Lossy Codec Testing: Impact of 0 dB Peak Levels

Hard clipping at 0 dB affects lossy codecs; 96kbps reveals minimal impact, supporting the safety of 0 dB true peak masters.

### Testing the Viability of 0 dB Full Scale Ceiling for Lossy Codec Platforms
* 🎚 Testing the viability of a **0 dB full scale ceiling** for masters played back on platforms with **lossy codecs**.
* 🔊 Lossy codecs can introduce clipping when peaks exceed 0 dB, emulating a **bad digital-to-analog converter that hard clips at zero**.
* 🎧 Blind A/B tests were conducted to discern between the **clipped and raw versions** of a master with **peaks exceeding 0 dB**.
* 🔈 The amount of clipping introduced by a 0 dB full scale ceiling was found to be **-60 dB**, which is a **minimal percentage of the total dynamic range**.
* 📉 Lowering the ceiling level to **-1 dB** was considered, but based on the test results, a **0 dB full scale ceiling** was deemed safe.
* 📣 The inability to discern the clipped version in the blind A/B tests led to the conclusion that **0 dB is a viable ceiling level**.
* 🧠 "If evidence comes to light that contradicts this conclusion, or if new knowledge emerges, further tests will be conducted to inform the **mastering process**."

I set my limited ceiling to 0 dB full scale, but platforms with lossy codecs. I play it back with peaks over 0 dB. Now we're going to check what this sounds like by using this testing method. You'll notice I've got my original master here, a codec auditioner, and a hard clipper. And if this master here can't pass some tests, I'm going to have to change a few things. We're not going to be doing 0 dB true peak masters anymore. We're going to figure it out. We're going to debunk this once and for all to figure out which peak level is actually best. But something I know is definitely the best is my audio newsletter. Okay, if you're a pro in the industry, you're going to want to be on it. Another 200 people have given it five-star reviews, so make sure you join. There's a link in the description. Now let's jump into this. This track here is INX Cuff You Up. It is an awesome master. I got to do Bob Horn.

mixed it. It sounds great. 0 dB true Peak. Just listen to this. We'll turn off the codec just so we can have some fun. You know, you guys know I love the tunes I get to work on, and this is a really cool [Music] one. It's a cool tune. It's a really, really cool tune. Now, what have I got set up here? Let's go over to the codec. We're going to be going into this codec 96 kilobits per second that goes out of send into two channels. One is raw with the phase inverted, the other one has a hard Clipper on it. Why a hard Clipper, would you say? So here's the thing. When you've got these Peak values going over 0 dB, because we've got lossy codex, they have to leave the computer through a digital analog converter, and that digital analog converter in the modern era, good digital analog analog converters can handle these overs with them buffer, or they set that they attenuate that rate slightly so the analog waveform isn't click so it

might be a decibel pad, a one debel pad, or whatnot. Um, however, cheap converters don't bother to deal with that sort of arithmetic. They will just say, Oh, it's zero dB, it's gone over, we're hard clipping that, and that doesn't sound good on show systems. Now, as a mastering engineer, it's our job to make sure it sounds good on as many systems as possible, so we're going to the lowest common denominator, 96 kilobits per second. The Peaks will be hard CLI. This will emulate a really bad converter that just Clips everything at zero. However, we need to also contextualize this. Spotify, if you're a free Spotify user, you're using 128 kilobits per second on your web app, or 256 if you're paying, and then you've got different bit rates based on desktop, mobile, and tablet. So, not everybody's going to hear the worst version, but we're going to work with the worst version, try and get the worst version sounding as good as possible.

