The isoPower products have a different noise signature than a standard DC/DC converter. Since the internal Tank Oscillator runs at 180MHz, and when rectified on the secondary this turns into 360MHz there is high frequency noise riding on top of the on portion 600KHz PWM control signal. The 600kHz ripple and 360MHz noise can be controlled to some extent with bypass capacitance, adding more low capacitance low inductance bypass will reduce the high frequency noise, and adding more bulk capacitance will reduce ripple. What I recommended in the data sheet is not aggressively trying to control noise and ripple.
The noise is hardest to control, but of such high frequency that most applications ignore it. Whether this is a problem for you may depend on the details of your design, I would suggest running this past the AD7794 applications people.
the AD7794 is a high resoluion device so a low noise supply is required for best results. The 5V from the USB is quite noisy. This is the reason why the AD7794 is noisier when the 5V USB supply is used. The AD7794 has been evaluated with low noise linear regulators where the performance of the AD7794 is optimised.
The frequencies originating from the ADUM5000 are high and outside the AD7794's useful band. However, they may alias back into the band of interest. Some RC filtering is advised on the analog inputs and reference inputs. If the analog inputs are buffered, then add a series resistor of 1K, a 0.1uF cap from AIN+ to AIN- and a 0.01uF cap frm each input to GND. If the analog inputs are unbuffered, the RC values must be limited. Consult the AD7794 datasheet for suitable values.
On the reference inputs, add a 1nF cap from REF+ to REF-. The reference inputs are unbuffered so large RC values cannot be used.
The power supply pins should have a 10uF and 0.1uF cap to GND. A small series resistor of 1.5 ohms or so will help to reduce any ripple seen by the ADC.
I agree with Mary's comments.
The ADuM5000 creates ripple on the supply at about 600kHz due to the control loop, which may be in band for your measurements. The internal power oscillator runs at 180MHz which is likely high enough not to cause direct probelms. The ripple can be further reduced by increasing the size of the large bypass cap between the ADuM5000 Viso and GNDiso lines. Higher values will not have an adverse affect on the operation of the power supply.