[kukrisna]

Hey All,

I did a bunch of searching here, reading various threads, and links they sent me to. Still a bit confused.

1) What is the difference between oversampling and upsampling exactly? I read both wikipedia articles but am still confused due to the specificity of the writing.

2) Am I correct in assuming that they can both be done via software and the external DAC? Or does that just apply to one, and not the other?

3) If it is so beneficial to up/oversample, why are there still many who do not?

4) How would a CD Transport up/oversample? Do some do or is this all handled via the DAC if the DAC has such capabilities?

5) I'm running OSX on a MacBook Pro with digital out to my Presonus Central Station. If I wanted to hear the benefits, am I correct in assuming that I will have to change the output in CoreAudio in OSX from its current position of 44.1kHz/16-bit to something higher?

Thanks!
Keith

 

[Andrew_WOT]

In layman's terms:
- Oversampling is a multiplication of original sample rate to an integer x2, x16...
- Upsampling is a stretching to a specific target - 48, 96, 192...
Afaik oversampling is done at the hardware level only, upsampling in both, software and hardware.

 

[Alleyman]

Upsampling vs. Oversampling for Digital Audio - page 2 — Audioholics Home Theater Reviews and News

A good read. I found it via google after I opened your thread. Although this is page 2 of the article, it focuses on the differences between the two and their advantages/disadvantages.

 

[kukrisna]

thanks for the link and the info fellas - that cleared things up

 

[gregorio]

The article is basically accurate although could give the unintiated the wrong impression. For example the bit about Jitter. In a few very specific cases it is possible that upsampling could improve the perceived quality. The original music would have to have been recorded on poor quality ADCs which were badly clocked in the first place. Even then, it is highly unlikely that upsampling is going to cure or even improve the problem. As a general rule, what is on the CD is the best quality you are going to get. Upsampling cannot improve on this quality but it can reduce the quality. The only reason to upsample is if a cheap DAC has such a poorly implemented reconstruction filter at 44.1kFs/s that it has to run at 96kFs/s because it is cheaper to implement at decent filter at 96k than at 44.1k.

Gregorio

 

[gyrodec]

As it is impossible to create a brick-wall filter at 44.1Khz that will not have significant phase issues in the audible zone, moving it to a higher fs is always a good idea, no matter how well the filter is made. The question is - is the errors created during the upsampling process (it is never perfect, obiously) more or less significant than the issues created by a steep filter near the audible band. That, is down to each DAC builders abilities.

 

[PhaedrusX]

Quote:

Originally Posted by gyrodec /img/forum/go_quote.gif As it is impossible to create a brick-wall filter at 44.1Khz that will not have significant phase issues in the audible zone, moving it to a higher fs is always a good idea, no matter how well the filter is made. The question is - is the errors created during the upsampling process (it is never perfect, obiously) more or less significant than the issues created by a steep filter near the audible band. That, is down to each DAC builders abilities.

what do these significant audible issues sound like?
i've seen the oscilloscope graphs, yet when i listen to my NOS DAC i can't actually discern any anomalies.
i would think that if these phase issues were as perceptible as some would have you believe, NOS designers and listeners would have abandoned the theory behind it a long time ago.

 

[gregorio]

Phaedrus, gyrodec's post is essentially correct, although of course the word significant in the first sentance is open to interpretation. In the early days of digital audio these filters were implemented very poorly because the processing power wasn't available. There were obvious audio artefacts, ring modulation making the audio sound quite harsh was the most obvious. Other phasing issues could include narrowing of the stereo image and loss of low mids. Today these artefacts have been largely (but not entirely) eliminated. These artefacts are unlikely to be noticable unless you have good ears, a high quality monitoring environment and poor quality DACs.

The one area where I will disagree with gyrodec is that it's always better to upsample. Some of the best DACs are as close to perfect as makes no difference, even at 44.1kFs/s. With high quality DACs there are no advantages to upsampling, only disadvantages.

