|
Post by scumbum on Mar 4, 2017 16:39:23 GMT -6
Is the Pro Tools Normalization plugin pretty good at not degrading the the quality of the audio ? Or are there better Normalization plugins out there ?
|
|
|
Post by popmann on Mar 4, 2017 17:11:03 GMT -6
Normalization can't be done in real time....ie, via "plug in".
There is no difference in "quality"--it's literally finding the peak and turning the volume up or down until that peak is the given value. It's not altering the audio anymore than your fader is.
|
|
|
Post by ChaseUTB on Mar 4, 2017 17:18:21 GMT -6
Is the Pro Tools Normalization plugin pretty good at not degrading the the quality of the audio ? Or are there better Normalization plugins out there ? Yes it is non destructive and its accurate. You can normalize your -12dbfs peak audio file to -.3dbfs Peaks and then reduce the -.3dbfs peak file back to -12dbfs with the same plug and flip the phase and they will cancel. You can also use trim or gain. Clip gain is so easy... print your mix at whatever it is peaking at ( for mixers who mix low volume as well as ones who love volume ) Then get out a meter and use the clip gain line or info volume slider and add or subtract until your mix Peaks at whatever you like dbfs wise (-8dbfs or -6dbfs or -12dbfs if you are Wiz ๐ฌ)
|
|
|
Post by scumbum on Mar 7, 2017 14:06:00 GMT -6
So if using Normalization is non destructive then there is absolutely no reason at all to record loud levels because you can always turn them up later come mix time using Normalization . Ronan Chris Murphy said some converters even before 0 start to distort a little so even getting near the top is not good . You could pretty much just leave your mic pres gain knob all the way down if you wanted all the time . What level when recording at 24 bit is actually TOO low . This would be based on the mic pres noise floor ?? Lets say a CAPI VP26 . How low can you record with a VP26 before noise becomes a factor ? Heres a beautiful chart I found ,
|
|
|
Post by wiz on Mar 7, 2017 14:47:18 GMT -6
that chart is why my mixes peak where they do 8)
cheers
Wiz
|
|
|
Post by ragan on Mar 7, 2017 15:17:43 GMT -6
I'm gonna admit something embarrassing. I don't really know what "normalizing" is and I've never done it.
Thank you.
Ragan
|
|
|
Post by wiz on Mar 7, 2017 15:24:17 GMT -6
I'm gonna admit something embarrassing. I don't really know what "normalizing" is and I've never done it. Thank you. Ragan The audio gets scanned, its peak level noted (e.g. -12.1) and then you "normalise" that to a defined level.... e.g. make the -12.1 peak, now -0.1 ... so the file gets 12dB of gain added to it. cheers Wiz
|
|
|
Post by scumbum on Mar 7, 2017 15:55:54 GMT -6
I'm gonna admit something embarrassing. I don't really know what "normalizing" is and I've never done it. Thank you. Ragan Do you use pro tools ? if you do go to the audiosuite plugins , and you'll find Normalize . Highlight a section of audio and if the plugin is set at 0 it will then change the highest peak of that audio to hit right under or at 0 . It's just quick volume control , but for doing comparisons of stuff like mic pres you can match levels exact .
|
|
|
Post by ragan on Mar 7, 2017 16:06:06 GMT -6
I'm gonna admit something embarrassing. I don't really know what "normalizing" is and I've never done it. Thank you. Ragan Do you use pro tools ? if you do go to the ย audiosuite plugins , and you'll find Normalize . Highlight a section of audio and if the plugin is set at 0 it will then change the highest peak of that audio to hit right under or at 0 . It's just quick volume control , but for doing comparisons of stuff like mic pres you can match levels exact .ย I do, thanks. And thanks wiz for the explanation. I guess that's why I've never used it, not really very necessary to me in my workflow.
|
|
|
Post by jeremygillespie on Mar 7, 2017 19:18:44 GMT -6
I've never understood the normalization thing, mostly because when I learned, the guys that were tracking were always setting their levels to tape so that when they put all the faders to 0 on the console - you had your basic rough mix.
