It is board with DAC, of course analog power supply is separated. This spikes seen after LT3045 on 1.0V XMOS power rail, seems to be strange for such good regulator, anyway it sounds very good instead of PWM regulator. I'll try to power it from separate 1.5V transformer winding, may be LT behaves like this because of big voltage drop from 5V to 1VWhich PCB are you using? One way to avoid processor switching noise in the analog circuits is to have completely separate power traces for digital and analog, along with sufficient bypass filtering on all chips - both digital and analog. One common practice is to place two different capacitor values in parallel - the frequencies that one cap lets through will be filtered by the other. Bottom line: Since the processor draws a spike of current on every clock pulse, the power rail voltage (both power and ground) will see the increased voltage drop due to that current spike
As a result of experiments, I came to the conclusion that the most striking result is given by separation of power supply of VDD(core) and PLL_AVDD rails. Awesome improvement for sound. Now using two 1.0V regulators for these rails. It may be very usefull to users with different versions of XMOS, just look at schematics and datasheet (btw there is an simplified schematics in it, where these power supply pins are using the same regulator, but in text there mentioned that "The PLL_AVDD supply should be separated from the other noisier supplies on the board. The PLL requires a very clean power supply").