Jump to content

AIFF, WAV, bits and samples...


KeithJames
Go to solution Solved by Holger Lagerfeldt,

Recommended Posts

Finishing my first project, an 11 song album, on Logic Pro 10.7.9 freshly loaded on new mini mac (haven't gone to .8 yet).

So, I have been presented to think about something I haven't in a long time.  All prior projects were submitted to publication in WAV and in 16 bits.  I don't know what the sample rates were, probably whatever was default.  At some point, I read that 16 bit was the best to submit for mp3 distribution and I think that it didn't matter whether you submitted AIFF or WAV, I had been submitting all WAV.

But now, I realize that my entire project has been done under different settings, seemingly Logic Pro defaults:  I checked and all the recording and mastering was done on files set to AIFF, 24 bit, 44.1 sample rate.

QUESTION:  Are these good settings to submit my mp3s for streaming publication or should they be reduced to 16 bit...and what is the difference in outcome to submitting WAV or AIFF (AIF)?

Thanks,

KeithJames

Link to comment
Share on other sites

Hi Keith,

16 bit 44k1Hz WAV is still the norm for some aggregators (CD Baby, last I checked), or they'll convert from whatever you send them and to 16 bit 44k1Hz, which you don't want them to do.

You want to control that process yourself or your mastering engineer should.

WAV and AIFF contain the same sound, the difference is just the container format. WAV is the slightly more compatible format, AIFF being less common on Windows.

Some aggregators might reject AIFF, although it's perfectly compatible.

However, almost all modern aggregators and distributors accept 24 bit 44k1Hz or prefer it over 16 bit 44k1Hz.

16 bit is still preferred or even required by many radio stations due to old Dalet database systems. So it's common to make both versions, sometimes a less limiting/clipped version for radio.

MP3 is a "dead" format today, officially died in 2017 when inventors Fraunhofer killed it.

Streaming doesn't use MP3s, but formats like AAC (Apple), Ogg Vorbis (Spotify), Opus (YouTube), etc.

Just in case: never deliver MP3, AAC or any other lossy format unless specifically asked for it, as you risk transcoding, i.e. double lossy encoding in the process.

For any lossy encoding it's better to deliver a high resolution source file, i.e. you're better off sending 24 bit PCM for the aggregator/distributor to encode themselves.

So except for maybe AIFF rather than WAV, you're all good.

PS. Logic Pro has or had a bug (haven't tested it recently) when importing MP3 files, where it hard truncates the contents to 16 bit. This means with Logic, specifically, it's better to use a 16 bit dithered source for MP3 conversion, but that's more of a quirk/bug than something you need to think about in general.

  • Like 2
Link to comment
Share on other sites

  • Solution

No and yes.

No, because you'll probably be double-dithering, and the file has already gone from 32 or 64 bit float to 24 bit integer, most likely with dithering or noise shaping.

So the best would be to open the the original sessions and bounce again, this time directly to 16 bit with a single round of dither/noise shaping.

Then you'de need to check/compare every single bounce to the original bounces to make sure the session/bounces sound the same and no glitches happened.

If it's a digital mixes and you have the same starting position, you should be able to do this quickly with null tests.

That's the mastering engineer talking.

-

Yes, because double-dithering on a 24 bit file to 16 bits for all intents and purposes isn't a problem.

However, for the latter hack you normally wouldn't do this in Logic.

Instead you'd use a standalone app, like the batch converter in Izotope RX.

Not just for speed, but because it allows for more advanced dithering options, such as overload protection (due to double dithering on an exitsing file), auto-blanking, and selecting a TPDF (flat) dither instead of noise shaping - which is a good idea when double-dithering since double noise shaping is potentially worse.

In any case, it's not a biggie IRL no matter what you do here.

  • Like 1
Link to comment
Share on other sites

5 hours ago, Holger Lagerfeldt said:

double-dithering on a 24 bit file to 16 bits for all intents and purposes isn't a problem

… because basically no one who is older than, say, 20 will ever be able to even hear this additional  faint noise around 20 kHz.

