Effect of power supply ripple on ADC readings?

Question

I want to use an ADC (AD7124 that is a 24 bit sigma/delta) for 14 bit readings of an external voltage. My power source is a fairly stable 24v DC supply and the ADC runs on 3.3v. I will be using the internal 2.5v voltage reference of the ADC.

I'm planning on using a MC34063 switching dc-dc converter to reduce the voltage down to 3.3 volts. Some say this is a bad idea because of the voltage ripple and spikes on the output of switching supplies.

The only effect I can think of that voltage ripple has on ADC readings, is on the internal voltage reference. According to the datasheet, the internal reference has a "power supply rejection" value of 85 db. I'm guessing power supply ripple is attenuated by 85 db and then appears as an error on the internal reference.

Question 1: Is this correct? If yes does this change with frequency? The datasheet has nothing on frequency.

Question 2: How else does power supply ripple affect ADC readings?

Question 3: If using a step down switching supply is a bad idea, will adding a L200c linear regulator solve the problem? (L200c is probably overkill but I need the current limiting function and that is the only thing I have available)

Thank you

Answer 1

PSRR always reduces with frequency, but let us put this in context:

The converter has a very low maximum sample rate, and you care about noise in this band so switching artefacts are unlikely to cause issues (check out the filterung done within the device); it is possible that the ripple of the switch mode device may give some intermodulation products at the reference, but they will be uncorrelated with the sample rate of the converter (i.e. they are highly unlikely to be synchronous).

Note that from DC to < 20kHz (the maximum available sample rate), PSRR is unlikely to change in any meaningful way for the reference.

Using multiple samples and averaging will eliminate such noise; that said, it might be a good idea to feed the Avdd pin through a ferrite to eliminate the majority of the noise.

I am not sure a post linear regulator is actually necessary when you consider that the reference has a tolerance (absolute) of +/- 0.2%

Answer 2

I use switching regulators (Linear Technologies, many different ICs) to supply very sensitive and highly demanding ADCs from 500 kS/s up to 150 MS/s sampling rate. All ADCs work properly (up to 16 bits precision). There is no big value in using linear regulators after switching ones.

LT switching regulators (in my designs) operate on frequencies from 0.5 MHz to 3 MHz, using MLCC (ceramic) capacitors for filtering.

The proper filtering, decoupling and PCB design makes almost perfect and clean power. You may need to add one or two filters between switching regulator and your ADC.

The PCB design is a key factor for success: you must put all AC currents of the regulator far from ground / signal paths of your ADC. The PCB must have at least 2 layers, preferably 4 layers.