Here's some info from an Audiophilia review:
PS Audio linked up with an eccentric and extraordinary engineer, Ted Smith, who is the lead designer and original prototype builder to create this ‘revolutionary’ DAC.

1. [The PCM notation x/y is short for x bits per sample of the amplitude of the analog signal at a sampling rate of y kHz. DSD uses a delta-sigma modulation method instead, in which samples of the delta (∆) (difference between successive samples) of the amplitude are measured and converted to a 1 bit per sample stream (of 0s or 1s) at a very high rate of 2.8224 MHz (denoted by DSD64), or at twice that rate, 5.6MHz (denoted by DSD128). The DS converts all fed to it into DSD128. 1]↩

The DS operates on an entirely different basis from the PWD or any DAC that I know of, and I point out right away that its main selling point is not to encourage people to go out and buy (waste money on?) expensive DSD high-resolution audio files: It’s to demonstrate that even CD 16/44.1 resolution PCM files can reveal extraordinary new details/harmonics when played properly, that is, in this new way. It is not about what format the audio is recorded in or converted to before it hits the DAC, it is about how you play it back: Whatever digital format you throw at the DS (PCM/DSD) gets converted by the DS to a 1-bit DSD digital stream, converted to analog and played bit perfect as such. This requires upsampling/processing the digital input into DSD format, and even DSD files too get upsampled to 30 bits at 10 times DSD rate then, for noise shaping, filtered/downsampled to twice the native DSD rate (yielding 5.6MHz) as the final 1 bit DSD product. The DS uses no off-the-shelf IC DAC chip such as the PWD’s superb Wolfson WM8741, or the very highly regarded ESS Sabre ES9018 used in many recent DACs that currently can handle/play DSD as input. In fact it uses no traditional such chip at all: Instead, it uses a field programmable gate array (FPGA); a Xilinx Spartan 6 FPGA.

This allows PS Audio to run proprietary software that does all the switching, digital conversion and filtering throughout. They are not unique by using FPGA in a DAC unit; but they are unique in the way they are using it, and for what purpose: to process (entirely) the digital signal to yield what is their ultimate goal—1 bit DSD format as output for anything originally handed to it and with no jitter. But, to accomplish this, they also (of course) have to deal with power supply and clock issues which involve other highly proprietary components/methods within the DS. The DS uses only one master clock, for example, PS Audio arguing that this reduces jitter even further. It is a very serious, very important and very expensive component of the DS: it is a voltage-controlled crystal oscillator (VCXO) made by Crystek Corporation. And, no, it is not an ‘atomic clock’—which quite frankly would be as silly as using Einstein’s General Theory of Relativity to ensure that one arrives on time to an 11:00AM meeting in room 311, when walking from one’s own office in room 310 next door at 10:59AM. Of particular note is that the path from digital input to analog output on the DS is much simpler and elegant than for the path required when using the very finest of PCM based processors; check out the diagram ‘Liberation from a Chip and Complexity’, under ‘Overview’ on the website.

One advantage of DSD as a format is that it is simple to just low pass filter it for conversion to analog. But, of course, this is easier said than done: PS Audio’s method of conversion at the end is very sophisticated involving the use of a passive (versus active) low pass filter via a passive audio output transformer. This results in another notable change from the PWD: the volume control is no longer a digital volume control built inside an IC DAC chip. This is actually a very important new feature. I have lately been using a passive preamp/attenuator for my PWD (instead of going directly from PWD to amps) since it sounds better (cleaner and more transparent) on my system than ‘direct to amps’: The volume level of the PWD is only bit perfect when at 100%, which is how it can (and should) be set when using an attenuator. DS volume control is bit perfect at any volume level. Thus, plugging directly into your amps and using the DAC’s volume control is now an even better justified option than it was for the PWD.

In any case, PS Audio claims that the increase of details and harmonics revealed/unmasked by the DS is a major improvement over traditional high-end DACs, and that this improvement is very significant even for CD rips, which are of course PCM at 16/44.1, and are the foundation for most people’s digital audio libraries. On the PS Audio website it states: ‘True DSD core engines (compared to the standard multibit Sigma-Delta converters followed by random lower quality multibit converters) offer advantages in simplicity, linearity, and in analog-like overload characteristics that avoid PCM’s hard clipping potential and a PCM processor’s propensity to mask subtle details.’

Sound Quality

From the DS, cymbals, bells, blocks and other percussion had a rich and noble, long, natural lingering decay, and voices revealed a kind of resonating halo surrounding the singer offering extraordinary natural acoustics such as reverb. In general, individual instruments of any kind seemed to have more space and air surrounding them with the natural acoustics of their venue on display. Subtle details including external noises in live performances, such as clapping, breathing and tapping, were apparent more often and in a much stronger way. The imaging and transparency alone were mind-boggling. All of this was very noticeable. A sizzler cymbal, for example, sizzled away with a beautiful lingering decay.

Peabody - I heard a lot of the same things but he says it better