True Peak limiters - do they REALLY sound worse ?
Вставка
- Опубліковано 31 бер 2023
- I hear lots of people saying they hate the sound of True Peak limiters, but is this really true ? Or are they testing them in a way that is misleading ?
For more about inter-sample peaks and when they can be audible, click here:
productionadvice.co.uk/inters... - Розваги
It was 10 years ago when I learnt so much from you. Thanks Ian.
My pleasure !
Awesome video! Thank you so much for all the information!
Thanks Ian for the really in-depth video of this topic, it has been puzzling me for a while now. I've heard many big-name mastering engineers talking about this subject - or not really talking but more like telling in super confident way - not to use TP at all and never really going in to details why, and only saying "it (TP) sounds worse or bad".
Pleasure - as I say in the video, TP limiting isn’t necessarily needed, especially if you use over sampling - but I do think many people are misjudging the way it sounds because of the way they’re testing
Thank you so much for this video and the others you make. You are awesome.
Thanks John !
Very helpful and interesting! Thanks!
Great ... When I get to the end of the video.. find that what I had discovered was the best way of keeping it clean and loud (using fader automation)...
In extreme cases also use clip gain to pull down sections.
🤓🤗Very clear and valuable information and comparisons! Now I can say that I understood True Peak processing better...I use FabFilter L2 but in some cases Bx-limiter True Peak....Thanks for the light
Pleasure !
Clear explanation. Thank you
Pleasure !
Hi Ian, I was one of the people who asked this question about true peak on one of your live streams. These examples are excellent. Thanks. Cleared a lot up for me.
Pleasure, glad it helped !
Loudness is fun and all, but i think people forget that there is a volume knob. However there is no "get rid of distortion" knob...
A valid point well made...
Haha thumbnail game is on point!
😂
Nice analysis, Ian. One variable I think you missed is the second, additional stage of Pro-L 2's true peak limiting, which p14 of the manual says "any true peak overshoot introduced by the ultra-fast limiting process itself, will be smoothly attenuated afterwards, ensuring the final true peak output doesn't exceed the current maximum true peak Output Level". It's very difficult to know how much of an audible effect this may have, but it's worth noting. Thanks for making the video!
I haven't looked into that specifically, but in this case the main distinction is whether TP limiting is on or off - Pro-L 2 definitely allows ISPs through when it's off. Or did I miss your point ?
@@ProductionAdvice Pro-L 2’s TP mode has two serial stages: 1) limiting triggered by true peaks (I assume via some kind of fancy oversampled sidechain?), then 2) “smooth attenuation” on any remaining TP overshoot (see manual p14). I wonder whether this “smooth attenuation” could be what some people don’t like the sound of - but I also don’t know how audible it could be (I suspect very minimally).
@@HawkReachGames OK I see. Yes possibly but honestly I think most people just enable it, hear more limiting (and don't like it) and never try it again. You'll be hearing that "smooth attenuation" in the TP examples in this video presumably, and in the LUFS-matched example I don't hear a problem, personally
Fantastic video Ian, thank you very much. One question about using automation on the ISL channel: wouldn’t that ~1dB increase during the first part of the song increase the integrated loudness and make the loudness normalization reduce it further down and end up being lower in level during the loud parts?
Slightly, yes ! But less than the heavily limited version so still a better option, IMO
@@ProductionAdvice Makes sense, thanks!
Hi Ian, thanks for the video.
I would have liked to hear about the impact of oversampling on reducing ISPs, especially in sample limiting. To be more specific, to what extent the increase in oversampling rate would reduce ISPs and therefore, increase loudness?
Oversampling also allows sample limiters to be more sensitive, so you’ll get very similar results as using a TP limiter
Hi Ian, super interesting point of view here.
Here's mine, I never seen a super loud master with true peak limiters. All tracks I bought from Bandcamp, Beatport or whatever, all have positives true peaks values, going from smaller values to bigger. I'm using Wavelab as well and here what I measured on, for exemple, TheWeeknd - Sacrifice, +0.6dB TP and an other that can match the music used in this video, David Guetta ft Benny Benassi - Satisfaction (new version) hit +1.5dB TP, for 5.8LUFS Integrated.
And I have a guess about the fact most of mixers/mastering engineers don't care about true peaks and isp, is maybe due to the way platforms normalized.
I wonder if they level down before going on the conversion of the audio file. If so, that would make sense because if we turn down volume before converting, means it should not go above 0dB TP due the big amount of reduction for going to, just example, -6LUFS to -14LUFS. I assume this goes this way cause even super loud mastered track with no TP limiters sounds not bad on platforms.
Nothing sure about that, just my guess and thoughts. Spotify or so does not say a single word about where the normalisation process is applied. I do not know about Tidal and other Hi-rez platforms tho, I think this go same way ?