that's going to carry only forward to being better at higher bit rates. Now let's run this, and I'll test to hear what that clipping sounds like. Okay, what clipping is being introduced. So we've got this phase inverted, beautiful. The hard Clipper here, the lossy codec is on. Let's take a listen. Now I'm trying to dance so I can't. It's just clipping. Um, that's not good. That hard ing, how loud is that hard clipping is another question. Then we'll contextualize this and do some blind ABS to see if we can hear it before tweaking or potentially changing um our actual peak level. So at the most it hits -60 DB. Interesting, that's good to know because the hard clipping isn't the worst thing about this whole process. If you're actually in a 96 kilobit per second um lossy format, you're getting much worse than that hard clipping if it's going over zero. You can solo the artifacts here and listen to the difference. What information is actually

being taken away, and it's much louder than -60 DB below full scale. So the idea that all that information is being taken away when you go to 996 kilobits per second is much more damaging to the audio than the hard clipping itself. That isn't an excuse for it, that's just an observation. Now what's important is that we do a blind AB test. So what we got here is H for blind AB, and we're going to solo between the raw and the clips. We're not going to have any meters there to show us which one's playing. We're just going to jumble it, show the track, and then guess which one. We're going to do it 10 times and see if we can pick out the difference between the clip version and the raw version. Okay, let's [Music] go. It should be a crown me feel you should do a tongue me feeling like this. Guess I got a lucky. I'm going to loop this kick because I know if I can get this kick.

right that's where I'll be able to pick it up. [Music] that top one is the clipped one. Boom that's one out of one let's do it. [Music] again that top one is the clipped one. Oh no it's not okay 50% let's do it again. [Music] I think the top ones that clipped. [Music] one I got to you down you it should be a crown got me feeling like this my mind got me feeling like this. I think I'm not sure which one's clipped on is the bottom on the clipped one. No I'm getting it wrong okay so only one out of three let's go for see if we can pick it out even more let's go. [Music] again down you it should be a crown me feeling like this you should do a me feeling like this let's see let's have another listen C you C you C you C you C you C. [Music] you I'm going to say the solo at the

bottom is the Raw. One, the clip on. I'm getting it wrong every time. Oh, I'm getting it wrong three times out of four. I'm really struggling to pick out which one is the clipped one and which one isn't. Is the Clipper still on? Yeah, the Clipper's on. So, I'm in a bit of struggle town. [Music] Here, I got to check something. One moment, let's have another look. Is my codec on? It is on H. I. I can't pick out the difference at the moment. I really want to test myself, see if I can. I got to luy down. C you up. It should be a crown. Got me feeling like this and then P. My Mind. [Music] Come on, let's get this. [Music] Right, that bottom one is the clicked one. That's two out of five. That's um, you know what, I can't say that. I I I know what's going on there. Um, which means how negligible is that clipping 'cause remember like I said, I did that null test, that clipping is 60 DBS.

below as in the amount of harmonics it's introducing. The way it's changing the signal is 60 DB below the loudest point with full scale. So, the artifacts introducing are very low. It's very far below the total available dynamic range. And actually, funnily enough, Dan W just recently put out a video on decibels, and he did a great great deal explaining the logarithmic value of what we see here when we're metering. So, -60 DB isn't linear scale. 60 DBS below full scale. If we actually use this decibel converter here and we go6 -60 DB, it is 0.01. So, it's a tenth of a percent of the available amount of dynamic room there is because it's logarithmic. Okay, that's just a converter he used in his video and was a great explanation. So, the amount of signal being introduced is only one tenth of the available amplitude in that signal. I hope I'm converting that well that well that that train of logic. So, 0o DB in theory even

with the hard Clipper chopping all those tops off. Okay, you can't, I couldn't pick out if you guys picked it out. That is great. That said, I want to see if I want this test to come out with something else. I want you guys to see in a null test. So we're going to take get rid of the blind tests here at what point or how much of that clipping disappears based on the ceiling level. Cuz I got this test set up here already. The null test is there. So what we're going to do is we're going to bring this output level down at 96 kilobits per second codec auditioning and have a look and monitor what 0 dBel ceiling does what .3 what5 what1 introduces. So that way we have an idea of what different ceiling amounts actually introduce on clipping when it's lossy converted. That's .3. It's not a big difference between 0 and 3. .5 instead of being up at -60 we got 80.