(Heck, many of those "renowned" old mastering engineers never even noticed the horrible 15.something kHz buzz from their CRT displays when mastering albums or remastering reissues, while this buzz leaked into their masters and thus onto the CDs… :D)
(For the record, I was about 30 when I realized that I can't hear the CRT buzz anymore.)

  • Like 1
  • Haha 1
Link to comment
Share on other sites

5 hours ago, Holger Lagerfeldt said:

batch converter

Also for the record: MacOS has a batch converter built in, it's called afconvert.
For example, I've set up an Automator workflow to batch convert 24/44.1 wave/aiff files to CAF with ALAC compression. Those work then flawlessly with Logic, and in the long run it saves me dozens if not hundreds of GB storage space.

The basic command inside the "Run Shell Script" Automator action is pretty simple:

for f in "$@"; do
afconvert -f 'caff' -d alac -t "$f"
done

Set the workflow input to Audio Files and save as Quick Action. That's all it needs.

For more basic info on afconvert see e.g. moellerstudios.org/how-to-use-afconvert 

  • Like 2
Link to comment
Share on other sites

Thanks Holger and loukash. I pretty much understand what is being said here. I can indeed go back and do re-bounce from the earliest stages in 16 bit. I think a big takeaway here for me is that I should not be using dithering on the process of bouncing and arranging my clips until it’s all done and I’m in the master stage and then dither once at the end on the single track. I have been keeping my dithering level at the highest and it’s been on for all the bounces and etc. through the project so I’m over “fizzing” things I guess.

KeithJames

  • Like 1
Link to comment
Share on other sites

I should add that one of my concerns was that submitting the wrong bit rate or double+ dithered files - now realizing that I have been doing this - would cause some horrible distortion down the road when streamed from the wrong place.  From your answers this does not sound to be the case or a worry…., and regarding double dithering just that I have been reducing quality and adding a layer of too much “hiss” so to speak

  • Like 1
Link to comment
Share on other sites

8 minutes ago, KeithJames said:

double+ dithered files - now realizing that I have been doing this - would cause some horrible distortion down the road when streamed from the wrong place. 

Depending on the streaming/download format and the lossy bitrate, the high frequency noise added by dithering will be usually stripped anyway. For example, AAC 256 kbps like in the iTunes Store will shave off frequencies above ±17 kHz if they seem "irrelevant" to the compression algorithm, depending on their loudness and complexity, so that there's more bitrate left for the frequencies that a human ear can actually hear. The dither noise is usually above that.

As every so often, I can only recommend Ian Shepherd's blog:

  1. productionadvice.co.uk/when-in-doubt-dither
  2. productionadvice.co.uk/when-to-dither

(Ian did the remastering on a couple of reissue projects that I produced some 12-15 years ago, and eventually he encouraged me to start with mastering myself. :)

  • Like 1
Link to comment
Share on other sites

Just found another interesting article on dithering here:
producerdj.com/lies-damn-lies-and-dithering

And I learned something new: white noise dithering is actually better than noise shaping. In iZotope RX that would be TPDF vs. MBIT+. In Logic: POW-r #1 vs. #2/#3
Did a few quick tests right now and based on what I see in the spectrogram, it seems so.

So: When in doubt, dither. :)

  • Like 1
Link to comment
Share on other sites

12 hours ago, loukash said:

Depending on the streaming/download format and the lossy bitrate, the high frequency noise added by dithering will be usually stripped anyway. For example, AAC 256 kbps like in the iTunes Store will shave off frequencies above ±17 kHz if they seem "irrelevant" to the compression algorithm, depending on their loudness and complexity, so that there's more bitrate left for the frequencies that a human ear can actually hear. The dither noise is usually above that.

Indeed, however this is irrelevant to whether dither works or should be applied or not. The de-correlating effect of the dither on the audible periodicity of the truncation distortion has still taken place on the source material used for lossy encoding.

