Mixdown to mp3 peaking at 0db

I have a limiter (ozone maximizer) set at -0.3db true peak, and the stereo out peak levels are consistent with this number when I mixdown to wav. For some reason, mixdown to mp3 is showing peak levels of 0db. I’ve been trying to troubleshoot this for a while, and I found that dithering, bitrate, sample rate, and “high-quality mode” don’t seem to affect this at all after testing several different values. Not sure where the extra 0.3db is coming from - if cubase is somehow normalizing the audio or if the encoding somehow introduces it. Maybe I misunderstand this issue and someone can shed some light on where I went wrong if that’s the case.

Process of compression to MP3 is happening after the limiter and all the processes. MP3 is lossy data compression - it’s changing your uncompressed audio, thus creates additional peaks.
iZotope RX is the only professional software that I know that allows for “limiting” MP3 files while rendering. Of course I don’t know them all.


I see, so it might be compression artifacts. I tried mixdown to mp3 with the limiter at different ceilings: both 0db and -1db peak at 0db after mixdown, and -3db peaks at -1.2db… It’s hard to believe that compression artifacts would add 1.8db of peak, is this just the nature of mp3 encoding? It’s also strange to me then that it never seems to surpass 0db.

Yes, it’s normal to add that much and even more and it’s out of our control. I quit exporting to MP3 from Cubase a long, long time ago. Besides iZotope RX I didn’t find any other mp3 converter that would limit peaks.

I see, I’ll go ahead and stick with wav and use RX for mp3’s. Thanks

1 Like

What you experience is Dithering.

how it could sound in those frequencies that lack bit loss. but it’s difficult cubase has UV22HR, it’s an extra noise that is added when downgrading to mp3. and also what the others say happens, but the audio file. the low noise added to compensate is the DB you see hiding distortion. So dithering compensates the audio film but noise to hide distortion all the “bits” that are lost when you go from say 24 bits to 12 bits, then there is a halving of the data loss in the audio file, there are only 12 bits left to read from and not 24 .
Then consider all the data loss that occurs when converting wav to mp3. it’s like the most realistic pc game you play for minecraft. Cubase has UV22HR fum works well. but isotope I know is sharper and has more formats. but the file is effing messed up when it’s mp3, I’ve tried to fix to mp3 to all the material that wav has again, through RX10 adv where you synthesize the sound that doesn’t exist. with rx making prediction on the sound wave and drawing in innovative sound waves. but rx does a good job there too. therefore you have to adjust the må3 file after a format change for the worse. but cubase has all the tools to make it sound good. it’s mp3 file :slight_smile:

Read aboutDithering and how to fix it and what happen in the world of 0 and 1:)

Here is a good video

And wavelab have izotope MBIT+ Dithering

MBIT+™ Dithering.

I have this on i track i put out, hate it. Will se if i put it back and fix it, coz the artifacts is most notable in the beginning.

I appreciate the explanation, but I experience this issue both with and without dithering applied. My knowledge is that mp3 does not have bit depth as uncompressed formats do, so dither is not necessary.

Ditherimg is just a layer noise when downsamplimg to mp3 if you go from 24bit or 16 bit 1 bit is 6dB to a mp3 with lower And alot of material is in the upper range coz of the waves a alot smaler and therfor more that goes away, the only thing dithering is doing is to try to even out the wider steps that result when you remove as much material as you do from 24 bit and 16 bit. this can be clearly seen in RX. in many cases you don’t need dithering because it can destroy more than it does good. I downsample in RX and check if any dithering is needed and if there are many data error points that are offsets that need to be fixed before and after downsampling as there can be some distortion with and without dithering. after that I synthesize the mp3 file from the RX resymtj module instead of dithering and only fix data error points in the audio file before. so in my opinion, to get a perfect mp3, it is so low or low dithering where you have control over data error points and fix them, to finally soften the file by synthesizing the sound wave so that it corrects itself and solves distortion and other effects. then I run the world’s best in my opinion clipper RX declipper mpdul to take and round off the peaks so that they are nice and without distortion peaks. and then it sounds to me at least that you got back quite a lot of stability and high frequencies, as well as some low ones.

if you have hand techniques to check mp3 files and fix them for best results and stability. so please write down how you do it, as I will probably overwork it.

Some pic to make it more clear how i think

Sample rate x bit depth x the number of channels = Bit rate

44.1kHz x 16 bit x 2 (Stereo=left and right channel) = 1411.2 kilobits per second.

As you can see, the bit rate of a CD quality audio file is 1411.2 kbps.

So to get the cd quality i need 1411.2 kbps but meny streaming sites put down it to 128 kbps 256 kbps or 320kbps. And it demens and on top of that they compress it once more and optimize air and peak levels for their services without caring about defects that may occur. I’ve had that problem at some point where you can clearly hear the defects when it’s up on the site.

so the question is also what is the most optimal bit pch hz and downsampling to get an mp3 file that is as standard as possible after downsampling.

what do you others say and how do you get the best mp3 files possible.

And for 64bit processing. Men og izotope can handel that, only some stockplugims. I only use 64bit then recording my hardware. 32 bit float when mixin crazy headroom, and have never Max that out hehe.

hope you understand what I mean and that I express myself correctly with the problem and how can solve the problem from mixdown to finished product in mp3 format.

otherwise you can tell sp I’ll see if I can explain more clearly

I can se a Gais to have it on somtimes, to get more material in to mp3 om the high end of the spektrum it gets more rounded sound, everything is noize upp there anyway hehe and its easy to hear defects coz of data loos and more smaler waves that splits.

What is the teoretical holy grail and WBS for this problem. ”WBS= working brakedown structor”