Cheers and thanks for such a very well detailed video. Sorry for my average explanation, hard to tell such a complexe topic when not the native language.
As far as I know conversion come first and then normalization is applied (maybe I am wrong). ISP generation are also sensitive to codec compression settings "IME", very easy to test with the apple mastering tools by playing the audio file and changing audio codec quality. Lower the quality higher amount of ISPs. Thanks for the video Ian , cool stuff as always! 🙌
@@antblanc Where did you get that info? I'm looking for ages haha
So very interesting if it converts first, lots of masters should be "worse", no ? Cause I often send masters with gain like +1dB TP or so, and it sounds good on spotify or anywhere else. I mean I ear a lot of people saying it's going to get into a ton of distorsion and so… When I actually do some test with changing codec, it sounds bad with plugin from Apple or even the native one from Wavelab, but once online, it's a whole new story. That's what keep me wondering.
Unfortunately I’ve been told by a reliable source that several of the most popular players reduce to fixed point *before* normalisation 😕
I don’t know exactly which ones, but Spotify recommend a -2 dBTP maximum for content above -14 LUFS, which makes me think… 😕
@@GrumpyGr3g hi mate! I hear you. Normally hearing Spotify from the desktop app , normalization is enabled by default, and we probably are listening the high quality "conversion" and normalized to -14 LUFS (if you have the normal mode enabled). Surprises start in other scenarios, for example and in my case, in my car , using the mobile Spotify app (normalization is not applied) , connected to bluetooth to my car audio system, lots of loud masters distorts as hell, my theory is that in that scenario Spotify plays lower quality audio (maybe vorbis 128kbps, or similar) , I guess because that is less data bandwidth and faster to play. That is what I have experienced from my tests at least :)
@@antblanc Yup, lower data rates definitely make things worse. It should be possible to figure out if the desktop App clips the decodes now you mention it, maybe I should do some tests…
I tend to use 8x oversampling with Pro L2 and that tames a lot of the ISP issues
Agree, oversampling is a useful alternative to ISP limiting
@@ProductionAdviceI mean practically speaking it’s the same thing, true peak metering and by extension limiting is just 8x oversampling, look into ITU R bs.1770-4 where considerations for true peak metering are given and it is literally 8x oversampling, some limiters oversample the sidechain to provide accurate gain reduction without introducing the filter caused by the oversampling.
The fab filter is a very nice sound its got a softer quality. I would only use it to gain stage my limiting a bit for its flavor i would not smash it like you did.
To be clear, I never push limiters this hard. It was a real-world example from one of my clients, which I wanted to use because otherwise people would accuse me of manufacturing an un-realistic situation.
Very interesting. thank you. after watching this i wonder, would you say that if you are someone that utilises clippers in mastering then you have even more incentive to use true peak limiters as you truly aren't chopping off more than you need to? thanks again
Clippers are a whole other topic... but if the clipper you're using lets ISPs above zero through then yes, personally I'd use a TP limiter as well.
(Fwiw I only use clipping on a small proportion of masters when I need the sound. Having said that, some limiters have a clipping stage built-in - see this video: ua-cam.com/video/oOUhsHertJI/v-deo.html )
@@ProductionAdvice thanks for this 👍🏿
Great video as always Ian. I have a question though. If you are using the Waves WLM plus last in your mastering chain, and are bouncing a 24 bit file down to 16 bit (for Soundcloud or for CD's) would I then need to use a dither plug in after the WLM. For example using a Waves L2 (and just use it's dither function rather than any further limiting). Many thanks in advance, Scott.
You only need dither once at the end of the chain, so either the L2 or WLM would be fine for this purpose (Because the WLM shouldn't change the audio in any way) Hopefully that makes sense ?
Thanks for responding. If i was going from 24 bit session and bouncing down to 16 bit wouldn't i need a dither plug in alongside the WLM plus as presuming the WLM doesn't have dither built in? @@ProductionAdvice
Hm my bad, I thought WLM+ had dither. So if you want to use it as your final limiter you'll also need a dither plugin afterwards, sorry for the confusion.
No problem at all Ian. I thought as much.I use the WLM plus to get more level then an L1 or L2 as the dither, leaving it at zero on the threshold and out ceiling. The WLM plus i use your advice and set it for minus 1.0 true peak. Appreciate your help, and love all your vids @ProductionAdvice @@ProductionAdvice
true peak sounds better to me
Same !
That thumbnail... that's the first time you've done a youtube face, right? I look forward to 🤔&😲in future. Let us know on the podcast perhaps if it really does make a difference to views!
I noticed it too, this thumbnail is lit 🔥
😂 Actually no, I did a couple last year - but glad you liked it !
I get most of my loudness with saturators and clippers and enabling True Peak limiting after a clipper introduces a noticeable crunchy distortion.
So maybe it's that some people have a way to go about mastering that doesn't go well with TP limiters.
