Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: In which situations oversampling the input of a DAC is beneficial? (Read 5374 times) previous topic - next topic
0 Members and 2 Guests are viewing this topic.

In which situations oversampling the input of a DAC is beneficial?

Hello hydrogenaudio. I have a question that has been discussed a lot in the past years but, but even after a lot of reading, it's still unclear to me. Sorry for asking once again.

Assumption 1) Sometimes or some DACs will internally (hardware) oversample and filter (what kind of filtering?) its input to match the sample rate that it uses internally for processing.

Then, based on this assumption and on the assumption 2) that its limited hardware used for oversampling is inferior to a virtually unlimited (CPUs running at GHz in PCs) software-oversampling, one could infer that you should, whenever is possible, oversample the input to match the sample rate that the DACs uses internally (and do manufacturers provide this information?), thus bypassing the DACs inferior oversampling/filtering that would otherwise result in more losses/distortions to the original signal.

(all of that would theoretically not apply to NOS DACs)

Whether or not the reasoning above will yield in perceived higher sound quality is a whole different question. I just wanted to know if the reasoning above is actually what happens and, if so, in what situations/context. And, if so, why not always software-oversample/filter? (provided that the CPU will not get overly taxed).

Thanks!

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #1 –
It is more useful to view the oversampling filter, and the following DAC, in conjunction, as the actual converter. The performance of the entire arrangement can be assessed and compared with other such arrangements. The oversampling becomes an implementation detail.

The perceived quality is an entirely different matter. Either we are talking about objective quality, or about subjective opinions.

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #2 –
Quote from: F u r u y á on
Assumption 1) Sometimes or some DACs will internally (hardware) oversample and filter (what kind of filtering?) its input to match the sample rate that it uses internally for processing.
A proper DAC will do reconstruction ... with a reconstruction filter. In the most trivial case that's upsampling followed by a lowpass filter. (A real DAC will have more stages including an analog one.)

Quote from: F u r u y á on
Then, based on this assumption and on the assumption 2) that its limited hardware used for oversampling is inferior to a virtually unlimited (CPUs running at GHz in PCs) software-oversampling
No, you are comparing apples and oranges here.
One is a hardware block programmed to do one job while the other is a processor with a general instruction set that has to run the pieces of software you throw at it.

Quote from: F u r u y á on
one could infer that you should, whenever is possible, oversample the input to match the sample rate that the DACs uses internally (and do manufacturers provide this information?), thus bypassing the DACs inferior oversampling/filtering that would otherwise result in more losses/distortions to the original signal.
You usually can't even do that since the DAC won't accept data at the rate that it internally processes it. You can still do resampling to higher sampling rates and then the DAC's filters will operate at higher frequencies as well - above your resampler's filters.

Quote from: F u r u y á on
(all of that would theoretically not apply to NOS DACs)
It does, but such DACs simply lack the digital reconstruction part.

Quote from: F u r u y á on
And, if so, why not always software-oversample/filter? (provided that the CPU will not get overly taxed).
Some audiophiles are obsessed with bit-perfectness. Another reason might be that your operating system could also resample, Another negligible point is that DACs can actually perform worse at higher sampling rates...

So yeah, you can pretty much always resample.
"I hear it when I see it."

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #3 –
Quote from: F u r u y á on
assumption 2) that its limited hardware used for oversampling is inferior to a virtually unlimited (CPUs running at GHz in PCs) software-oversampling
Assumption 2 isn't always true.  It's probably true for DACs that use a single chip DAC internally, but there have almost always been DACs that dedicate more hardware to the oversampling and that hardware can use "better filters" (or more accurate processing of it's filter) than most software oversamplers running on a general purpose CPU.  I put "better filters" in quotes because there's no perfect filter and hence there's room for differing opinions about what kind of filter or which parameters of a filter provide the best results.

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #4 –
And regardless of which chip has more power, what really matters is that the DAC has enough power to do the job required.

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #5 –
It's pretty common for DACs to run at a constant internal frequency such that doubling the sampling rate you feed them halves the oversampling ratio such that nothing changes.  Which makes sense, if you design something to run optimally at 5 MHz, it would be dumb to run it any any other frequency.

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #6 –
Thanks guys and gals.

Please correct me if I'm wrong: basically, it doesn't matter whether you oversample or not since an oversampling DAC will always oversample. (?) (even if you input the max sample-rate it supports).

And, most importantly, analyzing the DAC as a whole (the box, not the chip) is often more useful than analyzing the individual parts inside the "black box". (?)


What about the difference in oversampling techniques (in-DAC versus out-of-DAC)? Is there any situation/context where there is a benefit with oversampling out-of-DAC? (excluding situation below)

After reading a bit (or should I say, a lot :D) I found out that NOS DACs are the ones that truly benefit from oversampling (less early roll-off in the audible band). Is that correct? (Archimago's TEAC review - PCM 24/384 section)

Some manufacturers are including a NOS mode ("Bit perfect" mode, iFi; "Filter: OFF" mode, TEAC). I wanted to ask what's the pros/cons of oversampled input + NOS DAC versus non-oversampled input + oversampling DAC, but I'm afraid this should be asked in a different thread?

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #7 –
The NOS DAC provides a marketing buzzword and allows the manufacturer to charge a price premium.  It may also give you the dubious pleasure of trying to hear differences between various upsampling algorithms, if you're really, really bored.

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #8 –
Quote from: F u r u y á on
Some manufacturers are including a NOS mode ("Bit perfect" mode, iFi; "Filter: OFF" mode, TEAC). I wanted to ask what's the pros/cons of oversampled input + NOS DAC versus non-oversampled input + oversampling DAC, but I'm afraid this should be asked in a different thread?

