There are so many variables and complications in the path from analog to digital to analog again that it’s really hard to find useful generalisations.
The main one you can rely on is this: most of what people don’t like about the sound of digital is the compromises involved in low-pass filtering. The lower the filter, the steeper the drop-off, and the less computation you have available to do the work, the worse the impact on the sound. That’s the main reason that high-res PCM can sound better than CD: the audio might not have been subjected to a less-than-excellent sub-22kHz LPF. These filters are needed during both recording and playback, unless you’re somebody who enjoys the particular sound of unfiltered “NOS” PCM playback.
DSD has an advantage in terms of filtering. Because the sample rate is so high, the LPF cut-off can be set way above the limit of human hearing and the roll-off can be very gentle. That’s most of why good DSD doesn’t sound “digital”. DSD playback is also vastly simpler in principle than PCM is, and the playback filtering is purely analog.
But DSD has a disadvantage in that it requires complex computation during production (sigma-delta modulation or SDM) to make a 1-bit sample depth give you more than 6dB of SNR in the audio band.
All this computation for low-pass filtering and sigma-delta modulation is a primary reason why new firmware for the Directstream DAC models sounds different. These DACS take in PCM or DSD and then perform LPF and SDM to create new DSD bits which flow through the output stage. Changes to the LPF and SDM algorithms and/or parameters change the sound.
That said, I agree with everybody else who said that the distribution format is one of the least significant things in determining how enjoyable that recording is to listen to. The contest is usually won or lost well before anybody decided whether to output as PCM versus DSD.