Simple mastering tips no one talks about... BUT THEY SHOULD!

[@naoredri4444]

I just wanted to say that you are truly the hero I have been waiting for. The level of details is insane, and stupidly high. Thank you! As a bedroom hobby producer, your content is incredibly valuable to me. You shed light on topics that are not often discussed or explored in depth. I've watched almost all of your videos, and they have all amazed me or taught me something new!

[@panorama_mastering]

My pleasure! Thank you for watching and enjoying!

[@a2bs333]

Exactly the info I needed today. I have two questions. 1) What is a good TP range for mp3 and streaming services encoding in your opinion? When mastering from an MP3 beat would I be overcooking by using a sandwich limiter technique Ozone TP (on) limiter, clipper at 16x Oversampling and final FF Pro L2 limiter at 16X Oversampling? Thank you in advance 🙏

[@shaneslack-multimedia9268]

1). True Peak really is not as focused on as most people think it is as it changes based on so many variables (SR, Analog Playback Medium), and is a byproduct of the taking a digital signal and bringing it back to analog. Usually a TP limiter tends to really thin out the master or not just sound that good so I know for a fact that most of the mastering engineers will not sweat over a delta of + 0.10dB over your ceiling setting. (Just pay attention and have good practices for your ceiling). If you actually audition and measured most of the original masters of songs you get on streaming services (as to avoid streaming compression/loudness normalization algorithms each may apply differently), you will notice TP reading all across the board even from the best masters and mixes. However it's up to you if you want to worry about TP. 2). Cascading Limiters/Dynamic Processors really depends on how hard you are hitting each limiter (How much GR is being applied) and the sound you're getting from either of them. If you are going to stay with keeping TP Limiting, make that the final limiter in the chain as even with oversampling limiters do add distortion and having a TP limiter before a non-TP limiter will ultimately just nullify and cause new over-peaks.

[@a2bs333]

@@shaneslack-multimedia9268 Oh wow thank you for that in depth reply. I was under the impression that some encoders can add up to 2dB to the peak after reading many articles and watching engeneering videos. I get that many masters will differ as not everyone cares about the normalisation applied and how it affects the sound. I am concerned about avoiding the streaming services' limiters. I only ever apply 2-2.5 dB of gain reduction on master bus clippers and limiters. I read that TP level is quite important in the context of avoiding unwanted clipping.

[@shaneslack-multimedia9268]

@@a2bs333 The problem with streaming service normalization/limiting and trying to avoid them is the fact that both no two streaming services even have the same protocol and that there has been extensive history of them changing their protocol(s) without any sort of public facing documentation thus leading to a bunch of trial-and-error experiments having to be done by engineers to find out those changes themselves - this was actually one of the points discussed in AES TD1008. I might not be understanding the question presented since in my experience and discussions it seemingly has been stated time and time again that this is something where it is better to master for what the song calls for in it's own right versus trying to match a specific streaming service's guideline or penalties based on LUFS/Peaking (especially since most services are starting their transition to offer loseless playbacks with even opting to turn off normalizations by default)

[@jensen777]

This could not come at a more perfect time. I just finished a mix and im taking a little break before mastering and this pops up!

[@panorama_mastering]

You got this!

[@Euterpe-rb7vd]

Very professional and well presented..thank you!

[@panorama_mastering]

You're very welcome!

[@leckel1996]

Wonderful content! I would like to make an argument for higher sample rate mixing/mastering though. One of the upsides of upsampling first is you can actually get away with turning oversampling off in most or all of your plugins. So you actually end up doing less harm to your music. Also, the oversampling filters will be up at 40kHz at 96kHz vs 20kHz at 44 or 48 which is at the threshold of our hearing.

[@everestwitman]

Seems legit. I have been on a rabbit-hole dive of trying to figure out if upsampling mastering sessions is the way to go, and I agree with you here. Might be worth performing an experiment, mastering the same song with the same processing one upsampled in the mastering session and then downsampled for delivery and the other kept at the original sample rate (but upsampled by plugin processing.)