 

11 hours ago, loukash said:

Just found another interesting article on dithering here:
producerdj.com/lies-damn-lies-and-dithering

And I learned something new: white noise dithering is actually better than noise shaping. In iZotope RX that would be TPDF vs. MBIT+. In Logic: POW-r #1 vs. #2/#3
Did a few quick tests right now and based on what I see in the spectrogram, it seems so.

I don't want to step on anyone's toes, but that article is a bit of a rambling rant.

Let me nuance things:

Izotope isn't hiding or obfuscating anything when it comes to dither. To me it's quite clear what's going on, how, and why, based on their documentation.

Allow me to paraphrase myself from another post:

Dither Amount - "None" actually means no dithering, but still noise shaping of the distortion, however it can lead to some minor audible artifacts.

Noise shaping isn't just equalized dither, as the article claims. In fact noise shaping and dither are two processes, which is why it's possible to have noise shaping without dither.

Colored dither is merely equalizing the dither noise, but subsequent quantization can create new spectral distortion artifacts.

Noise shaping adds an adjustment or feedback loop between the shaping and the dither, unlike a simple equalized dither. This allows it to shape both the dithering noise and the quantization noise at the same time. That's why it's so effective, not just because it's shifting the frequency spectrum in some variation of an inverse equal-loudness contour.

TPDF (triangular probability density function) can be obtained by setting noise shaping to none and dither amount to high. Decently effective and no artifacts per se, but a slightly higher noise floor.

This isn't hidden at all by Izotope, like the article claims, and why would they.

Auto-blanking - Turns off dither depending on the circumstances. "When Quantized" is sort of intelligent in the way it detects already dithered signals to avoid double-dipping. So that fixes another worry in the article. "On Silence" (as in literally no input signal) and "Off" should be self-explanatory.

Limit Noise Peaks - Mainly relevant if you use the Ultra/aggressive settings. This fixes another complaint the article posed, so it's actually avoidable.

Suppress harmonics - Relevant only if you're not dithering and therefore truncating instead (for some reason). Enabling this function can lower or rather shift the harmonic overtones caused by the quantization distortion.

-

Finally, TPDF isn't "better" than noise shaping. From a purely objective standpoint a good noise shaping algorithm like MBIT+ is better.

But subjectively, I concur. I don't like (aggressive) noise shaping either - in exaggerated listening tests. I'm not sure I could tell in an ABX blindtest, but the fact that there's a theoretically audible difference I don't like, is enough for me.

  

13 hours ago, KeithJames said:

Thanks Holger and loukash. I pretty much understand what is being said here. I can indeed go back and do re-bounce from the earliest stages in 16 bit. I think a big takeaway here for me is that I should not be using dithering on the process of bouncing and arranging my clips until it’s all done and I’m in the master stage and then dither once at the end on the single track. I have been keeping my dithering level at the highest and it’s been on for all the bounces and etc. through the project so I’m over “fizzing” things I guess.

KeithJames

Just to clarify: You should always dither when going from a higher to a lower bit depth.

But first and foremost you should try to stay at the highest possible bit depth instead of dithering multiple times.

When you're forced to use multiple rounds of dither, then try to use flat/TPDF dither and avoid noise shaping until at least the last round of quantization.

Logic doesn't have a TPDF option, so POW-r#1 is the closest, despite being equalized (but not noise shaped) dither.

  • Like 2
Link to comment
Share on other sites

Thank you Holger for this further information. It is helpful without a doubt. I am an artist by sound ear and musicality, but to call myself who is not even a computer native at the age of 66, a technical music producer is a far cry, but that is what I do from start to mastering; I do it all. So I am not cognizant of enough technicalities to be too worried about fixing my current project to some of the things I’ve have learned here. Going back through and redoing all the steps will hazard, the delicate piece of artwork I’ve already created. But I still am very interested in learning more about the technical side and definitely have learned things from you about dithering here which will inform me going forward, and I very much appreciate your time in helping me!

  • Love 1
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...