How to use Sample format / Sample rate converter (Doubt)

[Cultura-Difusa]

Hello there!

Firstly let me thank the Enzo for his work. Thank you.

Now my question:

Do the signal processing plugins (Sample format and Sample converter) get applied when enabled to all conversions indistinctly, or only to input files with different sample rate/format to the desired output.

That is it.

Thanks in advance!

[enzo1982]

fre:ac always changes the sample format to 32 bit float when the sample rate converter is active, even if the source sample rate already matches the target. I have plans to change that in a future update, though.

The sample format converter on the other hand does not change anything when the source format already matches the target format.

[Cultura-Difusa]

Sorry to bother you.

You mention:

---

[...] always changes the sample format to 32 bit float when the sample rate converter is active, [...]

and

[...] sample format converter on the other hand does not change anything when the source format already matches the [...]

---

---

1. Correct me if I am wrong; but do you mean that?:

If sampling rate of the source is different from the output (with sampling rate plugin enabled), then, gets converted. And that being all things equal, it does not.-

and (that)

Regardless of the input sample format, if the plugin is enabled (sample format), the output is allways changed.

---

2. And yet, if the previous is correct, I still have a question:

Does not the "sample resolution dial" (8-32) affect the output?

I am asking these questions, because I have some lossless files from people who think more than 44 khz and 16bit are "necessary", and I want to resample those files to sane rates/formats (16/44) doing the less harm possible.

Thanks in advance!

[Lee-Carre]

@Cultura-Difusa

I have some lossless files from people who think more than 44 khz and 16bit are “necessary”, and I want to resample those files to sane rates/formats (16/44) doing the le[ast] harm possible.

Now that you've stated your actual goal, and relevant context, that helps a lot.

You want the sample rate converter filter, not the sample format converter. They do different things.

1. Correct me if I am wrong; but do you mean that?:

If sampling rate of the source is different from the output (with sampling rate plugin enabled), then, gets converted. And that being all things equal, it does not.-

[…]

Based on my reading of Enzo1982's reply, above, this is incorrect. The sample rate converter always converts, even when source & target already match. Converting to 32-bit float (and back) will likely result in rounding errors, and the converted file won't match the original file.

So, instead, you'll want to determine which files have sample rates/resolutions which are greater than you'd like (perhaps use FFprobe, part of FFmpeg), and only have the sample rate converter filter enabled for those files (the ones from friends). The rest of the time, you want both the sample filters disabled.

Might I suggest setting up different configuration profiles; one for non-conversion, and one for conversion, so that switching between them is easy (and it makes it more obvious if you've got the converter active, in case you don't want it enabled).

[…]

Regardless of the input sample format, if the plugin is enabled (sample format), the output is allways changed.

Again, based on my reading of Enzo1982's reply, this is incorrect, for the sample format converter (but would be true/correct if you'd asked about the sample rate converter).

So, if you want to leave 44kHz@16-bit files untouched (except for whatever transcoding you're doing), then you want to ensure that the sample rate converter is disabled. Since what you're trying to do requires a rate conversion, not a format conversion, you want the sample format converter disabled at all times.

Does not the "sample resolution dial" (8-32) affect the output?

Yes, it would. This would be the bits-per-sample value.

The exact effect it has, depends on if you're changing the rate or merely the format. Again, different things.

Thanks in advance!

This presumes that you're gonna get an answer, or any response.

---

[…] people who think more than 44 khz and 16bit are “necessary” […]

Direct them to watch Episode #2 (Digital Show and Tell) of Monty's Xiph tutorial videos, in which he thoroughly debunks (with demonstrations, using analogue test equipment) this fallacy, and explains what >48kHz and >24-bit(-per-sample) are actually for.

Also point out that, unless they're encoding losslessly from a high-resolution source (directly from an ADC outputting >44kHz and/or >16-bit, else transcoding from other files which were generated that way), then they're just ‘up-converting’ which makes the files larger without actually adding any fidelity (same as interpolating an image to increase the pixel-count; software can't add detail which isn't already in the original/source file). If so (up-converting), they're simply wasting resources (CPU time, storage, network capacity) with no gain.

Your optimal solution, is to convince/persuade these friends of their folly, so that they don't generate bloated files in the first place. Then, no need for any conversion, and everyone benefits (including the friends) 😀.

[Cultura-Difusa]

Thank you, I will re-read your answers, understand properly what I want, and set up my rig/profiles accordingly.

Again, thanks.