Would someone explain what "up sampling" means and how it can affect the sound?

I am looking at buying a DAC and after reading a few articles, there seems to be a question that comes to mind: what is up sampling, and does it affect the sound ? Some audio companies have up sampling in their DAC and some do not.
I just want to make an educated decision about what DAC I purchase that, hopefully, will not be obsolete in a few years.
Thank you for all your comments and answers.


Showing 2 responses by bombaywalla

upsampling is a digital domain process where the bit rate of the digital music stream is changed to something higher by an integer number. In redbook CD the bit rate of the music stream is 44.1KHz.
It appears to me that you are not versed in digital signal processing hence going into too many technical details is not advisable.
But, due to the 44.1KHz sampling rate of redbook CD, the analog filter at the output of the DAC needs to have a very steep filter skirt to attenuate the DAC clock energy. Well, making such a steep skirt analog filter has the major disadvantage that the in-band filter response (i.e. the response of the filter in 20Hz-20KHz audio region) attenuates the high frequency response of the music/program material + it creates a lot phase shift which destroys the integrity of the music signal.
One way around this is to upsample the music signal by an integer number, say, 2 to 88.2KHz. Now, the filter skirt of the analog filter following the DAC can be much less steep. This makes the analog filter easier to design, it adds less phase shift inside the 20Hz-20KHz audio band. So, from an analog filter design perspective this is a win.
What about the digital music bit-stream perspective? The way upsampling is done they zero-stuff. So, imagine your bit-stream was a 11111 at 44.1KHz & we want to upsample to 88.2KHz. then the new zero-stuffed bit stream would look like 1010101010. Now we can clock this new bit stream at 88.2KHz & the logic1s would be encountered at the same time as they were when we clocked the data at 44.1KHz. So, we have increased the bit rate. but we've destroyed the original data - our orig data at 44.1KHz had a string of 5 logic1s; the new 88.2KHz bit stream has alternating logic1 & logic0. what's up with that? You are correct if you asked this question. So, here comes the important part of upsampling, smooth the new higher speed bit-stream, you have to run the 88.2KHz bit stream thru a digital filter that will try to estimate whether the zero that was stuffed should remain a logic0 or be flipped to a logic1 to make the playback music sound right. There are n number of algorithms to do this many of them are quite sophisticated & almost all of them are proprietary. it is this digital filter algorithm (often called an interpolation filter) that imparts its sonic signature onto the final up sampled music signal. You might like the way the manuf does this interpolation or you may not. And, herein lies the crux of the matter i.e. the effect of upsampling onto the higher bit rate music signal. It is also why so many manuf exist & different people like diff manuf & you rarely get a consensus on who upsampling technique is the best. It's a matter of taste. Often you even see people who have upsampling CD players & DACs where they simply turn this feature off with the press of a front panel button because they say it sounds better without upsampling.
Anyway, that's the skinny on upsampling & its effect on the music signal. sorry it had to be so long-winded but it was important to explain a bit of the background.       
Thank you, Almarg. :-)