[@soulofwaves]

That's a good point, but I'm actually having a little of headaches when working with samples. I mix in 48 khz and all the samples I own are in 44.1 and what I'm trying to do to reduce the impact of this redundant upsampling is to dither all the tracks with samples

[@paavoilves5416]

Dan Worrall has good content on sample rates, check it out

[@nj1255]

A side note to Tip 1: I don't like when plug-ins have forced oversampling/anti-aliasing without the option to disable it, since it adds unnecessary up- and down-sampling steps. If every single plug-in you use in a chain are made like this, the sample rate conversion steps quickly adds up. I rather use the FX-chain oversampling option in Reaper whenever I can, or simply don't oversample at all. I'm shocked that no other DAW has implemented this solution yet. It would save CPU and remove unnecessary signal degredation for the user, and also save a lot of work for plug-in developers (that work would be shifted over to the DAW developers, but I think it makes more sense to have oversampling as DAW feature rather than a plug-in feature). Justin and Schwa are the GOATs for implementing this in Reaper (and for many other reasons as well)!

[@panorama_mastering]

I get where you're coming from; however some plugins will oversample the side-chain signal (feeding the detection circuit) for a more accurate ballistics;

[@plastiktide]

Awesome channel! As an engineer who has mastered most of their own mixes for the last 15 years I always appreciate insights from a mastering specialist 🤘

[@elglieze]

ily man, always appreciate you sharing what is rarely talked about on yt 🙏🏼

[@lusid_music_uk]

I remember seeing someone test out this sample rate conversion theory in a video. (Forgot who it was now) But the test was basically sample rate converting a signal over 100 times then doing a null test. The result was the null was completely inaudible/ridiculously small difference. Key takeaway from it: even 100x sample rate conversion doesn’t degrade the signal in any meaningful way and can thus be ignored entirely. Spend your mental energy on more important things.

[@samphelps856]

Great work

[@n3gi540]

some of your videos are lit like this one 🔥🔥🔥

[@Fire-Toolz]

hey nicholas, i'm curious. i generally work in 48 because that's how i want my plug-ins to work. when i get a 44 mix, i assume that when i convert it to 48, i'm simply putting 44 quality in a 48 container. so my processing is in 48, and i'm using that 48 processing on the 44 audio. is this accurate? or are you saying that i am converting audio from 44 to 48 i am introducing low pass filters? and if i am, what are those filters? if something like 10hz is being filtered out, who cares right? does that *really* matter as far as phase? or is the low pass filter more harsh than that? i think i'm going to do a null test and see what i find.

[@tryingtotryistrying]

ehhh not sure why I'm replying to you cause I don't have your answer besides it can't hurt to do a null test! but also what I was going to say is my limited understanding is that 44->48 isn't like pouring a can of beer into a pint glass but breaking down the cans beer into its core ingredients then making it again but scaled up somehow for the pint. obviously don't trust me blindly on that. there is a site called SRC comparisons that show the aliasing different converters do on a 96 to 44 conversion- they are NOT built equally, so that's something to consider as well. like pull up rx audio, or r8brain compared to wavestudio 7 (whatever that is) and can see how using r8 to go back and forth isn't nearly gonna mess with it as much.

[@nana-ld4cr]

Thank you for the video! Do you always deliver 44.1 masters for streaming releases?

[@gazingwrmth]

Great Tips! One Question Though Where does one should put a clipper in the signal chain? Before Limiter usually?

[@memeswillneverdie]

ye usually, the idea is to shave off the peaks so your limiter isn't working as hard

[@ampersand64]

Usually before any dynamics processing. Peaks that are consistent in amplitude makes the compressors/limiters act more predictably. If the goal is to transparently increase loudness, clippers work well on short transients because they only affect higher frequencies. The more low frequencies that saturation affects, the more that the sum/difference harmonics creates interfere with the overall sound. This is why, for longer transients (containing the fill cycle of lower frequencies), limiters sound more transparent (since they act slower than clipping)

[@gazingwrmth]

Got it Guys! Thanks 🍻

[@GingerDrums]

Great video very honest and solid advice :). One stickler: Aliasing is pronounced "Aeli-ass-ing", not "aeli-eye-sing".

[@PharaohLawLess1]

Great video

[@jacquesancillon6635]

sorry but I hear a real difference between a project played and mixed at 48 . Then mastered at 192 or 96 ...the sound is clearer, and very more precise for a better adjustements !!!! I often work lower in samples because the needs of virtual instruments eat up a lot of cpu ( mac pro 12 cores mid 2012 ) . After stems creations, I can finalize with larger samples. Since I've been doing this, the results are much better, and the depth is much more real. What's more, I'm 64 years old and at my last audiometric test, to the amazement of the doctors, I can still hear up to 22-24 K Htz, and for an old guy that's not so bad Ok it's just my opinion and feeling , thank you for all the peaceful and constructive exchanges that i can read here, that's cool.

