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: Reduce by at least 3 dB before upsampling (Read 1693 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Reduce by at least 3 dB before upsampling

Here is an advice I saw many times and which I would like to know how to apply it in fb2k.
Upsampling can be done without causing damage if the audio level is reduced by at least 3 dB before upsampling.

I have set Resampler SoX in fb2k DSP Manager but I cannot find how to reduce the gain by 3db before the resampler.

This is the complete article that explains why.
Intersample Overs in CD Recordings
Enjoy life now, this is not a rehearsal.

Re: Reduce by at least 3 dB before upsampling

Reply #1 –
It seems nobody can provide real life examples with proofs where clipping caused by resampling (with decent quality resampler) is audible.

But you can use foo_dsp_amp ( https://foobar.hyv.fi/?view=foo_dsp_amp ) to lower volume before resampler in DSP chain. Also "Gain / Scale" DSP from foo_dsp_utility ( https://www.foobar2000.org/components/view/foo_dsp_utility )
If you are resampling on playback only, not on conversion to file, just lowering output volume in fb2k is doing the same. Because fb2k is processing signal in 32 bit float (64 bit float in 64 bit version), so no clipping is happening before conversion to fixed point.

Re: Reduce by at least 3 dB before upsampling

Reply #2 –
Quote from: Bogozo on
It seems nobody can provide real life examples with proofs where clipping caused by resampling (with decent quality resampler) is audible.

But you can use foo_dsp_amp ( https://foobar.hyv.fi/?view=foo_dsp_amp ) to lower volume before resampler in DSP chain. Also "Gain / Scale" DSP from foo_dsp_utility ( https://www.foobar2000.org/components/view/foo_dsp_utility )
If you are resampling on playback only, not on conversion to file, just lowering output volume in fb2k is doing the same. Because fb2k is processing signal in 32 bit float (64 bit float in 64 bit version), so no clipping is happening before conversion to fixed point.

I'm trying to do it in real-time and not to modify the file (no point of wasting space).
I'll try the plugin you suggested.
Thanks
Enjoy life now, this is not a rehearsal.

Re: Reduce by at least 3 dB before upsampling

Reply #3 –
This explains why unconditionally apply a fixed amount (e.g. 3dB) of volume reduction may not solve everything, it is a long article so don't skip and read until the end.
http://archimago.blogspot.com/2019/06/guest-post-why-we-should-use-software.html

The Archimago article above contains a track with more than 5dB of intersample over, the file can be downloaded below. Please read the forum replies as well because two members provided ABX results in subsequent replies:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-1763519


Also, John Siau made misleading claims about lossy codec induced overs:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-2103400

I made a correction below and John Siau haven't posted in ASR until now:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-2103649

Re: Reduce by at least 3 dB before upsampling

Reply #4 –
Quote from: bennetng on
This explains why unconditionally apply a fixed amount (e.g. 3dB) of volume reduction may not solve everything, it is a long article so don't skip and read until the end.
http://archimago.blogspot.com/2019/06/guest-post-why-we-should-use-software.html

The Archimago article above contains a track with more than 5dB of intersample over, the file can be downloaded below. Please read the forum replies as well because two members provided ABX results in subsequent replies:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-1763519


Also, John Siau made misleading claims about lossy codec induced overs:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-2103400

I made a correction below and John Siau haven't posted in ASR until now:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-2103649

Lucky for me I only have lossless 😀
Enjoy life now, this is not a rehearsal.

Re: Reduce by at least 3 dB before upsampling

Reply #5 –
Quote from: bennetng on
The Archimago article above contains a track with more than 5dB of intersample over, the file can be downloaded below. Please read the forum replies as well because two members provided ABX results in subsequent replies:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-1763519
There are just ABX results that just proove that people can hear differences between files with different peak levels.

Re: Reduce by at least 3 dB before upsampling

Reply #6 –
Quote from: Bogozo on
Quote from: bennetng on
The Archimago article above contains a track with more than 5dB of intersample over, the file can be downloaded below. Please read the forum replies as well because two members provided ABX results in subsequent replies:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-1763519
There are just ABX results that just proove that people can hear differences between files with different peak levels.

