Upsampling -- DcS?

I have been seeing this product recently -- the DCS Vivaldi Upsampler, I believe? Now to my understanding, I thought keeping audio at its native sampling rate and bit rate was for the best quality (aka bit perfect)?

I see all this talk about these upsamplers providing the best redbook experience. Anyone want to fill me in?


yes, it is true that in a non-overampling (NOS) DAC the music signal integrity is preserved as compared to in an up/oversampling DAC where an interpolation filter massages the higher bit rate data stream to smooth it out.

A well implemented NOS DAC can provide really very good sonics (I have one & really enjoy playback thru it). But, the NOS DAC does have a potentially serious disadvantage if you feed it a data bit stream at 44.1KHz. What NOS DAC designers do is that they put the analog filter (also called a reconstruction filter) at a much higher frequency that what theory calls for. For eg, my NOS DAC has the reconstruction filter set at 55-60KHz (when theory called for it to be at 22.05KHz). The reason they do it is that they know (& we know) that putting a very high order reconst filter with a 3dB bandwidth at 22.05KHz is going to create some major in-band attenuation & consequently some major in-band phase shift. This will destroy the high freq content of the playback (you lose the ’air’ in the music & very often the ambience of the recording venue). By putting the reconst filter at a much higher freq they avoid these pitfalls BUT the image of the 20Hz-20KHz signal present in the 22.05K-44.1KHz region that needs to be heavily attenuate is rather mildly attenuated by having the reconst filter at 55-60KHz. They cannot put this reconst filter too much higher as you would not get even this mild attenuation. So, you end up with a fair bit of ultra-sonic content. Luckily it’s above human hearing range but what it does is makes the music playback harsh/brittle & gives it that "digititis" that CD playback is so infamous for.

To alleviate this issue what many people do is to upsample the digital bit stream in your computer (because it has much more CPU power compared to a DAC or a portable device) to 88.2K/96K/176.4K/192KHz & then feed it to the NOS DAC. Of course your NOS DAC has to be capable of accepting a high bit rate data stream. Remember, it’s not upsampling the incoming data stream (as it’s a NOS DAC); it’s merely accepting a higher bit rate stream. Now what happens is that the higher bit rate signal has no music signal at the fold-over freq & the reconst filter attenuates the clock frequency harmonics. So, using an example: if you upsample the 44.1KHz data to 96KHz in your computer, you have your music signal in the 0Hz-20KHz region (let’s use 0Hz just to make it easier to explain), you have no useful information in the 20KHz-48KHz, you again have no useful info in the 48KHz-76KHz region & then, finally, you have the image of your music signal in the 76KHz-96KHz region. You want the reconst filter to attenuate everything above 20KHz but with the reconst filter set at 55-60KHz this is not possible. But, guess what? There is no useful information in 20KHz - 76KHz region so if the reconst filter slacks off a bit you do not get any ultra-sonic content (like you used to when you fed the data in at 44.1KHz). So, much less harm done to the listener when upsampling the incoming data stream. Since the reconst filter will start filtering everything higher than 55K-60KHz, your clock fundamental freq at 96KHz will filtered with some appreciable attenuation & then the clock harmonics at 192KHz, 288KHz, 384KHz will get filtered with a great deal of attenuation. So, now all of a sudden your music should start sounding much better & that harshness/brittleness in the high freq should be reduced a lot/gone. If you followed the above line of reasoning then you will definitely say that upsampling to an even higher clock rate (like 192K or 384KHz) would be even better. And, you are correct! I see a lot of NOS DAC owners upsample to 192KHz & 384KHz before feeding into their NOS DAC. For the 192KHz case, for eg, you don’t even see the clock fundamental till 192Khz - far away from the music signal of 20Hz-20KHz & there is no useful info in the 20KHz - 172KHz region. With the reconst filter set at 55K-60KHz, you get some really nice attenuation of any spurious signal hanging around in the 60Khz & above region.

As far as up/oversampling DACs are concerned, they are many very nice implementations that many listeners like. Several well-known brand names in the market. The crux of the up/oversampling DACs is that each manuf has a different algorithm to smooth out the up/oversampled data stream. And, you are going to like some of these implementations & you are going to dislike an equal number - it’s a matter of taste. You will have to hunt them down using friends’ opinion, conversations with manuf. listening tests, magazine reviews, etc. Up/oversampling DACs can provide a really very good playback experience (I happen to really like the Wadia 861SE implementation of this). In my experience I’ve found that up/oversampling DACs made with R-2R ladder DACs sound the very best. Others disagree & seem to like up/oversampling DACs using delta-sigma DAC implementation. The dCS implementation of an upsampling DAC uses a ring converter that is the best in the market for randomizing the noise hence the S/N ratio is really superb & the noise is much more uncorrelated to the incoming signal. It's a superb piece of electrical engineering But do you like their upsampling algorithm sonically? And, if you do, do you have really deep pockets?

Like I wrote, there is no consensus amongst us audiophiles re. up/oversampling DACs. You have to try several to see which one you like.....

Pretty well written post just above mine. I have a dCS Vivaldi up-sampler. All you have to do is listen with/with out it. It will only take a minute to hear the benefits it brings to a dCS system...

I think bit perfect is likely true, just doesn't work out given playback and recording processes that are actually taking place...