Mastering Tips for Improved Audio Quality

Always set the limit ceiling to -.2, turn off linear phase EQ, ensure oversampling for distortion, and use dither for 24-bit sessions.

### Mastering Tips for New Producers and Mixers
* 🎚 Set your limiter ceiling to **-.2 with 8 times oversampling** to avoid inter-sample peaks and prevent distortions on less advanced DA converters.
* 🔊 "Turn off linear phase EQ for general EQing applications as it can **affect transients**."
* 🎶 "Use **linear phase EQ** for mid-side processing to **maintain stereo field relationships**."
* 🎛 "For clippers, distortions, and saturators, **enable oversampling to prevent aliasing distortions**."
* 🎛 Always turn on **dither when exporting to lower bit depths** to avoid **truncation distortion**.
* 🎛 Working at the **native sample rate** of the material is generally recommended to avoid adding **unnecessary filters and processes** to the signal.

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

The Impact of Setting the Limiting Ceiling in Audio Mastering

2 with 8 times oversampling is recommended to avoid inter-sample peaks and distortions.

This setting complies with the documentation by the International Telecommunications Union, ensuring a maximum underr discrepancy and preventing the audio from exceeding the zero ceiling.

Tips for Using EQ and Oversampling in Audio Production

Turn on eight times oversampling to prevent the signal from going over zero, and avoid using linear phase EQ for general EQing applications.

Linear phase EQ is more suitable for mid-side EQs to maintain the stereo field relationship.

When using distortion or harmonic saturators, ensure they have oversampling to avoid aliasing distortion in the signal, which can create unwanted frequencies.

Maximizing Audio Quality with Sample Rate and Dithering

Expanding the sample rate can help reduce aliasing in audio, with higher rates making it inaudible.

Dithering is crucial when converting from a 24-bit to a 16-bit audio file, as it prevents truncation distortion and loss of audio quality.

Truncation distortion mainly affects the lowest bits of the audio signal when reducing the bit depth, which can significantly impact quieter parts of the audio.

The Importance of Dither and Sample Rate in Audio Mastering

Dithering adds noise to the lowest bit of audio, smoothing out the sound and reducing distortion.

Working at the native sample rate is preferable to avoid adding extra filters and processes to the signal.

The speaker's theory is that upsampling or downsampling introduces unwanted filters, but they acknowledge the need for further research on this topic.

The Impact of Oversampling on Signal Quality

Oversampling the signal at 96 and then back to 441 resulted in a 50dB difference, indicating an audible effect below the noise floor.

However, when applied to a full mix, the impact was negligible, with the signal being significantly lower in the noise floor at -7680dB.

This suggests that the oversampling effects may vary depending on the type of audio material.

The Impact of Sample Rate on Audio Quality

Running sessions at a higher or lower sample rate than the original production can cause noise and filtering issues.

A slight difference in sample rate may not significantly affect the audio quality, with the difference becoming audible at around -60 dB.

It's important to carefully consider the benefits before deciding to upsample all sessions to a higher sample rate.