Quote from: saratoga on 2024-12-06 21:47:08

In the real world sounds are not made of little connected straight lines, so it is not at all a natural way to do resampling.  Probably you mean "easiest to understand".  It is used because it is easy, not because it is good.

They're artifacts resulting from the error introduced by approximation of a curve with lines.  You may personally like that sort of distortion, but as many people have pointed out there are much better ways to accomplish the same thing than trying to find specific ratios of sampling rates that make the distortion less harsh or whatever you're looking for. 

Linear interpolation is not completely natural but it's the most natural interpolator because generating absolutely nothing above a frequency limit is much more artificial than generating filtered reflections (lower frequencies generate weaker and weaker reflections in linear interpolation, that's the filtering).

Linear interpolation is not completely natural but it's the most natural interpolator because ... In my opinion, at least.

btw, do you realize that bad resamplers, despite creating some new high-frequency content, also introduce significant distortion in the lower frequencies, sometimes even as low as 100 Hz or so?

I think I've heard some cheap toys that use those sampling rates like
(5000 hz, 6000 hz, 7000 hz , 9000 hz) instead of 8000 hz

The most natural thing to do is not introduce artifacts.  Then you don't have these strange issues with specific sampling rates making distortion more or less annoying to the ear. 

This is a common misconception among non-experts in this field. The most natural (as in "natural sciences") thing with waveform signals is something involving damped or undamped sinusoidal oscillations (like a good IIR or windowed-sinc resampler), not drawing straight lines between two sample points (like a linear interpolator).

btw, do you realize that bad resamplers, despite creating some new high-frequency content, also introduce significant distortion in the lower frequencies, sometimes even as low as 100 Hz or so?

Quote from: C.R.Helmrich on 2024-12-08 11:38:01

btw, do you realize that bad resamplers, despite creating some new high-frequency content, also introduce significant distortion in the lower frequencies, sometimes even as low as 100 Hz or so?

Out of curiosity: Why/how? Of course code bugs, but ... ?

This can be the case only with the mentioned case (when the output sampling rate is not a multiply of the input sampling rate).

Quote from: Klymins on 2024-12-08 12:24:53

This can be the case only with the mentioned case (when the output sampling rate is not a multiply of the input sampling rate).

Regarding additional frequencies created as a result of insufficiently attenuated aliasing (fold-over from high to low frequencies... answer to Porcus' question), that might be true. However, lowpass-filter-like attenuation of the low-band (the frequency content being upsampled) occurs in any case when using linear interpolation. Attached the average frequency response for an example, where the input signal is simple white noise (green)... which should, ideally, remain spectrally flat in the low-band but is 6dB down at the low-band Nyquist frequency when upsampling via linear interpolation (red). And 3dB down much earlier than that.

I, personally, would not like to have my upsampled waveforms 3dB down at such low frequencies...

The attached picture contains, in the lower right corner, the Matlab/Octave code using which I created this example.

Quote from: saratoga on 2024-12-08 01:16:59

The most natural thing to do is not introduce artifacts.  Then you don't have these strange issues with specific sampling rates making distortion more or less annoying to the ear. 

What causes these issues is the ratio between the sampling rates, not the sampling rates alone.

However, I want to try the pre-emphasis which flattens the frequency response as I wonder what does it sound like, therefore you reminded me to start a topic about that (topic#127058) - thanks.

I still find the frequency response of linear interpolation much more natural than the one of sinc interpolation. However, I want to try the pre-emphasis which flattens the frequency response as I wonder what does it sound like, therefore you reminded me to start a topic about that (topic#127058)

Quote from: Klymins on 2024-12-08 16:26:51

I still find the frequency response of linear interpolation much more natural than the one of sinc interpolation. However, I want to try the pre-emphasis which flattens the frequency response as I wonder what does it sound like, therefore you reminded me to start a topic about that (topic#127058)

Attached an extension of my experiment above (blue, was curious myself). Et voila, you've added a much more natural IIR oscillating filter.

Chris

@Klymins : You've been given lots of explanations, theory and reasoning since you first started posting in this site. And yet in most of your posts you try to tell the others that they are wrong because your preference is different.

I have a clear example of how aliasing sounds and why a bad resampling might feel better even though its quality is worse.
Check these two videos on youtube with the same song rendered one with ZOH and the other with SINC interpolator:
Song with aliasing (ZOH)
Song without aliasing (SINC)

Especially the start of the song has a very pronounced aliasing.
You might feel that the song sounds better WITH aliasing than without, but that does not mean better quality.
(Note: Since this is ZOH instead of linear, the aliasing extends to almost all the frequency range. Linear interpolation would kindly lowpass filter itself like what C.R.Helmrich posted)


Aliasing is an artifact that causes mirroring of the frequencies. There is no control of this mirroring, and it would rarely resemble what existed there before sampling it at a too low sample rate. 
The lack of control means that you can hear aliasing on a bass sound that should not have high pitched sound at all,  or you might hear a hihat that "magically restores" the missing part. 
So in the end it is a gamble if the aliasing will be an improvement or it will annoy and that's why pretending to use aliasing as a kind of sound improvement is a nonsense.

Mechanisms like SBR, used in MP3PRO or HE-AAC fullfill this functionality in a controlled way, and it still can sound bad in certain cases.

When you deny that audible aliases are artifacts, it is not opinion, it is misinformation.

I was trying to say that what is artifact and what isn't changes with the point of view

Quote from: Klymins on 2024-12-12 08:05:25

I was trying to say that what is artifact and what isn't changes with the point of view

And that is misinformation. At least here, when it is about audible artifacts - sure you could have argued that when you are down to way below audibility it wouldn't anymore qualify for the term, but that's not what you are doing: you keep denying that [this particular kind of] resampling artifacts constitute "artifacts" whatsoever. 

In this case, distortions created by a guitar amplifier (real or simulated), which are very widely used in some music genres like rock and metal, would also be artifacts.

Quote from: Klymins on 2024-12-12 13:08:56

In this case, distortions created by a guitar amplifier (real or simulated), which are very widely used in some music genres like rock and metal, would also be artifacts.

Creating music and reproducing a signal are different, you know? Two Wikipedia illustrations of halftone:

No artifacts here?

That's bullshit, but if you insist: it would make it much easier for you to give one categorical answer the following question.
Are the half-tone (the "dots") digitization artifacts or are they not?