[@Fire-Toolz]

anyone got an easy way to implement the high pass side chain filter? i know some compressors come equipped with one. should we manufacture it with multiple instances of the song with linear phase crossovers?

[@johndoe_1984]

Why don't people auto-hide their dock

[@Cuebur-q1o]

😊

[@cholkymilkmirage4984]

who tf had those thoughts about changing the mix sample rate omg i cant believe thats a real thing

[@panorama_mastering]

It's a thing...

[@schallfarben5614]

Well.. not a ME here but I know enough to say you over simplified the os topic. By upsampling the whole project you use the os algo of your daw, wich In most times is quite good, by that you can getaway (depends on how hard you are processing) with no os on single plugins and by getting around this you don't need to use their internal os wich is often not as good or lets say there are tremendous differences from developer to developer. So it's maybe not the best to "just" use the sample rate of the project but of course the difference can be small, on the otherhand, those little details are what you should be about as ME. my 2 cents

[@panorama_mastering]

I definitely oversimplified it; but for good measure; I still firmly beleive that if you can avoid it all together you're better off; even sometimes I won't OS my limiter; I'll use TP mode which will OS the sidechain signal and keep the original in tact; because the limiter isn't working hard enough that the aliasing distortion is noticeable

[@schallfarben5614]

@@panorama_mastering don’t know how to even answer to this.. You probably don’t use os on your limiter because you use pro l. It’s not about „believing“ I made a lot of a/b to find out. Tpl is the worst. to keep it short. So yeah maybe you think you didn’t oversimplified. I am sure tho an me with some more experience will say you did.

[@panorama_mastering]

RE: TP Limiting watch this; kzbin.info/www/bejne/bWHFopehms9sgbc

[@schallfarben5614]

@@panorama_mastering you reached out for fabien which is really smart as when it comes to dsp he’s one of the most knowledged people I know of. Still you kinda missed a really important point to me there. And again I am no ME, i only produce and have a really good me mastering for me because I know he does it better for different kind of reasons. The os in pro L uses linear phase filter and you should be able to hear this. It’s smearing the transients.. I don’t even want to start about true peak. Damn I am becoming a KZbin commentator but please dig deeper if you „breath mastering“ and not want to be schooled by producers

[@ampersand64]

I don't see your logic when it comes to up+down sampling. You condone plugin instance resampling, but shy away from import/render ersampling? The resampling itself obviously doesn't cause any side effects, because any arbitrary number of samples will simply be converted to an analog signal before reaching your headphones/speakers. You claim that the serial lowpass filters somehow affect the quality. What exactly is the effect you're referring to? If you're talking about attenuation of frequencies at close to 20khz, then yeah, that's a problem. But our ear's frequency response rolls off similarly, so this is pretty transparent. If I'm running my mastering project at 96khz, it's so that I can specifically avoid using oversampling on any nonlinear plugins, of which I'm bound to have at least 2. This seems to be the best solution, because oversampling on plugins uses low pass filters anyway. The filters on some plugins' oversampling options can be too gentle to totally attenuate ultrasonic frequencies, or linear phase (which can change the transients subtly), or introduce crazy phase rotation. If the filtering artifact you're worried about is phase rotation, I have no clue why you would recommend plugin oversampling. The choice is between stacking steep minimum phase filters in succession every time you use oversampling, or running the project with enough room to hold higher harmonics, and filtering them out at the end, possibly using a linear phase filter. FFT filtering of ultrasonic frequencies would prevent phase rotation, and luckily audio resampling methods like r8brain use this to produce ultra steep, linear phase, minimally intrusive lowpassing. Also, while phase rotation across the spectrum can be noticeable if it's extreme (such as the case of using 3 or 4 high-order filters), it's not necessarily a negative artifact. It can subtly and unpredictably change which frequencies dominate transients, in my experience. If you're worried about ringing at the high frequencies, don't be. Filter ringing length is frequency dependent, so higher frequencies have shorter ringing. The only way to avoid ringing is to use gentler filters and set them lower (like 18khz), and seek filters that minimize how noticeable that ringing is, like inverse chebychev (with a gentler rolloff at the cutoff). Using just 1 filter to upsample upon import, then another, better filter to export to any desired format seems like *the* way to go. This avoids the phase rotation that can eat up headroom. What's more, since compressors and limiters will catch inter-sample peaks more accurately, and you don't have to worry that you're rendering out a clipped MP3. Either way, all this lowpass filter talk is splitting hairs. Just about everything else we do to audio is more noticeable than ultrasonic filtering. Aliasing is definitely more noticeable! Same with bitrate reduction and MP3 encoding, not to mention saturation from speakers and the noise floor of amplifiers. If you take the time to carefully listen to what your processes do, then make good decisions, there's no wrong answer. VSTs good for ultrasonic filtering: airwindows ultrasonic, fabfilter Pro-Q3, Engineer's Filters by RS-MET, and r8brain for resampling/MP3 encoding. Don't take this like bad-faith criticism, lemme know if I'm missing anything. I'm here to learn!

