Gnarly MP3 compression artifacts

[MattCurious]

Hi everyone

I'm exporting a wav at 48kHz, 24bit and compressing to MP3 at 192 (I forget the measure, I'm terrible at this bit)

There's a little digital clipping on the master out.

On this occasion I'm getting that gnarly gurgling sound on hi hats.

What causes that and how do I stop it?

Cheers
Matt

[Danny]

Matt,

Did you uncheck Dithering when you exported from the DAW?. Dithering usually applies when transforming one bit rate to another. It sometimes causes noise in conjunction with quantization of your tracks.

Just curious. Why exporting at such a high bit rate? Are you recording at that rate? If so, it's going to tax your CPU and use more memory and drive space.

The highest Bit Rate I have seen libraries ask for in uploading tracks is 16 Bit /44.1 Hz.

After all, A CD is only 16 bit. Perhaps you have a good reason to record this way...

[MattCurious]

Hi Danny

I don't believe I did - I can give that a try.

The mastering guy I sometimes use asks for 48/24 on the wav file. But I then compressed to MP3 to get feedback on the work in progress from a collaborator.

Cheers
M

[mojobone]

Bit depth, sample rate; please don't conflate them with encoding rates which are expressed in kilobits/sec and really are just a measure of how much data is being thrown out.

If you're getting artifacts at such relatively high bit rates, odds are your levels are way hotter than they need to be. It's also possible that what you're hearing is a result of normally inaudible audio compression artifacts or inter-sample peaks that the encoder can't handle, but the cure is the same; back off the levels a bit. And some brickwall limiters aren't really, meaning that if you set the threshold at -.03dBFS, there's still gonna be some stuff up there between -.03dB and 0dBFS. A spectrum analyzer plugin can tell you if that's the case. Try giving yourself an entire decible of headroom; the consumer can easily turn up the knob at the other end if it's not loud enough to suit.

[andygabrys]

.....Why exporting at such a high bit rate? ...

some publishers are asking for 48 k 24 bit now FYI. But it also doesn't make sense to up sample and bit depth on export (unless its a one in a blue moon scenario) so.....

After all, A CD is only 16 bit. Perhaps you have a good reason to record this way...

Its up to you but I always record 48 k (cause just about everything I do is for TV) and at 24 bit depth.

Recording at 24 bit gives you more dynamic range, and while not really important if you are mastering loud as the dynamic range is cut drastically, but at least you don't have to run levels hot on the way in as the noise floor is very close to -bazillion dB in 24 bit.

YMMV.

What Mojo said about more headroom is super important. In fact he and I went through this about 4 yrs ago on this forum.

I don't usually go -1.0dB but its a totally safe way to go.

If exporting your hi-res with an output ceiling of that bugs you, you can go back to -0.2 dB or whatever you like and use one of the SONNOX plugins to convert to mp3 or AAC as they intelligently calculate the output level of the converted compressed file and pull down the input accordingly so you end up with clean mp3s. There are two SONNOX plugs that do it: https://www.sonnox.com/products one is less featured and somewhat affordable and the other is mega $$$ although they have a 40% off sale right now.

[MattCurious]

Thanks, guys - this is an area where I'm woefully lacking, so this is super helpful. I'll pull the levels back and see if that cures it.

[Len911]

The mastering guy I sometimes use asks for 48/24 on the wav file. But I then compressed to MP3 to get feedback on the work in progress from a collaborator.
Cheers
M

https://blog.soundcloud.com/tag/private-link/

I'm not sure why you need to even mess with mp3's to your collaborator. You can upload your wav to soundcloud, make it private and downloadable and just email the secret link, and your collaborator can download the wav. They can also stream it, though not as good as a downloaded wav, however they can comment on the timeline in the streaming version.

[Casey H]

True, but for simple things like opinions from others, mp3s, with their smaller size, are still easier to deal with. You can attach an mp3 to an email or even if you send a download link, it's a lot faster.

I think most people use mp3s for a lot of things other than sharing final mixes or tracks targeted to go into a DAW.

I use WFH producers for all my work and during the process, they always send mp3s for review and send the hi res wav or aif for the final, once approved. I would much rather (during the production process) get an mp3 as email attachment, click and play, then have to download a 40MB file.

Also, figuring out how to make mp3s sound reasonable is a necessary part of being in this business as you have to upload mp3s to sites like Taxi for submissions and many libraries ask for them for initial review.

My 2 cents.
Casey

The mastering guy I sometimes use asks for 48/24 on the wav file. But I then compressed to MP3 to get feedback on the work in progress from a collaborator.
Cheers
M

https://blog.soundcloud.com/tag/private-link/

I'm not sure why you need to even mess with mp3's to your collaborator. You can upload your wav to soundcloud, make it private and downloadable and just email the secret link, and your collaborator can download the wav. They can also stream it, though not as good as a downloaded wav, however they can comment on the timeline in the streaming version.

[MattCurious]

True, but for simple things like opinions from others, mp3s, with their smaller size, are still easier to deal with. You can attach an mp3 to an email or even if you send a download link, it's a lot faster.

Exactly that

It's not a problem I usually have, even with slightly hot tracks - there's obviously something about this one that's disagreeing with the MP3 compression (the wav is fine) - but Casey is exactly right; I want to know what causes it so I can avoid it in mixdowns for Taxi and others.

[Len911]

The mp3 codec determines what to throw away by using algorithms based on psycho-acoustics, frequency masking, and perception, and how much to throw out depending on the bit depth and sampling frequency you give it to work with.

Higher frequency sine waves would stand to reason to be more problematic than lower frequency sines, as there would be more of them, and if the codec thinks they are being masked, a cymbal being masked by a guitar or piano or their harmonics, because they think they are masked might throw more away, thinking it wouldn't be noticeable, though sometimes they are. It might help to either create a space in the cymbal frequencies for only the cymbals and eliminate the guitar and piano harmonics, or give the mp3 more to work with in terms of a higher bit rate and sampling frequency.

Mp3's themselves can be very small or up to nearly a pcm file with no loss. The size you use would also depend on whether one was still on dial-up or broadband, and the size of their hard drives.
If you are only allowed to upload mp3's, then you would want something like 320kbps, but if allowed pcm file like wav or aiff, there's not much difference in size.

You could try a different codec, Lame and Fraunhofer are the two most popular.