Peak level I believe anyone can hear.
I am just wondering about the up-sampling  and it looks to me for the moment at least that at best it will sound the same.
Just like upresizing an image. There will be the same interpolated data but nothing else so it’s not worth the hassle.
Enjoy life now, this is not a rehearsal.

Re: Reduce by at least 3 dB before upsampling

Reply #7 –
Quote from: Bogozo on
There are just ABX results that just proove that people can hear differences between files with different peak levels.
The change in peak level is a necessary consequence caused by clipping, in other words, there must be level differences between the clipped version and non-clipped version.

So when some people can hear this difference, it means some people can identify the distortion caused by intersample over: the distorted peaks.

Re: Reduce by at least 3 dB before upsampling

Reply #9 –
Quote from: bennetng on
Quote from: Gabriel Schwartz on
Lucky for me I only have lossless 😀
There are some other forms of clipping needed to be take care of, even if you don't use lossy formats, like this:
https://www.audiosciencereview.com/forum/index.php?threads/digital-filter-game.23795/post-810355

I’ve read that article.
My Holo Audio May KTE DAC has a separate DSD converter.
In any case I was not going to upsample DSD as this is going into the wrong direction.
In FB2K there is no way (that I know of) to upsample DSD without converting it to PCM first which defeats the whole DSD purpose.
After trying a few time with -3db reduction before SoX resampler and going up to 705/768 only for 44.1 to 384
Enjoy life now, this is not a rehearsal.

Re: Reduce by at least 3 dB before upsampling

Reply #10 –
Quote from: Gabriel Schwartz on
In FB2K there is no way (that I know of) to upsample DSD without converting it to PCM first
No software or hardware doing digital processing can upsample (e.g. DSD64 to 128)or downsample DSD (e.g. DSD128 to 64) without converting to PCM first. Including HQPlayer that you mentioned in other posts. Read this:
https://hydrogenaud.io/index.php/topic,124362.msg1029281.html#msg1029281

Re: Reduce by at least 3 dB before upsampling

Reply #11 –
I think the test files in the linked thread are valid for simulating effect of clipping. The loudness is the same, only peaks are different. You can't really test it any other way.

But I would be curious to know what kind of equipment was used for example by that melowman to ABX the two files. The difference in the signal lasts for just 17 nanoseconds. I can easily hear loudness difference when playing just the clipped peak. But it absolutely drowns to the background when it happens during the sudden transient. But if hardware distorts with the loud high frequency noise, that might be detectable.

Re: Reduce by at least 3 dB before upsampling

Reply #12 –
Here is another member with a positive result:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-1768870

I myself cannot hear the differences with these two files. What I disagree is that many people think that intersample over only happens on highly brickwalled loudness war contents.

As demonstrated, some genres can produce high intersample peaks even if the overall loudness is not very high.

Re: Reduce by at least 3 dB before upsampling

Reply #13 –
Quote from: bennetng on
Here is another member with a positive result:
https://www.audiosciencereview.com/forum/index.php?threads/lets-develop-an-asr-inter-sample-test-procedure-for-dacs.49050/post-1768870

I myself cannot hear the differences with these two files. What I disagree is that many people think that intersample over only happens on highly brickwalled loudness war contents.

As demonstrated, some genres can produce high intersample peaks even if the overall loudness is not very high.

Agree.
Its mainly placebo effect.
Enjoy life now, this is not a rehearsal.

Re: Reduce by at least 3 dB before upsampling

Reply #14 –
Quote from: Case on
I think the test files in the linked thread are valid for simulating effect of clipping. The loudness is the same, only peaks are different. You can't really test it any other way.
But how we can be sure that audible differences are caused exactly by intersample clipping and not just by volume difference of these peaks?

Re: Reduce by at least 3 dB before upsampling

Reply #15 –
I'm not sure what you mean. This is the exact same thing, only properly controlled to exclude possible DAC behavior. If the clipping is left to actual DAC its level would be unknown.
But I do not believe that the ridiculously tiny difference in the loudness for a few nanoseconds is audible. Any difference that something as weak as human hearing can detect must come from some larger difference in the output.