Debs there of artifacts being introduced -11 it's negative one is actually a pretty safe bet there's very little very I couldn't pick it out when it was -60 if it's all the way down here at80 100 remember this is the loudest part of the song um I think people have a very difficult time picking out that clipping is an issue so what does this mean I honestly thought these tests were going to lead me into changing my ceiling level going from negative 0 DBS down to negative 1 and just having negative 1 is my only deliverable so I don't have to have two separate deliverables one for distribution and another one if they want to service Apple digital Masters zero DB even with those overs at lossy formats and remember we were auditioning 96 kilobits per second so as low as it goes for um for common streaming platforms with their codecs I couldn't a be it is is that all me guys you need to tell me Nick it is

your fault. You couldn't hear the difference. We heard all six differences and got 100% on our AB tests. Cuz if you did, high five to you. I couldn't pick it out. Maybe on a different system I might have picked it out. I don't know, but that's the nature of this. This is why I wanted to do this test. Set it up. I have to have a little bit of humility here to say I couldn't pick it up in the blind ABs. I hope you enjoyed this video cuz this was just something I wanted to do cuz I heard somebody comment they said one of my masters was going over zero and I said no, that's cool because I have no issues with it. Um, something that I have done in sessions before and I do do in sessions, especially on loud EDM tracks, is I do run this codec here, okay? I do run this codec into a hard Clipper. Okay, that's an easy way to know where and how much of that clipping is being introduced cuz this isn't a very

a loud song like at its loudest it might be 6 LS. But in some like trap tunes, you can get to like 4 neg 3 LS. And the overs can be way more than a decel. It can be like 2-3 DBS over at 96. But yeah, there you have it. That was interesting. That was that was that an interesting video. Maybe not the exciting bang at the end to say we debunked and we're going to change our ceiling level. But um, negative 0 is perfectly safe in my opinion. At least if you've got a different opinion, let me know. And uh, yeah, let's run with that for now. And if I change my mind or not change my mind if evidence comes to light and I can make a decision that is otherwise to this and prove it to you guys or you guys can prove it to me and I can share it with you and we can do more tests and we can get a better result that helps inform us with new knowledge. We're definitely going to do that, okay? We will.

Debunking peak levels in mastering…

Hello. I'm Nicholas Di Lorenzo. I'm a mixing and mastering engineer and owner of Panorama Mastering Studio in Melbourne, Australia. I’m an Italian-Australian. born and raised in Melbourne.

This channel is for anyone involved in the music making process, artists, engineers, producers or musicians.

To stay up to date, hit the "bell" icon for alerts as I release videos weekly on all things audio, mixing, mastering and industry.

-

Want to learn more about what I do?
https://panoramamastering.com.au 
@Panorama_Mastering

Panorama Mixing & Mastering

Setting a 0 dB true peak may cause issues with lossy codecs, leading to hard clipping during playback.

The goal is to find the best peak level that sounds good on all systems, especially those with poor converters.

Different bit rates on platforms like Spotify also affect how the music is heard, so it's important to optimize for lower quality playback.

Analyzing the Impact of Clipping in Audio Production

Clipping in audio production can be more damaging in a 96kbps lossy format than just causing hard clipping.

The loss of information in the audio due to the low bitrate is highlighted as more detrimental than the clipping itself.

A blind AB test is conducted to try and differentiate between the raw and clipped versions, revealing the challenge of identifying the clipped audio in the production.

The speaker is having difficulty distinguishing between the clipped and non-clipped audio.

They mention that the clipping is minimal, at 60 DB below the loudest point, and that the introduction of the signal is only one-tenth of the available amplitude.

The speaker plans to conduct a null test to observe how much of the clipping disappears based on the ceiling level set at 0, 0.

The Challenge of Identifying Clipping in Audio Production

It can be difficult for people to identify clipping in audio, even in the loudest parts of a song.

The speaker conducted tests to try to identify the issue and expressed humility when unable to do so.

In audio production, using a hard clipper can help identify and manage clipping introduced by loud tracks.

Different types of music have varying sound levels, with trap tunes reaching as high as 4 neg 3 LS.

The video discusses the safety of sound levels, suggesting that negative 0 is safe, but encourages remaining open to new evidence.

The speaker commits to further tests and research to gain a better understanding of sound level safety.