[Hantra]

All:

After our DAC shootout, I started wondering about my view on non-oversampling, or filtering.  The DAC's that had high oversampling rates, and upconversion sounded very consistent.  There was a consistency of energy in the higher frequencies that perhaps is not there with my non-oversampling DAC.  

This makes me wonder what the filtering DAC's are actually adding.  I mean, we all know that DAC's with oversampling essentially "make-up" information to fill in the blanks of digital recordings.  But is that really real?  I mean, yesterday, all the DAC's I heard sounded good.  I suppose I am biased still to my DAC even though the Bel Canto did some things I like better.  

The reason I would choose mine over the BC is because I have lived with mine for a while, and have gotten used to the absence of some smearing that I notice with all other digital that I have heard.  This smearing is not that noticeable if you haven't really listened extensively without it.  Once you go to a DAC that doesn't do it, then it's hard to listen to digital otherwise.  If the BC DAC had not had the digital smearing, I would choose that DAC as the best, even though it didn't provide the quietest stage for the music to live within.

There is something about a musical note eminating from zero as opposed to somewhere higher on the amplitude scale.  It just changes things, and begins to sound more like music.  There is still the natural decay, but there is dead and complete silence between notes as well.  This too is hard to notice unless you have listened to something for a while that does that.  

It's sort of like Wayne once said about the Bybees.  He said "it gets rid of noise that you weren't previously aware of".  That's sort of what I experience with this DAC I have.  Now, the question is harder to answer when you ask which one is correct.  As I have only heard one other non-oversampling DAC, I can't really give an educated comment on this.  I have heard the 47 Labs Shigaraki, and although it was okay, I didn't like the system it was in.  It was setup on some Triangle speakers which I do not like, and so I can't tell you anything about the DAC.

For those who have, or have auditioned the non-oversampling DAC's, can you provide insight?  Do they sound a bit rolled off at the high frequencies?  If so, is that because of the lack of digital noise, making the HF's much clearer, or is that because they really are rolled off?  

These are questions that hopefully we can get more insight into than just my own.  I am not as well versed in this technology as I know many of you are.  I am also not sure that I am THAT interested in the technology as I am the sound.

Opinions?  Facts?

B

[WilliamL]

Hantra, et. al.

"Upsampling does not add any data that wasn’t present on the original recording but interpolates intermediate data points"

Yes, we know that it doesn’t result in the creation of new info, yet IMHO it struggles to realistically DO this "interpolation", albiet artificially. I personally prefer non-upsampling to upsampling/oversampling.

I would check out this white paper from Sim Audio. Its a darn good place to start for basic info.

http://www.simaudio.com/pdf/Upsampling.pdf

(BTW. We don't sell or promote Sim Audio.)

Cheers,
Bill

[Hantra]

Good article Bill.

Thanks. . .

So far, the only real determination I can do is to take an LP, and the same CD, and listen to the cymbals.  I don't hear any extra "air" around the cymbals as opposed to the vinyl.  In fact, they sound almost exactly like the CD.  So, all I can assume is that the extra "air" we hear is often times digital smear.  

Not to say it's not pleasing to some.  But when you have lived without it, it's hard to go back.

So, no one else here has a DAC that is non-oversampling?

B

[eico1]

the best excuse for oversampling is to keep jitter under control in stand alone dacs. Many non-up/over dacs will use low order analog filters that probably roll-off pretty well into the audio range, so some audible differences may be due to that fact too.

steve

[Hantra]

the best excuse for oversampling is to keep jitter under control in stand alone dacs.

While I will admit that I don't know that much about the subject, one thing I am sure of is the jitter issue.

Adding digital filtering dramatically reduces the time window you have for tolerable jitter.  It takes a LOT of time to do digital filtering.  I don't remember the exact numbers, but someone in the know may chime in here.  I think that in a non-filtering DAC, the acceptible window for jitter is 175 picoseconds.  In one with a filter, that dramatically reduces to something like 40 picoseconds.

B

[eico1]

Handra, what window are you referring to? Most src's have a very high input jitter attenuation.

steve

[Hantra]

Steve:

I won't pretend that I know what I am talking about, but I got that information from an article by Ryohei Kusunoki.  

"There are two axes on digitizing the sound. The time axis and the amplitude axis. In case of CD, they are 44.1kHz and 16bit. In other words, we have to press in the amplitude data into one of the 16bit stage at every 22.7 s. That produces maximum of +0.5 LSB error, and the digital audio starts by accepting this error at the beginning.

However, this error only concerns the amplitude axis and no amount of error was admitted on the time axis. Let me suppose that the accuracy of 16bit means how accurately the acoustic energy (time x amplitude) is transmitted by being distributed into each steps of 16bit. Then, by making the amplitude data more accurate, we can distribute the error onto the time axis.

     If we distribute 1/2 of the error,
1 ÷ 44.1kHz ÷ 216; ÷ 2 = 173 (ps)
     
This represents the maximum limit of the acceptable error (maximum limit of the jitter). (diagram 1)



All of the above is based on the basic sampling rate. When in 8 x oversampling and 20bit, that number would be 1.35ps (diagram 2). This is a totally impossible number to achieve for a separate type DAC which has to recover the clock by PLL. This means that under an average jitter environment, the oversampling can not operate theoretically, and lowers the accuracy within the operating field. In short, just by oversampling the original data, 16bit accuracy can not be satisfied anymore.

[eico1]

Looks like 47 labs babble, be carfull of you source for technology! At least to get the whole picture.

Anyway, he seems to be talking about the jitter of the DAC chip's clock, not the jitter on the aes digital signal which is where src can help. No one says you have to run the src to a high rate, you can actually use the chips at say 44.1 input and 48 out if the designer likes. This ratio is selectable on pro units and would be a good feature to tweek in different filters for the audiophile.

steve

[Hantra]

Well, there doesn't seem to be another side to the argument, b/c no one has really made a case refuting the non-oversampling argument of late.  

The only case is history, and that's like saying that we do it that way "because we've always done it that way".

B

[JohnR]

Erm... I'm not entirely sure what the arguments against over-sampling are, other than the claim that not doing it sounds better. I'd have to agree with steve about the quote you made. What that fellow seems to be missing is that when you over-sample, you don't need the same number of bits. That's how the "one-bit" convertors work, for instance. For that matter, digital power amps are also "one-bit" convertors of a sort.

A non-oversampling DAC unavoidably will have either 1. A steep filter with phase shifts and some attenuation into the below-20k region; or 2. artifacts present in the analog output because of aliasing. There is another factor too, which is that over-sampling DACs do the filtering just outside the audio band digitally, and only use an analog filter to remove the very high-frequency artifacts. Perhaps the noise spectra introduced by the upsampling *and* filtering is less pleasing in some ways than the errors of non-oversampled conversion. One could speculate that when these devices start using longer wordlengths for their internal arithmetic that this would change the subjective performance. Who knows

Incidentally, the "history" argument doesn't hold water, since we used to do non-oversamping before we did oversampling...

JohnR

[Hantra]

Perhaps the noise spectra introduced by the upsampling *and* filtering is less pleasing in some ways than the errors of non-oversampled conversion.

What exactly are the "errors" of non-oversampling?  Just curious b/c I really don't know. . .

B

[JohnR]

The ones I just posted above.