What is the point of upsampling?

[Mnyb]

What is this tread about again

Before i’ll unwatch ..

I thought the topic was the use/need of external upsampling to feed a DAC.?

* the tru and tested use of upsampling inside DAC chips ( since decades ) usually called oversampling is not in question ,it’s a practical solution to an engineering problem.

* feediing a DSP system , which one ? this diverges a lot and depend so much on how that is implemented ?

 

[solderdude]

Isn’t the whole point of upsampling to “help” the filter?

The upsampling is needed to move the mirror band higher up so analog post filtering so removal of the 'steps' and/or noise shaping noise is much easier and can be done without phase shifts in the audible band.

 

[Julf]

Something called Redbook (not Mao’s one)

And the evidence...?

 

[Julf]

As I recall albeit hazily from a couple of courses I took in college decades ago, discretization error, the difference between the analog value of a sample and the digitized sample, is white noise spread evenly over the entire spectrum up to Nyquist. If you upsample, you spread this noise over a wider band. Much of it moves outside the audible range and can be filtered out. You thus improve the SNR, probably not by much if your DAC has 32 bits, but more than zero. In the competition to have the best measuring DAC, this can make a difference.

Isn't that only valid for the initial analog-to-digital conversion, not subsequent upsampling?

 

[witwald]

Have you not understood the Shannon-Nyquist sampling theorem?

The Shannon-Nyquist sampling theorem is but one part of the reconstruction puzzle. Different analog filters (e.g., Butterworth, elliptic, Linkwitz-Riley, Bessel, etc., as well as filters of different orders within a given filter class) have different amounts of phase shift in their passband. Doesn't upsampling permit the use of an analog filter that produces less phase shift in the passband?

 

[witwald]

For playback purposes, we have yet to see any evidence of anything regarding an inherently better audible SQ from upsampling.
Lots of claims, no evidence, and it would be hard to see how it could.

I understand your point. But, does not upsampling allow for the use of an analog reconstruction filter that is measurably lower in regard to the phase shift that it introduces? Whether the differences are audible or not could be regarded as being somewhat of a moot point.

 

[Julf]

The Shannon-Nyquist sampling theorem is but one part of the reconstruction puzzle. Different analog filters (e.g., Butterworth, elliptic, Linkwitz-Riley, Bessel, etc., as well as filters of different orders within a given filter class) have different amounts of phase shift in their passband. Doesn't upsampling permit the use of an analog filter that produces less phase shift in the passband?

What is the issue with phase shift? It won't be audible.

Yes, as many have pointed out here, DACs do upsampling internally so they can apply digital filtering, requiring less stringent analog filters - but just feeding the DAC upsampled material won't bring any benefit.

 

[Sokel]

May be easier to tell if you zoom in on the transition band frequencies and place a marker at 22050hz.

Super sloppy but visible.

EDIT:Markers are at 20.000Hz and 22.050Hz

Files are also exported and available to everyone as soon as I get an app to share,it's too big for uploading here (60 M) here:

Khadas Filters.rar

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

Edit:added download link

 

[Blumlein 88]

Super sloppy but visible.

EDIT:Markers are at 20.000Hz and 22.050Hz

View attachment 227146

Files are also exported and available to everyone as soon as I get an app to share,it's too big for uploading here (60 MB)

Wetransfer.com is maybe one of the better free ones for up to 2 gb. Google drive will also work for files like you have. Dropbox.com also still has a free account for up to 2gb. Dropbox is one I use. I've used Google drive, but not as much as Dropbox.

 

[Sokel]

Wetransfer.com is maybe one of the better free ones for up to 2 gb. Google drive will also work for files like you have. Dropbox.com also still has a free account for up to 2gb. Dropbox is one I use. I've used Google drive, but not as much as Dropbox.

There's the files too:

Khadas Filters.rar

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

(I officially uploaded the first Multitone files? )

 

[witwald]

What is the issue with phase shift? It won't be audible.

The phase shift can potentially change the shape of the waveform. This may be important in some applications. As I mentioned previously, whether the differences are audible or not could be regarded as being somewhat of a moot point.

Yes, as many have pointed out here, DACs do upsampling internally so they can apply digital filtering, ...

Sure...but what are those DACs digital filtering?

 

[Blumlein 88]

There's the files too:

Khadas Filters.rar

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

Got it. I use linux so RAR files are fine, but zip might be better for universal access.

 

[Sokel]

Got it. I use linux so RAR files are fine, but zip might be better for universal access.

Noted!Next time...

 

[Julf]

The phase shift can potentially change the shape of the waveform. This may be important in some applications. As I mentioned previously, whether the differences are audible or not could be regarded as being somewhat of a moot point.

The shape of the waveform only matters to your eyes, not your ears.

Sure...but what are those DACs digital filtering?

Not sure I understand your question.

 

[Sokel]

(for the friends who ask)
The default filter (Apodiz Fast Roll-Off) is the dark blue-purple one.

 

[tuga]

The upsampling is needed to move the mirror band higher up so analog post filtering so removal of the 'steps' and/or noise shaping noise is much easier and can be done without phase shifts in the audible band.

So there is a benefit in / requirement for upsampling.

I don’t understand why it is being questioned all of a sudden?

 

[witwald]

The shape of the waveform only matters to your eyes, not your ears.

But, for example, if I need to obtain the values of the peaks in the waveform, the shape matters very much, doesn't it? There are other use cases than just listening.

Not sure I understand your question.

Stepping back a fraction, you mentioned that "DACs do upsampling internally so they can apply digital filtering". In a sense, I did not fully understand your statement, as the cause–effect relationship was somewhat unclear. Would it not be more correct to say that the DACs do upsampling (e.g., by adding zeros into the digital signal data stream), and then they perform digital filtering on that new data stream to remove the unwanted signal content that is contained in the new data sequence, essentially resulting in an upsampled interpolated digital version of the original waveform. The DACs do digital filtering as part of the total upsampling interpolation process, which is not what I took away from your description.

 

[witwald]

I don’t understand why it is being questioned all of a sudden?

The OP asked a question along the lines of: "I just can't fathom what upsampling is supposed to do? Like from 44.1 to 192 gets me what exactly in the end?" One answer that he got was: "There is no value in upsampling." That was a little too short to explain the situation to the OP, and hence this situation has resulted in this longish thread.

 

[pkane]

Got it. I use linux so RAR files are fine, but zip might be better for universal access.

Should probably take this to Multitone thread. But there’s no great need to compress the mth file, it’s already compressed.

 

[tuga]

The OP asked a question along the lines of: "I just can't fathom what upsampling is supposed to do? Like from 44.1 to 192 gets me what exactly in the end?" One answer that he got was: "There is no value in upsampling." That was a little too short to explain the situation to the OP, and hence this situation has resulted in this longish thread.

External upsampling can potentially improve on internal chip-upsampling, particularly when you bypass all internal filtering and modulation.

Many manufacturers are now doing away with off-the-shelf SRC and even D/A conversion and the measured performance they produce is putting them at the top of the scale.

Unfortunately not many bench tests are available of internal vs. external upsampling of the same DAC