오디오 품질 향상을 위한 마스터링 팁

### 초보 프로듀서 및 믹서를 위한 마스터링 팁
* 🎚 리미터 상한을 **-.2로 설정**하고 **8배 오버샘플링**하여 샘플 간 피크를 방지하고 저급 DA 컨버터의 왜곡을 방지하세요.
* 🔊 "과도 상태에 영향을 줄 수 있으므로 일반 EQ 애플리케이션의 경우 **선형 위상 EQ를 끄십시오.**"
* 🎶 "스테레오 필드 관계를 유지하기 위해 **미드 사이드 프로세싱에 선형 위상 EQ**를 사용합니다."
* 🎛 "클리퍼, 왜곡, 포화 상태의 경우 **오버샘플링을 활성화**하여 에일리어싱 왜곡을 방지합니다."
* 🎛 더 낮은 비트 심도로 내보낼 때는 항상 **디더를 켜서 잘림 왜곡을 방지**하세요.
* 🎛 일반적으로 신호에 불필요한 필터와 프로세스를 추가하지 않으려면 **자료의 기본 샘플 레이트로 작업**하는 것이 좋습니다.

항상 제한 상한을 -.2로 설정하고, 선형 위상 EQ를 끄고, 왜곡을 위해 오버샘플링을 하고, 24비트 세션에 디더를 사용합니다.

here are five mastering settings that new producers and mixes. And if you're doing any of these, simply changing them is going to improve your Masters not tenfold, but they're subtle improvements and they improvements that are free. You just have to click a button or move a thing, and it's done. You don't have to do any work for them. So let's get on with it. The first one, widely debated, what should you set your limit of ceiling to? I'm going to tell you, and then I'm going to show you and give you the whole scientific explanation. Should it be -1? No. Should it be -2? No. Plain and simple, a Time over sampling with a.2 ceiling. Now why specifically this? Let's get into the science of it. Okay, so in you've probably noticed if you've had your ceiling at zero, sometimes on your meters it reads over zero. Why is that? Well, because digital waveforms are discrete time have discrete sample bound time signals. And if we look over here, that's each of these sample points so

when your limit is set to here let's say where's my little marker here. When you limit a set to there, you're limiting that sample. But what you're not accounting for is this part over the top, which is the inter-sample peak. And that's the reconstructed waveform, and that's in discrete. That's continuously recalculated based on where the sample points lie. So then you get over-rates. This is a conundrum. We don't want that. We want to stay below the ceiling because then when we go out into the real world and people play it on their CD player or in their car or whatever system they have, their DA converter might not be as advanced as the ones we have in the studio and might actually not take favorably to these inter-sample peaks, and you get distortions. That's not good. Okay, moving a bit fast, but what are we meant to do here? So the first thing is if you look at ITB s177 um, this is all this is a

documentation by the International Telecommunications Union, and it documents exactly how to take or how to um meter the true peak value. And in this, it actually shows how much the M, the maximum underr could be. So, at four times oversampling, you got a half a dbel discrepancy. At 8 times, a 136 discrepancy. That's why with this here, I said do 8 times oversampling and a ceiling of 2. Okay, that's just as a general sort of thing because if you're doing 8 times oversampling on a 48 kHz session, um, you're going to have a a read error of .16 or .13. And then if your ceiling is at .2, you're never going to go over that zero ceiling. Perfect, that's solved. So what I'll do is I'll quickly show you this in action. So we've got the limiting going on here, super heavy limiting, and oversampling is off. We got the ceiling at .2, but we're going to get overreads by doing this. So there are overreads there plus4.

plus3 now we'll turn the over sampling on eight times over sampling, and what that's going to do is we've got oversampling that might hit zero at some point, but it's not going to go over zero. The next thing is turn off linear phase EQ. There is a use case for linear phase EQ, and I'm going to tell you in a second, but just turn it off as a general rule. All right, take a listen. This is natural phase, so it's an infinite impulse response, um, so everything that happens with the EQ filter is going to happen after the initial transient or that impulse. So have a listen to this. Kick Drum rings out, but that's part of the actual tone of the kick. Now listen to what happens before for the kick when I put it into linear phase. You see that little, that little, um, that little ramp up. That's a finite impulse response, and that's the linear EQ compensating for that phase relationship of that EQ boost. Now you're going to do a 16 dB boost at 62 Hz.

probably not, but the reality years once you start carving your track and actually EQing and manipulating it. Don't use linear phase EQ because all of those subtle little amounts and shifts with a finite impulse response that linear EQ needs to operate is going to murk your transience simple. So if you're using a left-right stereo EQ, don't use linear phase mode, turn it off. Now I did say at the start of this, there is a use case for linear phase EQ, and the reason why we want to use linear phase on mid-side EQs is the way mid-side is encoded and decoded. So you have your left plus right, which equals your mid, and your left minus right, which equals your side, okay, and that's the phase relationship. So left plus right positive, so everything that is similar. Left minus right, so left minus right means right phase inverted, gives you your side, and then to come back and you do the opposite and you get your left and right recoded or decoded back in a stereo. Now if your phase relationship on the side signal is