Oversampling does two things:  decrease quantization noise and decrease rolloff/imaging.

Oversampling ratios on most DACs are very high.  Perhaps 64, 128... even 1024x, whereas if you are oversampling you are probably talking about 2 or 4x.  If you take an old DAC like that 25 year old one discussed in the review you linked and oversample 2 or 4x, you can correct for the roll off from the anti-imaging filter on the DAC (at least assuming the filter is configurable), but you won't get the reduction in quantization noise because the oversampling ratio is very low. 

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #9 –
Quote from: F u r u y á on
Please correct me if I'm wrong: basically, it doesn't matter whether you oversample or not since an oversampling DAC will always oversample. (?) (even if you input the max sample-rate it supports).
That's a strangely worded question. :-)

Most contemporary DAC chips for Audio are Sigma-Delta-Converters. That principle is based on a relatively high oversampling factor, and additionally uses noise shaping. Such designs will always oversample, regardless what input sample rate you select, as oversampling is pivotal for the way they work. If you feed them an externally oversampled signal, they usually reduce their internal oversampling factor by a corresponding amount, so the overall oversampling factor ends up being the same. That doesn't mean the end result is exactly the same, though. It depends on whether the characteristics of the external oversampling filter match those of the oversampling filter in the DAC chip, but I wouldn't expect significant differences, as long as the filters are competently designed.

If the DAC uses a different principle, for example old DACs from the early years of digital audio, the oversampling factors are much lower, and the oversampling filter is usually implemented as a separate chip. Oversampling factors of up to 16x were used before the sigma delta principle took over the market. Here, it is conceivable that if you oversample externally, you end up with no oversampling in the DAC. Today that would be an exotic design.

Quote
And, most importantly, analyzing the DAC as a whole (the box, not the chip) is often more useful than analyzing the individual parts inside the "black box". (?)
That is certainly true. Analyzing the individual parts in the overall converter is a matter for specialists. There's nothing a normal user can usefully contribute, or gain.

Quote
What about the difference in oversampling techniques (in-DAC versus out-of-DAC)? Is there any situation/context where there is a benefit with oversampling out-of-DAC? (excluding situation below)
In theory, there could be such a situation when the in-DAC filtering is deficient in some respect. However, in practice the user is not in a position to determine whether that's the case, and what to do about it. If a converter doesn't do a good job by itself, the most reasonable reaction would be to replace it.

Quote
After reading a bit (or should I say, a lot :D) I found out that NOS DACs are the ones that truly benefit from oversampling (less early roll-off in the audible band). Is that correct? (Archimago's TEAC review - PCM 24/384 section)
This is true, but from a converter design perspective, and not from a user perspective. If the NOS DAC has a correct analog reconstruction filter, then you would have to change this filter in order to reap any benefit from the usage of external oversampling. If it doesn't have an analog reconstruction filter, it was misdesigned to start with. Both situations rather call for a replacement of the DAC, not for some kind of patching it up using external oversampling. You've got to decide whether you want to be a user or a DAC designer.

Quote
Some manufacturers are including a NOS mode ("Bit perfect" mode, iFi; "Filter: OFF" mode, TEAC). I wanted to ask what's the pros/cons of oversampled input + NOS DAC versus non-oversampled input + oversampling DAC, but I'm afraid this should be asked in a different thread?
Those manufacturers address an audiophile fad that runs counter to sampling theory. There's no reason to take that seriously in any way. As a user, feed the signal you want to listen to to the DAC in its original form. If the DAC is worth keeping, it will do a good job converting it to analog. If not, don't try to patch it up by trying to feed a "tuned" signal to it, it would be nothing more than random tinkering.

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #10 –
Quote from: pelmazo on
[
Those manufacturers address an audiophile fad that runs counter to sampling theory. There's no reason to take that seriously in any way. As a user, feed the signal you want to listen to to the DAC in its original form. If the DAC is worth keeping, it will do a good job converting it to analog. If not, don't try to patch it up by trying to feed a "tuned" signal to it, it would be nothing more than random tinkering.

Good opinion, I agree. DAC should do its job internally as its designer wanted/specified, and it should be feeded with content at one of the standard rates (44.1/48/88.2/96 kHz) as it is recorded/produced/stored. Seems to be a reasonable approach to those endless talks. And since 44.1 kHz is the CD standard enough (or very very close to enough) for complete Hi-Fi experience, good DAC should work with this content input without any hassle.

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #11 –
Thanks once again everyone for replying.

Special thanks to pelmazo; your answer gave me a lot of peace of mind.


And small digression of mine:
Quote
Technology cannot exist without engineering and hard science.

Audiophiles cannot exist without technology.

Yet, sometimes (or often) they choose to deny formal knowledge (science and engineering). Go figure.*

*They should be sentenced/punished/educated every time they deny sampling theorem >:(
*That applies to other contexts. For example, some people want to protest against modern "artificial"/"synthetic"/"no life"/"digital" technology. And in order to do that, they use their iPhone to post in Facebook 8)

Re: In which situations oversampling the input of a DAC is beneficial?

Reply #12 –
Quote from: F u r u y á on
And, most importantly, analyzing the DAC as a whole (the box, not the chip) is often more useful than analyzing the individual parts inside the "black box". (?)
Analyzing the parts is either done by professionals/designers to check if they meet design requirements or by audiophiles to ascribe magic properties to different parts, topologies ... and fallaciously generalizing those parts' qualities and performance onto the whole device. :P

At the end of the day you can use the most expensive and exquisite components and still build a crap device. Of course the easiest way to justify such a purchase would be the biased, anti-listening-test and -measurements i.e. audiophile approach. ;)
"I hear it when I see it."