I've been doing it like that ever since.
|
|
|
Post by scumbum on Mar 7, 2017 21:21:38 GMT -6
I've never understood the normalization thing, mostly because when I learned, the guys that were tracking were always setting their levels to tape so that when they put all the faders to 0 on the console - you had your basic rough mix. I've been doing it like that ever since. I was thinking today if you had specific levels you liked , -12 kick , -12 snare , guitars -20 .........you could normalize everything to your prefered level to get it right in the ballpark when you begin a mix . Then do your final tweaks to the levels with the faders and you'd probablly always be pretty close and save a decent amount of time .
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Mar 8, 2017 21:01:56 GMT -6
Normalization IS destructive. This is especially obvious with lower level values--even if those values are just the components of a waveform near 0 dB. Let's use an example.
Assume your normalization scan indicates the overall signal should be boosted by 1dB. This means that it should be multiplied by 1.122 and change. Now assume you've got part of a signal that has 4 bits of magnitude. Expressed as integers, you can code values from 8 to 15. Multiply those values by the boost of 1.122. A value of 8 turns to 8.976. A value of 15 turns to 16.83. But there's no such thing as a fractional bit. Those values must be rounded or truncated (they're usually truncated). So our original value of 8 stays at 8 and our original value of 15 now becomes 16. We haven't just changed the gain of the waveform--we've distorted it. This is less obvious in the louder parts of the signal, since the error is lower. But for every time a wave hits a max, it also passes through a min.
If you absolutely feel like you've got to change the gain, then clip gain is better. It's non-destructive and the math will happen in floating point, pushing the error waaaay down.
But the real question is why you want to normalize in the first place. Maximum signal level is only loosely connected to perceived loudness. The loudness you perceive in a track is a function of the actual content as well as the tracks before and the tracks after. There's no reason to introduce distortion into a track--especially when you're making it permanent. Final adjustment of overall levels should be moved to as late in the process as possible.
|
|
|
Post by ChaseUTB on Mar 8, 2017 22:20:55 GMT -6
@exponenetialaudio
AudioSuite 'Overwrite Files' Option Usually, audio suite plug-ins render audio non-destructively. When you process a clip, the original remains available in the clip list. However, if you click on the Processing Output Mode button near the top left of an audio suite plug-in, you get the option to overwrite files.
Gain is an AS plugin. So is normalize. I believe there are options for both. The excerpt above is from PT 11 201 book.
|
|
|
Post by scumbum on Mar 8, 2017 22:23:40 GMT -6
Normalization IS destructive. This is especially obvious with lower level values--even if those values are just the components of a waveform near 0 dB. Let's use an example. Assume your normalization scan indicates the overall signal should be boosted by 1dB. This means that it should be multiplied by 1.122 and change. Now assume you've got part of a signal that has 4 bits of magnitude. Expressed as integers, you can code values from 8 to 15. Multiply those values by the boost of 1.122. A value of 8 turns to 8.976. A value of 15 turns to 16.83. But there's no such thing as a fractional bit. Those values must be rounded or truncated (they're usually truncated). So our original value of 8 stays at 8 and our original value of 15 now becomes 16. We haven't just changed the gain of the waveform--we've distorted it. This is less obvious in the louder parts of the signal, since the error is lower. But for every time a wave hits a max, it also passes through a min. If you absolutely feel like you've got to change the gain, then clip gain is better. It's non-destructive and the math will happen in floating point, pushing the error waaaay down. But the real question is why you want to normalize in the first place. Maximum signal level is only loosely connected to perceived loudness. The loudness you perceive in a track is a function of the actual content as well as the tracks before and the tracks after. There's no reason to introduce distortion into a track--especially when you're making it permanent. Final adjustment of overall levels should be moved to as late in the process as possible. Clip gain ......you mean in Pro Tools the Digirack "Trim" plugin ?
|
|
|
Post by ChaseUTB on Mar 9, 2017 0:09:27 GMT -6
I am not trying to disagree at all and of course I could be mistaken or wrong regard destructive vs non- destructive and by no means am I tryna stir up anything, in fact I am eager to learn more hence why I continued the discussion.
