|
|
|
05-10-2017, 07:56 AM
|
#1
|
Human being with feelings
Join Date: Nov 2016
Posts: 23
|
First Reaper master ever: mp3 export clipping
Hey guys, finally closing my first full home recording production ever, basically a whole big pilot test. I'll probably let you hear the final product, it's a cover.
On to the meat: I made a quick essential mastering phase (eq, comp, multicomp, limiter). All ok up to limiting.
I had a -1db ceiling just for safety, threshold was about -8.9. The track is very dynamic (minimal vocal and guitar ballad, plus backing vocals). GR never went more than 2.5/3db on the loudest peaks. So up to there, I thought I was safe.
Then I exported mp3, 320 CBR, and the export slightly brickwalled the loudest peaks resulting in sensible clipping. Note I still dind't try a wav export.
Re-importing the track back in reaper, I noticed dbfs peaks are logically at max -1db, but the RMS fader goes positive, up to +6 in the clipping parts, and also slightly less in other slightly less loud sections when the clipping isn't perceivable but it seems like it still did happen.
So I'm asking myself, open-ended questions here:
1. what's happening and where am I wrong?
2. was it just the mp3's own compression to frak up my headroom, so should I set an even lower ceiling while rendering to mp3?
3. or did I squash the threshold too much on the -1 ceiling? so in that case, should I lower the ceiling, or set the threshold higher, or both, or what?
4. if nothing will clip while exporting to wav on the same settings, does it mean it's all the mp3's fault?
Up to you, any advice is welcome. Thanks
|
|
|
05-10-2017, 08:11 AM
|
#2
|
Administrator
Join Date: Mar 2007
Location: NY
Posts: 15,749
|
mp3 as you know is lossy, so it's interpolating samples when decoding, which can lead to overs that did not exist in the original. This is more likely to happen the closer the original is to clipping.
It's a rough approximation but imagine a sine wave where you chop off the tops of the peaks, then reduce to a discrete set of samples, then try to reconstruct the wave by interpolating between the samples. The interpolation is likely to restore some of the previously chopped-off peaks.
|
|
|
05-10-2017, 08:12 AM
|
#3
|
Human being with feelings
Join Date: Aug 2014
Posts: 11,044
|
It's a different scale.
If the peak meter is below 0.0dBFS, then your audio is not clipping.
The RMS meter is just telling you it's really loud, not that it's clipping.
|
|
|
05-10-2017, 08:13 AM
|
#4
|
Human being with feelings
Join Date: Aug 2014
Posts: 11,044
|
Quote:
Originally Posted by schwa
mp3 as you know is lossy, so it's interpolating samples when decoding, which can lead to overs that did not exist in the original. This is more likely to happen the closer the original is to clipping.
It's a rough approximation but imagine a sine wave where you chop off the tops of the peaks, then reduce to a discrete set of samples, then try to reconstruct the wave by interpolating between the samples. The interpolation is likely to restore some of the previously chopped-off peaks.
|
That's what I was thinking at first too, but then:
Quote:
Originally Posted by Rizzo
Re-importing the track back in reaper, I noticed dbfs peaks are logically at max -1db, but the RMS fader goes positive, up to +6...
|
|
|
|
05-10-2017, 08:27 AM
|
#5
|
Administrator
Join Date: Mar 2007
Location: NY
Posts: 15,749
|
I was answering technically/generically about whether an mp3 can clip if the source audio isn't clipping. But yeah, in this specific case, from the description the mp3 is not actually clipping. The RMS offset is +14dB by default, so it can read over zero even if there are no actual overs in the audio.
If it sounds like it's clipping but the meter's not showing red, it's probably just loss of dynamic range from the mp3 compression.
|
|
|
05-10-2017, 10:00 AM
|
#6
|
Human being with feelings
Join Date: Nov 2016
Posts: 23
|
Thanks everyone for replying, and also for explaining how mp3 encoding works.
I'll try exporting a regular wav and see what happens.
Also, in my master meters I have my display gain set to 0 and display offset to 14. Is this correct for monitoring?
|
|
|
05-10-2017, 10:31 AM
|
#7
|
Human being with feelings
Join Date: Aug 2007
Location: Near Cambridge UK and Near Questembert, France
Posts: 22,754
|
I had similar issues with my only track on Soundcloud. Distorts like crazy yet the original was clean.
