Would it be better for Replay Gain to be off by default?

[gevaerts]

tl ;dr summary:

It isn't extremely rare, it's only as rare as unamplified music - it's very common!

Yes, but such music is just a tiny fraction of all the music out there hence the compromise.

Maybe, but I'd say for a specific user either zero (or almost zero) tracks show the issue, or nearly all of them.

[cereal_killer]

Wow, impressive discussion going on here. I am no audiophile and my technical knowledge is not enough to fully understand what replay gain does, or how rockbox works regarding that, but I dare to put my 2 cents in.

I was surprised, hearing that a feature in rockbox, that is somehow processing and changing the sound output is on by default. I personally would prefer if everything is off and I could choose to switch a setting on if needed. Even if the effect is not noticeable for the kind of music I have on my player.

Surely the following comparison is not valid, but I wouldn't like an EQ setting to be on by default either. I like every sound setting turned off (no EQ, no bass or treble enhancement...) Obviously replay gain is a good thing and I think I need to look into that.

Greetings,

c_k

[saratoga]

I was surprised, hearing that a feature in rockbox, that is somehow processing and changing the sound output is on by default. I personally would prefer if everything is off and I could choose to switch a setting on if needed. Even if the effect is not noticeable for the kind of music I have on my player.

Its only on if you've scanned your files with replaygain.  And if for some reason you scanned your files with replaygain tags for purposes of not using replaygain tags, you can disable it using the setting we provide. 

Yes, but such music is just a tiny fraction of all the music out there hence the compromise.

Maybe, but I'd say for a specific user either zero (or almost zero) tracks show the issue, or nearly all of them.

Thats the annoying thing about defaults, they're never right for everyone.

[Julian67]

.........it's only as rare as unamplified music - it's very common!

Yes, but such music is just a tiny fraction of all the music out there hence the compromise.

It might be true that there has been more amplified music recorded than unamplified but there again it might not be (data? facts?), but  "a tiny fraction"?  The vast catalogues of Decca, Universal, Deutsche Grammophon, HMV, EMI, Mercury et al are so insignificant that they can be casually dismissed?

It's a truism that defaults can never be right for everyone.

I would say that RG with clip prevention (might change relationship between adjacent tracks but will anyone notice?) is a compromise between vanilla (never degrades audio; works for everyone) and current default (may degrade audio; works brilliantly for some but is very bad for others).  Obviously I think vanilla is the saner setting for a global default but the incumbent always has inertia and familiarity, and such is the nature of compromise....

I doubt very strongly indeed the validity of the presumptions that having unamplified music in a collection is unusual, or that having lots of it is extremely rare.  Reasoning based on those premises is inherently flawed (unless the resulting action applies only to oneself) because it's generalising from the personal and particular.

Maybe, but I'd say for a specific user either zero (or almost zero) tracks show the issue, or nearly all of them.

Clearly there are people who have only amplified music and probably some who have only unamplified, but can I propose that having some or lots of both in varying proportions is not at all unusual?

And if for some reason you scanned your files with replaygain tags for purposes of not using replaygain tags, you can disable it using the setting we provide.

Actually I'm not being perverse in applying RG tags to my collection.  A few years ago I did a big, laborious tidy up of my collection's directory and file names and metadata and being aware of RG realised this was a good time to try it.  But I found it impossible to assess RG without the RG tags    so I tagged my existing collection and modified my transcode process (flac to ogg or mp3) to include album gain for future encodes.  As stated earlier it's automated (scripted and aliased) so it's no extra (human) work.  After running into instances of distortion and sometimes bad clipping I looked at why this might be and realised that positive album gain values can leave you with four choices:

i) a pre-amp value so low that some tracks are ridiculously quiet

ii) the possibility of distortion/clipping

iii) applying clip-prevention and having the relationship between some adjacent tracks artificially altered.

iv) switch it off

So I went back to using the volume buttons for the most part, but if I make a playlist of rock or pop etc then RG becomes useful and I can toggle it on.

You might recall that in my original post I pointed out that beyond the actual benefits/disadvantages of RG there is the issue that it isn't obvious to any new or casual user of Rockbox that RG is even enabled because the menu item is highly ambiguous:

"Track Gain If Shuffling"  For anyone discovering Rockbox by using it (default male human behaviour since forever), or to anyone who doesn't know what Replay Gain is, then that is in no way a declaration that RG in album mode is on or that the player perhaps isn't doing what you think it is doing.

I'll repeat what I posted previously because I think it's valid and doesn't have a logical or practical weakness:

"I have no idea who the listener is or what they like so vanilla makes the best default and let the person doing the tasting add the syrup.  Definitely don't hide a raw chilli pepper in the middle."

[bluebrother]

This discussion is really getting useless. Both parties claim that their "default" type of music is a relevant portion but neither provides numbers. Just listing a couple of labels doesn't help any way -- there are enough labels to list for the opposite argument as well. Also, the number of labels or records is totally irrelevant -- if any the most common type of a typical Rockbox user would be of interest. We don't have any data for this afaik, but given that younger people are more familiar with computer stuff and therefore likely to be the majority of Rockbox users, and younger people (at least seem to) be rather on the pop side the current default matches simply a greater number of Rockbox users.

Instead of arguing about music types other questins should be answered: are there any side effects when enabling "prevent clipping" like degrading of audio quality for normal playback? How much processing power does enabling it by default cost? Are there other reasons to not enable it by default?

I.e. are there reasons good enough to change the default behaviour or are there important reasons against it?

[torne]

I had a look at the code and it looks like clipping prevention is "free", as it just uses the RG tag's recorded peak level to determine whether the level of gain specified will cause clipping and caps it if so. Unless I've misinterpreted the implementation the only downside will be that music with a positive gain adjustment will sometimes sound quieter than other tracks (and if clipping prevention is not turned on there, it *may* have produced audible clipping instead).

Let's just turn clipping prevention on by default as well.

[Julian67]

... music with a positive gain adjustment will sometimes sound quieter than other tracks (and if clipping prevention is not turned on there, it *may* have produced audible clipping instead).

As far as I can tell this is indeed the result of running RG with clip prevention.  The upside is no more potential for clipping or harmonic distortion, while the downside is that adjacent tracks on some albums will have different relative loudness than on the CD (or other source).  This is odd and defnitely not an effect anyone might seek out, but it's usually not going to be easily apparent like clipping or harmonic distortion and will usually require great familiarity with a  recording, or side by side CD vs File playback comparison, to identify it.  It's hard to see "a fault that will occasionally occur but probably not be noticed" as an ideal default when it's equally easy to achieve unmanipulated playback, but it's a huge improvement on current release default.

Old farts with beards talk about "a wire with gain" but I shaved this morning.