[@panorama_mastering]

I don't see your logic when it comes to up+down sampling. You condone plugin instance resampling, but shy away from import/render resampling? // Correct; this on a personal level; simply out of practice; reduce the amount of unnecessary processing; The resampling itself obviously doesn't cause any side effects, because any arbitrary number of samples will simply be converted to an analog signal before reaching your headphones/speakers. // Correct! in a single instance it won't! You claim that the serial lowpass filters somehow affect the quality. What exactly is the effect you're referring to? If you're talking about attenuation of frequencies at close to 20khz, then yeah, that's a problem. But our ear's frequency response rolls off similarly, so this is pretty transparent. // Correct! But compounded I'm happy to avoid it all together; If I'm running my mastering project at 96khz, it's so that I can specifically avoid using oversampling on any nonlinear plugins, of which I'm bound to have at least 2. This seems to be the best solution, because oversampling on plugins uses low pass filters anyway. The filters on some plugins' oversampling options can be too gentle to totally attenuate ultrasonic frequencies, or linear phase (which can change the transients subtly), or introduce crazy phase rotation. // I understand this completely; however; there are some limiters I opt to use that oversample the side-chain signal for let's say true peak limiting; and keep the original signal in-tact; I'm not oversampling unless I believe the aliasing distortion produced will be within 60dBFS of the incoming signal; that's usually only the case with clippers; or harmonic saturators specifically on the top end; If the filtering artifact you're worried about is phase rotation, I have no clue why you would recommend plugin oversampling. The choice is between stacking steep minimum phase filters in succession every time you use oversampling, or running the project with enough room to hold higher harmonics, and filtering them out at the end, possibly using a linear phase filter. FFT filtering of ultrasonic frequencies would prevent phase rotation, and luckily audio resampling methods like r8brain use this to produce ultra steep, linear phase, minimally intrusive lowpassing. Also, while phase rotation across the spectrum can be noticeable if it's extreme (such as the case of using 3 or 4 high-order filters), it's not necessarily a negative artifact. It can subtly and unpredictably change which frequencies dominate transients, in my experience. // This is a very good point; It's less about recommending the plugin oversampling but being purposeful with the decision to do so; Using just 1 filter to upsample upon import, then another, better filter to export to any desired format seems like the way to go. This avoids the phase rotation that can eat up headroom. What's more, since compressors and limiters will catch inter-sample peaks more accurately, and you don't have to worry that you're rendering out a clipped MP3. // please clarify on this point a bit more!? Either way, all this lowpass filter talk is splitting hairs. Just about everything else we do to audio is more noticeable than ultrasonic filtering. Aliasing is definitely more noticeable! Same with bitrate reduction and MP3 encoding, not to mention saturation from speakers and the noise floor of amplifiers. // Correct! Don't take this like bad-faith criticism, lemme know if I'm missing anything. I'm here to learn! // Definitely not bad faith criticism, this is why I MAKE these video's; I explore my curiosities; discuss my practices, then break them apart!

[@a2bs333]

Great comment I appreciate your insights and will be returning for your reply.

[@conrow1157]

Nobody Even hears this oversampling stuff..

[@Fire-Toolz]

if there's enough fold back you can definitely hear aliasing