Waveform tweaks

[m0dB]

Adjust the scaling of the waveform data: removing a * 2 factor which doesn't seem justified and undo a scaling curve applied to the waveform.filtered.all data which also lacks an explanation.

See my analysis here #12448

[JoergAtGithub]

The scaling for looks good now, but there seems to be something, that clamps the visual amplitude. See the following video of AllShader RGB R/L and notice the Peak Indicator of the VU-Meter:

VisualClampingAllshadersRGB_RL.mp4

For comparision, the next video is recorde with exactly the same setting but GLSL RGB stacked waveforms:

NoVisualClampingRGB_stacked.mp4

[m0dB]

I see what you mean. Can you send me this track so I can have a look at the values?

[m0dB]

@JoergAtGithub the scaling curve seems intentional. See the discussion @Swiftb0y dug up in #12448

[m0dB]

What visual gain are you using? On what Replay Gain do you have on that track? I got -9.59306 dB.

Here is looks like this:

Which seems correct, considering the ReplayGain. If I boost the visual gain with the replay gain (9.59306 dB =~ x3) I get this:

[m0dB]

I now made the gain compensation optional, so each renderer can decide whether to apply it or not. With this the amplitude is pretty consistent, only the "filtered" types are a bit different depending on the content, but I think for now this is as good as it gets. Should be merged to 2.4. Ping @ronso0 @JoergAtGithub

[mxmilkiib]

Cool cool. (On my phone and busy for a while)

Btw, because it hasn't been stated; the fact that the test track just above looks clipped before any gain.. is that just accidental? Like, that's the waveform like Audacity or whatever shows? Because it's maybe not the best example to use when specifically looking into waveform gain/height :) i.e. if the render somehow gets clipped itself, there's no telling that apart from the true waveform, or maybe the fact it looks clipped is a problem itself.

[Swiftb0y]

little modernization opportunities.

[m0dB]

@mxmilkiib wrote:

Btw, because it hasn't been stated; the fact that the test track just above looks clipped before any gain.. is that just accidental?

This is caused be the scaling curve that is being applied to the waveform and this is why it looks different in Audacity than in mixxx. I think it's a bad idea, but it's unclear to me if this is a bug or a feature, so I didn't want to change it.

[mxmilkiib]

Aah right, thank you, had kinda got the gist from the convo but that makes it explicit. I have always had this niggling feeling that Mixxx waveforms were ever so sightly.. "fatter".

Does feel like, if intentional, it would be for some take re UX. Maybe there's a convo somewhere from the date of the blame commit, if no-one remembers the rational.

It is kind of an imposition though, with no option to toggle it, or/tho is that what you've just added? I did wonder how this relates to the existing waveform scaling preferences..

I guess this in a way relates to #11833

(2p ramble; I've still not settled on a preference for scaling yet. Tbh, again, Pioneer is the club standard, they'll have had certified and paid UI people come up with those, certainly by the recent start of the era of 3000s, though obvs. their exact magic is private. Figuring that out using, like, a series pure tone samples, would be a fun for some but a super low priority side quest..)

Edit: to be clear; a (previously?) unavoidable scaling curve for the waveform graphic that makes some tracks look like they're clipping, that's.. really not very good, must be and have been confusing sooo many people.

[ronso0]

@m0dB Thank you, this is much more consistent now!

[Swiftb0y]

Code LGTM, though in the future, I'd really like to gain back some consistency by removing the scaling in the analyzer. Are you interested in looking into that in the future?