Maybe - I often use them both together, although not very aggressively.
So when you test, do you loudness-match as I do in this video ? Or just add it ? Maybe if you reduced the input to the clipper slightly to “share the load” between them it would work better.
BTW I’m pretty sure the “Aggressive” Pro-L setting used for this song includes some clipping to achieve the result
@@ProductionAdvice Well, I have never driven a limiter this hard, normally it's doing 1.5 to 2db of gain reduction, with sporadic peaks of 2.5dB.
Changing from TP to sample peak doesn't change the loudness, it just adds distortion so I'd need to back off the limiter input and get less loudness.
I see why you would want to reduce the TP values in a song like this where you're getting TP about 3.5dB higher than sample peaks, I can see that potentially becoming an issue with lossy compression.
But with 8x oversampling and without TP I get TP 0.2-0.3dB higher than sample peaks and when I use a streaming codec preview plugin I hear no apparent issues.
@@ProductionAdvice But I want to honestly thank you for providing an example where TP limiting is useful and makes a positive difference, if I hadn't seen this video I might have never considered using it but I think that I'll do some more experimenting rather than just ignoring the button altogether.
@@kelainefes Pleasure ! And yes, TP limiting is less necessary if you’re using oversampling. Strange that it causes distortion though, which one are you using ?
I'm more confused the more videos you do about it.
The approach I just happen to be using might be the right one for me anyway; mixing into a limiter with ISP limiting turned on, and getting everything to sound the way I want it.
Would seem a bit silly to turn the limiter on at the end of a mix and then snuff my transients and have to re-work them at that point.
Then I add a few dB of clipping to take the stress off the limiter, regardless of how I'm choosing to use it.
What say thee...?
That's exactly what I do ! Provided you don't push past -10 LUFS short-term, you'll be in great shape. Sorry the videos are confusing though... 😕
@@ProductionAdvice i was confused as well. i like yo Videos but 2 b honest i don’t even know what dis Clip was all about :) when i saw da waveforms: da files weren’t da same so i couldn’t really follow 🤔
@@giulioboobzilla Sorry 😕 Did you see the previous video ? It explains what ISPs are: ua-cam.com/video/-kIViigx8Rs/v-deo.html
@@ProductionAdvice no problem - U r welcum 😇 may-b i should watch dis video 1st. Thanx!
When it comes to chaining limiters would you put the true peak limiter before or after the sample limiter
Personally I just use a single TP limiter. But if you use one without this feature, you need to put the TP limiter last in the chain.
LOL, Dynameter is at 2… cray.
LOL, I almost blocked this channel until I realized the thumbnail was you! I'm not a fan of of this type of thumbnail, but still a fan of you. I hope this is an exception. What I like about your channel is precisely that it's NOT like the others who resort to poses. Just my opinion.
Sorry you didn't like it, I thought it was a bit of fun 🙂 The trouble is that thumbnails are important and there are only a limited number of times you can share a Wavelab screenshot...😕
@@ProductionAdvice Don't mind me. I also just realized what day today is. (Not sure if that's related to the thumbnail) 😉. Either way, love your channel and your presentation. Peace!
Look at you with the thumbnails!
The thumbnail is the best reaction this video has had 😛
If you measure commercial big name Artist released songs they all tend to peak over 0dBFS, so they obviously don't true peak limit.
True, although if you test (eg) UA-cam you’re seeing higher peaks because of the data-encoding.
Regardless, IMO they should…🙂
@@ProductionAdvice FYI I am testing CD's from the 90's to around 2019 and mainly Hip Hop.
Have u tried using oversampling and turning off the true peak limiter? And u still don’t have ISP In your master 😏
This is actually the way I often work, yes 👍
Online loudness normalisation doesn't work in reality, at least not on UA-cam.
In what way ?
The main issue with UA-cam I see is that they don’t increase the level of quieter material, and since there’s a lot
of broadcast and DIY material that can make the results a bit inconsistent.
Otherwise it works pretty well, in my experience
@@ProductionAdvice Ahh yes, I was thinking of making quiet stuff louder. I just realised there's lots of quiet music/videos that don't get boosted. It should work both ways really but doesn't. I guess you never notice loud music getting turned down as it will still be loud enough.
@@Dani-El. Yes, exactly 👍
the whole thing sounds SO distorted. I hear the difference and understand you chose a "real world" example, but that's so much distortion when trying to give a 'sonic' example of, reducing distortion. Confused why you chose this way over the top baked example.
It sounds horrible no matter what you do. Have we given up on fighting against the loudness war? -16 LUFS or BUST!
I haven’t given up, no…!
But to get ISPs to test effectively, things need to be pretty loud…
(Also some stuff sounds great at -16 but I don’t aim for targets)
@@ProductionAdvice As long as you are avoiding prioritizing loudness for dynamic range.