with 1.2v reference.

 

[MagicianT]

particularly choosing Rnr and Cnr in the "noise reduction" section (page 16).
This might as well be written in another language. No equation to plug into Wolfram. "Capacitive reactance" google results seem similarly aimed at engineers.

Read:

CNR is chosen by setting the reactance of CNR equal to .. RNR
at a frequency between 1 Hz and 50 Hz.

Of course, reactance is non-linear vs frequency, though it is not like CNR equal to RNR all over the range 1 to 50 Hz, but you free to select frequency point 1 or 10 or 30 Hz (depends on application) where CNR would be exact RNR.

For this example, RFB1 = 64 kΩ, RFB2 = 10 kΩ, RNR = 10 kΩ, and
CNR = 1 μF.

Very likely a typo, as on previous page they talk about 1k RNR.
In the essence, choose Rfnb1-2 depends on DC requirements, than set RNR as low as possible, limited by max injected current to sense pin (no data in DS, but usually 2-10 mA). All idea behind this noise reduction bla-bla-bla, is to lower AC impedance seen by sense pin as much as possible. In other words, NR network is necessary to remedy the culprits in design, to null out excessive increasing in noise in case of high impedance in series with sense input. Basically, LDO is low noise only when sense tied up to output, otherways it is not low noise anymore and needs additional components to keep performance at acceptable level.