The signal is “changed” but not “degraded”. I use a Sigtech with with the dCS elgar/972 combo directly into modified Melos monoblocks driving Dunlavy SC-Vs. On bypass mode vs. taking the unit out there is virtually no difference in sound quality. There may be some due to the additional cable (I use Orchids) but it is at or below that level.
Explaining why rapidly gets into the theory and implementation of digital filters. There are several challenges to implementation. First the mathematical theory is difficult and there are several possible choices for implementation. Second, the design needs to implement the math in algorithms (programs) that run on available chip sets. Finally, there must be someway to measure your room response, create the right algorithm and input it to the device to null out the speaker/room interactions. All of these factors impact the quality of any digital processing unit. For more info on Digital Signal Processing I recommend the technical papers on the dCS, Z-Systems and Sigtech web sites. For all of these firms their primary lines are professional digital audio processing.
The short answer for why a well designed DSP doesn't degrade the signal is that the algorithms have a bit precision that is greater and calculation rates that are faster than the Redbook CD standard. Sigtech’s internal bit precision is 56 bits compared to the CD’s 16 bits and with multiple 250 MIPS chips it can calculate a full bandwidth 2544 tap filter at the Redbook 44.1 kHz sampling rate.
The desired sound at your ear should be phase accurate and with a flat frequency response relative to that encoded in the original bit stream. The speaker and the room interaction “degrades” the sound through standing waves, first refection bounce, etc. All of this can be thought of as distortion created as a by-product of the room/speaker interaction.
The secret to good room treatment is to smooth the frequency response while damping the first and most of the second reflections while leaving the direct sound and later reflections to provide air. Sound travels at about 1 foot per millisecond. Reflections under 5-7 ms are perceived as colorations. Over 30 ms the perception is of ambiance, reverberation or echo. The best DSP units use algorithms that are designed with the psychoacoustics in mind as well as with variable amounts correction with respect to frequency so as to not over correct a speakers inherent response. If your speakers are already impulse coherent to begin with, like the Dunlavys, all of the DSP power can be applied to room correction. The Sigtech can correct up to the first 50 milliseconds.
“The Processor causes the speaker to emit a succession of wave launches (utilizing existing music signals) that have the same time-delays as the room reflections. Each delayed signal arrives at the listener's chair simultaneously with a room reflection, but with an inverted pattern of peaks and valleys...(which)... cancelled out the peaks and valleys in the room's reflections....In effect, the DSP pre-conditions the musical signal going to the speakers so the signal arriving at your head mimics the sound of an ideal speaker in an ideal room.”
Sorry for the long post but I hope this was reasonably clear.