The Well-Tempered Computer, an introduction to computer audio

Monday, January 24, 2011

Oversampling vs Upsampling

A recent thread on ComputerAudiophile.com has sparked a heated debate on Upsampling/Oversampling, whether it is a marketing hype or algorithm with true sonic benefits. 


I've came across this article by Pacific Valve & Electric Company's Vic Trola, a good read. 
Note: Pacific Valve imports and distributes Lite Audio, Audio GD, among others.
"Monday, June 12, 2006
What’s with Oversampling, Upsampling and Non-Oversampling (NOS) 
When we were vinyl, we had our own vocabulary: belt drive, direct drive, servo, anti-skate, counter weight, tracking angle – remember? Now that we are optical, a whole set of new words spook our vocabulary: Upsampling, Oversampling and Non Oversampling. Needless to say, these words are creating confusion out there in the press and among some audio jockeys (not you, of course) and they are used interchangeably, which is a big no-no.
Let’s first start out with some basics: oversampling is the realm of the conversion between digital to analogue and the upsampling is the realm of digital. Non oversampling is an enigma wrapped in a mystery. You cannot use these words interchangeably because they all mean different things.
You want a DAC to correct the anomalies in the conversion process. Take some of the earliest DACS, that were 16 bit and extracted data at a rate of 44.1 kHz / 16 bits. When the DAC converts data in the 20-20kHz range, it leaves some canard above the 22kHz range that must be cleaned up. If it did not, inter modulation distortion and ultrasonic frequencies would wreak havoc on your system – destroying tweeters, midrange drivers and turning mylar into chewing gum. In order to prevent this, a filter is applied. The filter that filters this out is called a “brick wall”, and it abruptly chops off frequencies at 20kHz. This filter, while not allowing the digital leftovers to blow your tweeter to smithereens, does cause audible distortions. The abrupt filter is responsible for some of the sonics coming from early vinyl copies of Fleetwod Mac’s Tusk that were digitally recorded and mastered or some circa 1980 digital recordings(Ry Cooder) . Cymbals sound like tambourines, highs are compressed as well as dynamics. If digital was going to make it into high end, something had to be done. Enter Phillips with its 4X oversampling player.
In a 4X oversampler, the digital signal is fed through a special digital converter that samples it four times. When a CD is sampled four times the sample rate becomes 176.4 kHz(44.1 X 4). When the filter is applied here, there is less of chance that the filtering roll off will make its way into the audible range. A filter applied at 176.4 kHz sweetens the music as it removes most of the grunge. As cost allows, oversampling pays off in a big way. The DAC 60 and DAC 38 use Burr Brown’s latest PCM1704UK chip which is 8X oversampling at 96kHz. The DAC 68 uses Analogue Devices AD1835 chip which samples as high as 192kHz. The AD1853 is fully compatible with all known DVD formats and supports 48 kHz, 96 kHz and 192 kHz sample rates with up to 24 bits word lengths. The higher bit lengths of these DACs allow them to act more in a linear fashion during the conversion process and results in superior sound quality.
Now I digress, to the DAC AH that is a non oversampling DAC using 8 TDA1543 chips with no digital filtering. From our discussion above one would assume that the grunge left over should have blown up tweeters and mylar alike. But, that’s not what happens. So what gives?
This has to do with the 8 TDA1543 that the DAC AH employs and the algorithm used for designing non oversampling DACs. The non-OS algorithm by definition introduces a slight roll off in the treble, so by the time you get up to damaging frequencies, the output of frequencies above 22kHz isn't really all that high. Hence, intermodulation distortion still makes its way into the DAC AH and other non oversampling DACs. This may explain why some non oversampling DACs sound compressed at the higher frequencies. The DAC AH modded, less so, only because the superior OP AMP compliments the conversion process in that it is able to open the sound a bit. Some NOS DACs add a digital filter – but that really defeats the whole purpose of the NOS DAC anyway – its kind of like adding MSG to a low salt diet.
So what is “upsampling”? First let me take away some of the confusion. There is belief in many an audiophiles’ minds about the power of upsampling or oversampling to create data. It is not possible to create more accurate information than is already in the digitally sampled signal. Think of blowing up a digital picture: if the data is not there, blowing it up does not increase detail.
Upsampling differs from oversampling in that upsampling is between the transport and the DAC. In the case of the Lite DAC 39, it sits between your transport and a DAC – like the DAC 60 or DAC 38. In this case the DAC 39 feeds the DAC 60 a 96 Khz signal instead of the 44.1 kHz. In theory, when a superior upsampler is mated with a DAC that can handle the upsampling, the true 96kHz 20 bit processing can be achieved. The resulting filtering process operates in a linear fashion which yields better sound quality. So, it has to be a careful match. An upsampler should only be used with a DAC capable of handling the high sample frequency."