Hi!
In Ozone7 tutorial, an iZotope guy said that if you're going to upload your song to Soundcloud, then while mastering you should set the mastering limiter/Maximizer's ceiling to -1dB or even -1.5dB (to prevent your song from clipping over 0dB after soundcloud compression).

Do you agree? Do I need to set the ceiling so low for soundcloud? I thought that -0.1dB or -0.3dB will be enough...

 

I've always done -0.3dB. Try it, if there is clipping, lower it. Experiment till you find what works for your track.

 

1db will sound better - more dynamic range [providing it isn't already pancaked] - If you really feel 0.7 will make all the difference then !

 

Macvideo talked about this in a webinar they did last week. They had 2 of the guys who created Ozone 7 on. One of them spoke about the extra processing these websites like soundcloud do on your uploaded file. In their testing they said this extra processing these sites do can introduce clipping that was never there in your final mix. So with their testing they figured -1.0 or -1.5 db is the safest way to avoid this.

 

i use -3.0 db

 

The limiter ceiling shouldn't be lowered. The limiter(Ozone) or anything else shouldn't be messed with at all. You don't have to alter the character of your master by changing limiter/compression thresholds within a plugin. If your master is naturally over -1.0dB you simply lower the global master output only so you can render a copy for Soundcloud that's not quite as loud.

 

Last edited: Dec 28, 2015

-0.5dB is for compressing to LOSSLESS, and even then it's not guaranteed to prevent clipping.

Low- bitrate encoding for web changes a lot so it will almost certainly clip after compression (if you limit to -0.3dB). -1dB is a MINIMUM. If you wanna check, try compressing your track to 128k mp3 OR redownloading it from soundcloud and checking the peaks on that,

Leading example above. In general, try -1.5 - -3dB for lossy instead. It's much less likely to clip.

 

first of all, use oversampling, inter-sample limiting/true-peak-detection, lookahead
ceiling for CDs/FLAC should be less than 0 dB...usually -0.3 dB
mp3 about -1 dB or even lower (-1.5 dB)
Some people say, SRC to 44.1 first, than limiter, than dithering with(out) noise shaping.
No need to use limiter at high sample-rates, because:
1) you have internal oversampling, often oversampling can be much more than your maximum affordable sample-rate;
2) you save nothing at high-rate (aliasing/distortions, for example), at different sample-rates behaviour is diferent. For example, 44.1kHz: you have distortion at 1.26 kHz and no distortions at 1 kHz; and 96kHz - you have distortions at 1 kHz and no distortions at 1.26 kHz. So, possibly less aliasing, but could be more distortions in audible area. Depends on everything.
3) Downsampling (even very super linear-phase) may cause smallest peak shifting. They are 0.x dB, small, but they there are. Minimum phase SRCs causes much more peak shifting.
4) Limiter and dithering are the final stage. So, SRC is better as pre-final, pre-limiter process.

 

-0.3dBFS here. Has never failed me in over a decade.

 

i think that's too mush lol, usually -0.1 will solve 90% of problems, -1dB is moore really more than enough to prevent clipping and iZotope has this option "True Peak Limiting" wish is already working great for me, with -0.0dB Ceilling...

 

The clipping comes from the compression artifacts, so it's not that dramatic (since it's basically "noise" anyway). -0.3 is ok.

No, the lower the ceiling, the less dynamic range is left.
But if you hear a difference between 0.0 and -0.5 (ceiling!), then you most likely have an overcompressed track...

 

you guys are crazy .!.

 

sometimes i just wanna my track to sound loud but not tha loud and i just put the threshold @-3dB and the whole track is at -2 to-4dB sometimes i overcompress not to the point that i can tell the diffrance with at least 3dB of dynamic range...

the point is it always depends, but letting the track breaths even if for a little is good, i don't wanna kill it with compression...

 

Interesting... I was always curious of what the "processing" included during upload. I figured it was converting to some sort of universal non-flash / flash HTML5 audio codec but never knew they actually added additional compression. That seems a little sacrilegious...

 

Thank you guys for your info!
Ok, so I'll be making shore that tracks I upload to soundcloud have ceiling not higher that -1dB.

I understand that it's better to upload a wav file to soundcloud than a mp3, right?
But is their a difference if I upload 32 bit wav, 24 bit wav or 16 bit wav? [16bit wav will have dithering done by me]

 

If you dither before you upload, the dithering will get destroyed , &/ it will sound worse with dithering because the lossy codec may try to compress the dither noise as well. Try 24/16bit wav WITHOUT dithering.

 

In terms of levels, there are no requirements that are special to soundcloud. No clipping, but that goes without saying for every application.
But there in fact is quite some stuff to keep in mind when mastering for soundcloud, so I wrote an article about it.
You can find it here:
http://www.auralnode.com/mastering-for-soundcloud/

 

Dithering for mp3:
In loud places it gets destroyed by compression, in quiet places/reverb tails it can be preserved. I saw my sonograms and there are no compressions in tails even with 224 kbps (lame 3.98 and higher), so quiet places and tails are dithered and no quantizations (I use MBIT+ noise shaping bby iZotope, that's probably why it is preserved(a lot of high-frequency power), don't know about dithering without noise shaping).

About 16/24/32 bit:
32 bit integer (fixed) has about 192 dB dynamic range (-192dB up to 0 dB), so it is better than 32 bit float.
32 bit float is same as 24 bit (-144 dB up to 0 dB), but audio data is preserved if you clip over 0 dB. Test: add a lot of gain on fader to get it clipped over 0 dB, export wav 32 bit float, open in DAW this wav and lower gain on fader - there is no clipping at all, it is clean. This thing doesn't work with 24 bit or 32 bit int and 16 bit.
24 bit - 144 dB as mentioned above. Minimum digital recording/digitizing/mixing/mastering standard. AD and DA converters, unfortunately, support 24 bit only for AD/DA processes, but dynamic range is usually up to about 120-130 dB (if we take the most professional audio interfaces, studio converters), not full 144 dB, I know only one ADC device (not DAC) that has dynamic range up to 153 dB (it is 32 bit), but don't remember the name (found on gearslutz forum).
16 bit - about 96 dB dynamic range. Without dithering: sounds more digital, quiet places dissapear (harsh crisping for most lowest parts), quantization noise and digital harmonics and other non-linear digital stuff are produced when you export non-dithered 16 bit. With dithering - small amount of maskin noise added to signal, sounds closer to analog, wider dynamic range (more than 96 dB!). Noise shaping for dithering makes noise less noticeable at mids area, but more noticeable at most high area and ultrasound + loudness increases.
For different sample-rates dithering with noise-shaping maybe different, so you have to dither after SRC downsampling to 44.1 kHz.

 

Last edited: Dec 29, 2015

No No NO, nice facts, but wrong interpretation. 98 db is more then sufficient in the final stage, quit places should not dissapear and harh clipping? dithering is actually audio fidelity taken to a extreme point.

If you like theory's that much think about this: 96B actually isn't a integer number, it can work with as much decimals as u want.
spread it out over an 20 hz- 20khz spectrum, again frequencies are not bound to integer numbers either.
so that is a hell of a lot possibilities, don't u think?

Good article, 40 Hz to about 16kHz now there is a pain

 

Last edited: Dec 29, 2015

I Usually export my track for soundcloud purpose -2dB, because soundcloud conversion ruins my track everytime. That is the best compromise for me.