Shifting because of your EQ, it's going to change the stereo field. So simple, no linear phase EQ for general EQing application. But if you're using midside EQ, use linear phase mode so you don't warp or ruin your stereo field relationship. The next one is if you have a Clipper, if you have any distortion or harmonic saturators, make sure they have oversampling or at least run some tests to make sure you're not getting aliasing distortion back into your signal. Now, that style of distortion, I'll show you what happens. So, we got this signal here, it's playing a high-frequency sine wave. Now, we're going to add a Clipper. Oh, wait a second, what are all these other frequencies popping up? Well, those are the harmonics from the Clipper that don't fit into the sample rate because this is at 48 kHz. And the harmonics are above 48 kHz. It folds them down, and those frequencies are not harmonic. They sound really weird. So, what do we do well

we expand the bandwidth of the sample rate which this Clips. So if we go two times, we remove some of that alizing. Four times, you get rid of enough of it that it's inaudible. And then as you go higher, it gets rid of more of it. So simple as this, do run this test. Run a signal generator, get your harmonic saturators in, add a high frequency, and then clip distort. Add just add some saturation and see how well this works. Now you wouldn't want any noise above about 80 because then you're starting to get into territory where you might start to hear it. Simple as that, a small little setting over sample Clippers and saturators. The next one is always have dither turned on. All right, if you're running a 24-bit session, practically it might go down to 16 bit if you need to export it for CD or CD baby. Always have dither on and I'll explain to you why. All right, let's just play a 24-bit file then a 16-bit file with no

dither, and then a 16-bit file with dither, and then I'll actually tell you exactly what it's doing. Okay, so what you're hearing there is actually truncation distortion. Truncation is the process of discarding the lowest bits when low bit dep. So you got 24 bits, and then you go down to 16, and what's going to end up happening is numbers are rounded to the lowest value, which isn't a big issue with loud signals. So everything in the that's super loud in the top part of your signal isn't going to be affected. It's the very lowest bits that get discarded because of rounding. So when you're rounding from, let's say, 200 point .9, which might be a particular amplitude signal, down to 200, that's a read error of about .5%. You're not going to see that at all. However, when the lowest bits truncate, let's say your value at 24 bits is 1.9, and it has to go to a whole

number of one that's a read error of 90%. And that's what ends up causing that truncation. Distortion. So the way dither works is it adds a little bit of noise to the very lowest bit of information, and what that noise does, it actually randomizes those amplitude values so that way it's. Smooths out the tail. So by having that noise in there, it actually. Smooths out the tail, so you don't get that crackling. Distortion. So simple add dither, okay, it's not going to make the rest of your master sound great, but those little tals, that reverbs at the end of things when you're going down a 16bit, people will hear it, and you don't want people to hear that, that sounds no good. Now number five. I don't know if I'm right or wrong on this because this is something I practice, but then I did the tests and the tests didn't disprove me, but they didn't approve my hypothesis or the way I work, and that is the fact that you

should always work at the sample rate of the material or the session's material. So if you produce something in 441 or 48, guess what? Master it in 441 or 48. And the reason or my theory behind it. And then I'll show you the tests because the tests don't necessarily approve my theory. The theory is that every time you upsample or downsample, resamplers use all-pass and high-pass filters in order to add new sample point values in. Okay, so you're adding a low-pass or an all-pass depending on whether you're going up or down in sample rate. And that's not necessarily desirable. You don't want to add more filters. You don't want to add more processes to your signal. So I was always of the nature, and I still am of the nature, because I still got to look into this further for you guys, but as always, the nature work at the native sample rate. Tests are so important. So I'll show you something fun I did because this is how I

spend my time is I got a signal. Okay, so look at these files up here, and then I oversampled it to 96. Then I took that 96 and two times oversampled it, and then I oversampled it back to 96, and then back out at 441, which is the native uh sample rate of that original signal to null test. Okay, and I did the same thing with the shifter sync wave, which is just that sync wave that's at one impulse. Okay, now with the impulse when I did this and I went up to 96 and then back down and all around, there was a 50 DB difference in the signals, which to me is you can't hear it in here, but 50 DB of a constant signal beneath the noise floor is audible in most mixes and Mark. So you can see here the peak value is -50 n. Okay, so that's the original, and then that's the oversamp, that's the upsampled, oversampled back down and out version. So they're both. So I've got a -50 under which I

wouldn't want to be upsampling anything or adding extra processes and resampling. Because of this, however, I thought, That's great, this is proving my theory. Let's get a full mix and do the exact same thing, because if it's like this on a single impulse, it's going to be worse across a whole track. However, guess what happened. It was actually better on this material to the point where it was so low in the noise floor, I think it's negligible. So look at this. So you've got your normal 441 version, and then you've got your 441 that's been upsampled to 96 um two times over sample. Then it's went back to 96 and then out at 441. I wanted to add as many processes in there to replicate what you do in a session if you upsampled it and then added a processor with oversampling, and then you had to export it. Look at this. I'll test see that it's at -7680 even in the

