Getting it wrong

We've been covering how a DSD/SACD converter works and we learned a few days ago that just about all modern DACS, certainly those that advertise 24 bit resolution, use a type of DSD on their outputs. It isn't exactly 1-bit, but it is DSD-like. So important is this technology and so widespread is its use in modern high end audio DACS that it might help us gain further clarity by simply visualizing today's DACS as PCM to DSD converters. One easy way to think about DSD is, as I have mentioned before, the fact that regardless of how the bits are managed (1-bit, several bits, various versions of PDM) they all have something in common that is not shared by PCM: the ability to stream the output through a simple low pass filter and hear audio. For sake of understanding let's just stick with our 1-bit explanation because it's easier to get a handle on the concept without getting overly technical. Just to refresh your memory, a DSD system uses what is known as PDM to encode its signal - where the density of the 1-bit signals gets more for louder levels and less for quieter levels. Here again is our visual of what this looks like. Also remember that one of the reasons this modulation is used, rather than the fixed code method we find in a standard CD, is that DSD is capable of amazing low level detail - as seen in its ability to render many more low level detail bits and wider dynamic range than can be handled by any other means. It's helpful to note that any DAC is capable of full output - but not every DAC can handle the quieter details and it is those quiet details that are added when we add more bits. By this I mean for you to understand that the maximum output loudness in digital audio can be achieved by even a lowly 8 bit DAC but the low level details, what's happening down in the quietest passages found in a 144dB signal range can only be handled with a 1-bit encoder/decoder. So it is these low level details that we'll want to look at. Because DSD systems don't rely on a fixed code to assign a stepped voltage, the perfect question here is "what does it use"? It uses what's known as error feedback to get those quiet passages right. Let's take a closer look tomorrow at how getting it wrong gets it right.