Never bothered to get to the bottom of it but I havent bothered to put anything else on there since. Shame.
I gather at least part of it is down to Soundcloud itself messing with the tracks.
__________________
Ici on parles Franglais
|
|
|
05-10-2017, 12:38 PM
|
#8
|
Human being with feelings
Join Date: Aug 2014
Posts: 11,044
|
Quote:
Originally Posted by ivansc
I had similar issues with my only track on Soundcloud. Distorts like crazy yet the original was clean.
Never bothered to get to the bottom of it but I havent bothered to put anything else on there since. Shame.
I gather at least part of it is down to Soundcloud itself messing with the tracks.
|
They have a really low bandwidth, or they did.
They recommend limiting to -0.5dBFS, but I've had one track that distorted way below that anyway. Years ago they told me they were going to have better algorithms for encoding, but I don't know if that ever happened.
|
|
|
05-10-2017, 12:43 PM
|
#9
|
Human being with feelings
Join Date: Nov 2016
Posts: 23
|
Guess I solved the problem. With the first applied settings, the track with slightly brickwalling in wav as well. I was just squeezing a bit too much. SO, dialed down the attack and up the threshold until the compression felt bearable to my ears and liking.
Now threshold is at -6 db. The final result isn't as loud as before (averaging at -3db and just the peaks at -1db) but it's way less compressed and I like it. About 10 db of dynamic range, analysis says. Is it alright for listening standards, or should I squeeze it a bit more? I do like it.
|
|
|
05-10-2017, 12:46 PM
|
#10
|
Human being with feelings
Join Date: Aug 2014
Posts: 11,044
|
Quote:
Originally Posted by Rizzo
Guess I solved the problem. With the first applied settings, the track with slightly brickwalling in wav as well. I was just squeezing a bit too much. SO, dialed down the attack and up the threshold until the compression felt bearable to my ears and liking.
Now threshold is at -6 db. The final result isn't as loud as before (averaging at -3db and just the peaks at -1db) but it's way less compressed and I like it. About 10 db of dynamic range, analysis says. Is it alright for listening standards, or should I squeeze it a bit more? I do like it.
|
10dB DR is plenty squeezed!
|
|
|
05-10-2017, 12:55 PM
|
#11
|
Human being with feelings
Join Date: Nov 2016
Posts: 23
|
Ignorant question here: so how could it be labelled as squeezed already, but it's still way less loud than its reference commercial track?
Or, is there a way for "loud" *and* "dynamic"? Again, just getting my feets wet on mastering.
|
|
|
05-11-2017, 12:02 AM
|
#12
|
Human being with feelings
Join Date: Aug 2014
Posts: 11,044
|
Quote:
Originally Posted by Rizzo
Ignorant question here: so how could it be labelled as squeezed already, but it's still way less loud than its reference commercial track?
|
It all comes down to arrangement and frequency balance.
|
|
|
05-15-2017, 04:17 AM
|
#13
|
Human being with feelings
Join Date: Mar 2016
Posts: 444
|
Quote:
Originally Posted by Judders
It all comes down to arrangement and frequency balance.
|
^^^
Check your low end - quite possible that you have way more lows there than the reference track, but your speakers can't reproduce them. Otherwise, it is also possible that your lows are more smeared in time, while in the reference they are shorter. Make sure no sub-bass comes from parallel FX (reverbs, delays). Basses take a lot of energy and achieving the right balance is tricky. A useful technique is to EQ the hell out of the base drum's lows to focus it in certain frequency (usually corresponding to the song's key). It's like tuning and dampening the drum. Of course actually tuning and dampening a real drum (or choosing a more appropriate sample) will sound better, but you will still want to further EQ the low end when loudness is of major concern.
There are also psychoacoustics at play. Generally the more energy you move into the high-mids, the louder it will sound, but that may also make it sound harsher. Achieving the right frequency balance is a very nuanced thing and there is way more to cover.
Last edited by avocadomix; 05-15-2017 at 05:41 AM.
|
|
|
Thread Tools |
|
Display Modes |
Linear Mode
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT -7. The time now is 08:42 AM.
|