Clip gain is amazing. Trim is an insert plugin ( destructive or non ) I say Non here is why:
I don't like the sound of the Apollo preamps with more than 30-35 Db of gain. Seeing as I don't have a console or mixer ( to my knowledge ) I have nowhere else too add gain ( read gain as vocal volume in HP, I track conservatively in PT, music im recording to is peaking around -6 to -10 ). Even With the HP output around 3 o clock on the Apollo starts distorting really bad in my both my shure sh440 and sennheiser 280 pro hd ( pro hd what does that even mean ๐ Shit sounds great tho ๐ญ ) And the Vox are not audible because I like to track loud... I insert a trim plug as the first plugin and boost 6-9 Db on the vocal recording track. When I trim the audio up 9 Db and back down it's the same file... what happens when you flip the phase?
Better yet, I am going to record one vocal print +9 Db of gain via trim plugin and write it to the file then trim it back down -9db print that ( render ) and then flip the phase against the original.... if there is silence this proves non destructive editing, correct? ( I'm asking )..
Now if normilzation is not overwriting the file as per my excerpt a couple posts above, wouldn't this process also fall under both categories depending on the personal PT preferences?
Also if normalize applies different gain to different bits or integers how does this happen and why? Last but not least just how destructive are we talking? Thanks in advance @exponentialaudio ๐
|
|
|
Post by scumbum on Mar 9, 2017 9:34:13 GMT -6
So it looks like Clip Gain came out in PT 10 . I'm running 8 , so I don't have it .
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Mar 10, 2017 10:41:16 GMT -6
The real difference between normalization and other gain-management tricks (clip gain, gain plugins and so on) is that classic normalization is destructive and replaces the values in your sound file with new values. That's problematic for reasons I described earlier, and of course it violates the unwritten law of never screwing with source tracks.
Any other gain adjustment does create distortion (even faders), but it's much less harmful for a number of reasons. All mixing these days is done in floating point math (the old TDM stuff may have been the last generation that wasn't) which by its very nature keeps signals in a range that gives the best possible result. So there is distortion--and it does accumulate over a long chain of effects and gain tweaks--but it tends to stay at much lower levels. And since it appears only in the final mix result, this means that you can tweak the mix and not make things worse (because the first link in the chain is still your original pristine audio). If you normalize, you start with a distorted signal and then keep adding to it as you move down the chain.
And there's alway a good reason to deliver at least 24-bit files to the mastering engineer (floating-point files are even better), since that engineer has the knowledge and the tools to prepare your final mix in a way that gives a satisfying result with a minimum of schmutz.
|
|
|
Post by popmann on Mar 10, 2017 15:19:12 GMT -6
I really don't like the implications of "if you absolutely feel like you've got to change the gain".
You DO, IME. Unless you recorded the tracks to your studio nominal level....or someone else recorded to theirs AND theirs is close enough to yours to not bother.
If you DON'T....then you have to immediately take all the faders to their bottom 1/3rd of their throw where there's far less granularity in movement....you can't insert your analog gear via hardware insert because your IO is calibrated to your studio's level. Actually--I would argue that's usually what DEFINES your studio's nominal level. You can't use analog modled plug ins that expect a certain level as X voltage.
Anyway-tape return channel trim has been a function of literally every mixer I've worked on my entire life. Analog or digital. Done for somewhat different reasons on analog than digital, but it's always been step on of mix time to get all the channels getting their feeds at the right levels. So, the implication that it's a some weird harebrained idea is....patently historically incorrect.
THAT SAID....file normalization is not the way to do it for a different reason. I honestly don't KNOW that it damages the sound in any meaningful way....why would it not be the same 32bit volume math as doing it any other way? I guess theoretically it could be that writing the 24bit file back embeds something "lossy".... But, I've never tested because I've never done it for a different reason: The reason it's not the best way to do it is that it gives too much "value" to an absolute momentary (potentially one time) digital peak level. Which isn't how levels are set. OP has noticed this by his assertion that he wants drums "normalized" to a different level than guitars....this is actually WHY normalization is sort of a dumb way to do it. A Distorted guitar and a snare should be roughly set to a pretty similar window, but because of the nature of that snare it will mean there are some higher absolute peaks....where the dirty guitar doesn't have much in the way of high absolute peaks.....it's average peak will actually BE pretty near it's peak....where the snare's average peak might be 10db below some occasional peaks.