Gregorio

 

[PhaedrusX]

Quote:

Originally Posted by gregorio /img/forum/go_quote.gif Phaedrus, gyrodec's post is essentially correct, although of course the word significant in the first sentance is open to interpretation. In the early days of digital audio these filters were implemented very poorly because the processing power wasn't available. There were obvious audio artefacts, ring modulation making the audio sound quite harsh was the most obvious. Other phasing issues could include narrowing of the stereo image and loss of low mids. Today these artefacts have been largely (but not entirely) eliminated. These artefacts are unlikely to be noticable unless you have good ears, a high quality monitoring environment and poor quality DACs. The one area where I will disagree with gyrodec is that it's always better to upsample. Some of the best DACs are as close to perfect as makes no difference, even at 44.1kFs/s. With high quality DACs there are no advantages to upsampling, only disadvantages. Gregorio

sure, i wasn't arguing against the imperfect theory behind NOS designs, or the flaws that show up in the sine waves. i was just genuinely curious as to how those errors are translated into the auditory realm.

it seems that these issues were more problematic decades ago, when the CD was first introduced as a media format, and are less significant these days.

i just question how "significant" the audible issues are when, as you say, it is largely inaudible nowadays. that, and "harshness" seems to be a characteristic that is absent in most modern NOS designs.

 

[eruditass]

i see advantages as being: allowing dac to work in native format without upsampling being done locally, better timing

disadvantages? a little more processing power if upsampling is an integer multiple.

 

[gregorio]

PhaedrusX - If you think that half the listeners out there are listening to 128kbps MPegs and $5 ear buds, the discussion about upsampling and reconstruction filters is all a bit pointless.

It is a valid point for discussion amongst the audiophile community though, as of course the only thing standing between the mastering engineer and your amp/speakers is your DAC. Just the difference in price between DACs is a good indication that they're not all identical and some are significantly better than others.

 

[12Bass]

Quote:

Originally Posted by gregorio /img/forum/go_quote.gif With high quality DACs there are no advantages to upsampling, only disadvantages.

FWIW, using ASIO with Foobar2000, I've found that 88 kHz or 96 kHz upsampling with Secret Rabbit Code sounds noticeably better than nominal 44 kHz with my modified Echo Gina24 and RME DIGI96/8 PST sound cards (both have LM4562s in the output, Gina has capacitor mods as well). The soundstage seems somewhat broader at 96 kHz, and subtle details seem more apparent, mostly in the highs. The two units use the AK4393 and AD1852 respectively, which I wouldn't class as substandard DACs. My suspicion is that, contrary to what the engineering sheets might say, in reality the low-pass filter artifacts interfere more with lower sampling rates, producing audible phase/frequency response anomalies.

 

[PhaedrusX]

Quote:

Originally Posted by gregorio /img/forum/go_quote.gif PhaedrusX - If you think that half the listeners out there are listening to 128kbps MPegs and $5 ear buds, the discussion about upsampling and reconstruction filters is all a bit pointless. It is a valid point for discussion amongst the audiophile community though, as of course the only thing standing between the mastering engineer and your amp/speakers is your DAC. Just the difference in price between DACs is a good indication that they're not all identical and some are significantly better than others.

right...you've lost me. i have no idea where you're coming from here.

these are pretty basic common sense statements of yours, and i'm wondering where in my post you saw any opposing opinions on any of these issues you've just mentioned.

 

[gyrodec]

I wanted to clear up what people though I said - largely because what I said was so very easy to read that way. I mentioned significant phase errors caused by filters near the audio band, and yes, these phase errors are significant, unfortunately there is little agreement on how audible they are. High-pass filters at the low bass end, usually caused by the limitation of your (and my) speakers do cause very audible effects. They are not dramatic, but once you know the signature, you can repeatably pick it up. (Mostly its a slight muddling of image and tone an octave or two above the cutt-off.) The audibility of low-pass filters at 20KHz is much less certain. There isn't much useful information up there to start with (3Khz is the highest actual note on any instrument), and people hearing also tail off up there. So, there are issues caused by the filters, but it is debateable that they cause actual issues for listeners. Also, I didn't say you should always upsample, I said I would always upsample if the upsampling algorithm caused fewer issues than the 44Khz filter point - and that is an implementation issue not a theoretical one, so it will vary DAC by DAC.

I feel gegorio and I are probably closer on this that readers my feel, at least form an engineering standpoint, we just have different listening experiences and biases.

 

[gregorio]

Gyrodec - I agree. My experience is mainly with pro gear rather than consumer equipment.

It would seem that the trend in good quality consumer DACs is to stick in a crappy reconstruction filter at 44.1 or not to have a 44.1 reconstruction filter at all and then upsample everything so it passes through the 96k filter instead. This isn't the route to high fidelity because you are always going to introduce errors with the upsampling process. The other option is to get hold of 24/96 files in the first place so they don't need upsampling but this means that music files are going to be 4 times as big as they need to be with no benefit to the consumer except that the manufacturers don't have to implement a decent 44.1k filter. I still maintain though that 24bit is completely pointless for the consumer.