loudest bits here. So I'm that through and that's what I'm getting. Those are the readings I'm getting to me, to me, and I have to be honest with when I give a recommendation that is negligible so you can actually run your sessions at a higher or lower sample rate than what you originally produced them in. My advice would be to keep it native because I'm still not conclusive on this and this was one of this is going to be my first piece of advice but it ended up becoming confusing to even me when I was setting up this video. But basically, you shouldn't be too stressed about this right now. I got to look into it a little bit more to me adding any difference of noise or difference in the filtering or representation of the mix that you're getting can be problematic so I try and avoid it so you should try and avoid it maybe where you do not work at higher sample rates than what the mix is at however I got to do a little bit more investigating or if you know something.

let me know in the comment section below because this test even if it's loudest it's -60 and that gets that's when it starts to get into the realm of could be audible. But even then, that's such a slight difference that nobody's going to pick it up on an A. So that final way you might be up your Masters might not actually be up your Masters. How's that for an end to the video? Uh, I wanted to show you that because I'm not always right, and everything is always in constant evolution, and I'm always trying to improve myself. So when I come across something like this that doesn't approve my thoughts and doesn't disprove it, it just adds a little question mark where I go, Hm, maybe I can oversample all my sessions and work at a high sample rate. What's the benefit of that? Well, that's another discussion for another video, but I'm going to leave you there because I'm pretty puffed out and yeah going to get back to some sessions. So until next time, take care.

5 Mastering Settings That F*ck New Producers & Mixing Engineers

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

오디오 마스터링에서 제한 상한 설정이 미치는 영향

샘플 간 피크와 왜곡을 방지하려면 8배 오버샘플링으로 제한 상한을 -0.

이 설정은 국제전기통신연합의 문서를 준수하며, 최대 언더 불일치를 보장하고 오디오가 0 상한을 초과하지 않도록 방지합니다.

오디오 프로덕션에서 EQ 및 오버샘플링을 사용하기 위한 팁

신호가 0을 넘지 않도록 8배 오버샘플링을 켜고 일반적인 이퀄라이징 애플리케이션에는 선형 위상 EQ를 사용하지 마세요.

선형 위상 EQ는 스테레오 필드 관계를 유지하기 위해 미드 사이드 EQ에 더 적합합니다.

왜곡 또는 고조파 포화기를 사용할 때는 원치 않는 주파수를 생성할 수 있는 신호의 에일리어싱 왜곡을 방지하기 위해 오버샘플링이 있는지 확인하세요.

샘플 레이트 및 디더링을 통한 오디오 품질 극대화

샘플 레이트를 확장하면 오디오의 에일리어싱을 줄이는 데 도움이 될 수 있으며, 레이트를 높이면 들리지 않게 됩니다.

디더링은 24비트에서 16비트 오디오 파일로 변환할 때 잘림 왜곡과 오디오 품질 손실을 방지하기 때문에 매우 중요합니다.

잘림 왜곡은 주로 비트 심도를 줄일 때 오디오 신호의 가장 낮은 비트에 영향을 미치므로 오디오의 조용한 부분에 큰 영향을 줄 수 있습니다.

오디오 마스터링에서 디더와 샘플 레이트의 중요성

디더링은 오디오의 가장 낮은 비트에 노이즈를 추가하여 사운드를 매끄럽게 하고 왜곡을 줄입니다.

신호에 추가 필터와 프로세스를 추가하지 않으려면 기본 샘플 레이트에서 작업하는 것이 좋습니다.

발표자의 이론에 따르면 업샘플링 또는 다운샘플링은 원치 않는 필터를 도입하지만 이 주제에 대한 추가 연구가 필요하다는 것을 인정합니다.

오버샘플링이 신호 품질에 미치는 영향

신호를 96에서 오버샘플링한 다음 다시 441로 오버샘플링한 결과 50dB의 차이가 발생하여 노이즈 플로어 아래에서 가청 효과가 있음을 나타냅니다.

그러나 전체 믹스에 적용했을 때는 그 영향이 미미했으며, 신호의 노이즈 플로어가 -7680dB로 현저히 낮아졌습니다.

이는 오버샘플링 효과가 오디오 자료의 유형에 따라 달라질 수 있음을 시사합니다.

샘플 레이트가 오디오 품질에 미치는 영향

원본 프로덕션보다 높거나 낮은 샘플 레이트로 세션을 실행하면 노이즈 및 필터링 문제가 발생할 수 있습니다.

샘플 레이트의 약간의 차이는 오디오 품질에 큰 영향을 미치지 않을 수 있으며, 그 차이는 약 -60dB에서 들리게 됩니다.

모든 세션을 더 높은 샘플 레이트로 업샘플링하기로 결정하기 전에 이점을 신중하게 고려하는 것이 중요합니다.