So, the discussion to be had is what's the least destructive way to adjust the gain of tracks at the beginning of a mix....PRE fader/insert/channel. I used the DSP24's 56bit fixed mixer's channel trim....then when I moved to software, Cubase's input channel trim, which is right at the top of the mixer channel next to the polarity like I'm used to from analog and later digital mixers....Logic and Protools not having trim/polarity functions as part of the mixer is a really shit move, IMO, from a position of workflow. While it's easy to say "just insert a Gain (in Logic that handles both trim and polarity) across all the channels--that still doesn't tell me from looking at it what's being done. Is the snare bottom reversed polarity? Now I have to open a plug in to see? I suppose I could use two....and just know the first is polarity and the second is trim....and default them to deactivated....I digress.....point is, the channel input trim is how I've always done this. I would love to hear if there's abetter way--I've taken to using the Region Gain, which is effectively the "clip gain" that Cubase and Logic have had for decades without a catchy name....since Logic requires a plug in to trim....
|
|
|
Post by ChaseUTB on Mar 10, 2017 15:44:45 GMT -6
I hate to say this my mentors and teachers all taught me to save as then save new session correct -> Song Name - Artist - (date ) - 1a static mix prep - normalize. Then the new save as or copy session opens.
I was then was taught to listen, unclip and faders/ busses/ by trimming or gaining. Then highlight and select all audio and Normalize everything to -6b dbfs Peak, then commence the static mix. So all those steps are introducing distortion.. good or bad distortion ? I assume bad
Funny thing is I know how to track correctly and als have always used clip gain or trim plugins ( if needed in other ppl sessions ) so I have never incorporated the normalize all audio to -6dbfs .... idk maybe because I used less compression then but the mixes that were normalized always exploded. Even a mix of my music I did had the same effect when all files were normalized. Then I normalized to -.3dbfs and limited about 2db and that is one of the loudest songs I have ( I don't even think I went through with the mastering process aside from proper dither and src )
Wow this is throwing me for a loop right now especially the Avid's own textbooks saying it's non destructive...
Why would one teach another this method of inducing distortion via normalize etc is the proper way if not? This was 2012 PT 10 and a expensive to me school...
I worked with a couple of my teachers outside school ( who have credits and nominations and they were taught by their studio mentor the normalize thing.. Maybe it was an advent of more and more people making music but not tracking properly so they used the normalize to " gain stage " idk that's the only thing I can think of maybe is a " time saving shortcut " ๐ค๐ค๐ค @exponentialaudio thanks this is kinda throwing me here. Good to know and I appreciate the breakdown and by all means accept your stance as truth that's why I am kinda at a loss here as I know students are being taught this everyday ๐ต๐ต๐ต
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Mar 10, 2017 19:01:10 GMT -6
I worked with a couple of my teachers outside school ( who have credits and nominations and they were taught by their studio mentor the normalize thing.. Maybe it was an advent of more and more people making music but not tracking properly so they used the normalize to " gain stage " Back in the analog days it was a good idea (and considerate as well) to provide tapes at a known maximum level. And you'd typically provide a tone at the beginning of the tape that would allow the gain knobs to be set in such a way as to match that. But it's important to remember that the meters used on tape machines were VU meters of some sort. The very nature of the VU meter is to integrate the signal over a short interval of time in order to show a value that's very close to RMS. Peaks (of the sort that normalization looks at) are quite deceptive. You can have strong peaks (especially for the higher-frequency components of a signal) that will be very strong, while the overall signal is still perceptually very quiet. By the same token, you can have a signal comprised of mainly low frequencies that may appear to need boosting. But when you normalize that signal, it will be really loud. So I think those teachers who recommend the practice of normalization are extrapolating from something that was a real need in the analog days. But digital is really different for a number of reasons and normalization is sort of like using a buggy whip to make your car go faster. It might be a better practice to decide on an operating level, set your meters to an RMS representation and then adjust trim so the the signal is approximately at that level.
|
|