Become a Site Supporter and Never see Ads again!

Author Topic: Normalization issue  (Read 2682 times)

0 Members and 1 Guest are viewing this topic.

Offline Twenty8

  • Trade Count: (13)
  • Taperssection Member
  • ***
  • Posts: 482
  • Heading for a better thesis
Normalization issue
« on: February 04, 2020, 11:24:23 PM »
Rather than name my DAW off the bat, I am looking for an answer to something that has been perplexing my processing lately.  I run 4 channels.

My normalization was the DAW's stock normalization process (normalizing without using a common gain across all channels).
Then I changed to normalizing to a custom dB or normalizing to RMS, peak RMS, or overall peak RMS.
I generally select all 4 channels to normalize at the same time during this process.

I can not figure out why I get RMS and VU levels in the red AFTER these processes.  Seems like this is a new problem for me.  Multiple recordings now, probably the last 3 or 4.

My solution was to reduce the overall VU post normalization, or *gasp* use envelopes before normalization, then render small chuncks of problem areas to figure out how to get the final copy to just under 0 dB.

But now... here I am.
I would like some help figuring out root causes to smooth and speed up my work flow.

Which process do you use to normalize?  How do you personally normalize and get results for more than 2 channels under 0dB?
Should I do each channel individually or all channels selected at once?
And lastly... any clue why my normalization process(es) is/are taking my RMS and VU values over 0dB even though I assume they would not?

Again, I generally run 4 channels.
Although I would like to hear from users of any DAW, I use Reaper.
Open/Closed ~ AT U853 (subc, h, c, o):
Decks ~ MixPre 6ii, A10, M10
Pres ~ 2x Oade mod UA-5 (W+, ACM+)
Also ~ Line Audio CM3/CM4/OM1 : AT3031 : AT831 : DR2D

Offline jerryfreak

  • No PZ
  • Trade Count: (31)
  • Needs to get out more...
  • *
  • Posts: 6205
  • The plural of anecdote is not data
Re: Normalization issue
« Reply #1 on: February 05, 2020, 02:11:11 AM »
normalize each track to the same RMS level, independently

not sure about reaper, but in soundforge i can scan a track for levels, so say i have 2 tracks

L: 8.9 dB peak, 25.7 dB RMS
R: 6.5 db peak, 23.8 dB RMS

if i want to safely keep peaks under zero, i got about 8 dB to work with on the left channel and about 6 dB to work with on right channel.

so if i  individually normalize both channels to 18 dB RMS, the left channel will gain about 7.7 dB, and the right channel will gain about 5.8 dB (more or less), and if the mix is similar in both channels, it should have a relatively balanced sound with equal RMS in each channel

soundforge would let me normalize to even higher levels as it can be set to automatically apply dynamic compression on the few peaks that exceed 0 dBFS. often times ill do that and compress a handful of extreme peaks by a few dB
Unable to post or PM due to arbitrary censorship of people the mod doesn't like. Please email me using the link in my profile if you need to connect

Offline aaronji

  • Site Supporter
  • Trade Count: (9)
  • Needs to get out more...
  • *
  • Posts: 3853
Re: Normalization issue
« Reply #2 on: February 05, 2020, 06:45:52 AM »
Are you using four-channel playback? If not, why bother with normalizing the four channels to begin with? Mix it down to stereo and then normalize those files (since stereo seems not to present an issue).

Personally, I would not generally normalize all channels to a consistent level; those differences between channels give the recording some of its spatial aspect.

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (15)
  • Needs to get out more...
  • *****
  • Posts: 15683
  • Gender: Male
  • "Better to love music than respect it" ~Stravinsky
Re: Normalization issue
« Reply #3 on: February 05, 2020, 12:52:25 PM »
I changed to normalizing to a custom dB or normalizing to RMS, peak RMS, or overall peak RMS.
I generally select all 4 channels to normalize at the same time during this process.

I can not figure out why I get RMS and VU levels in the red AFTER these processes.

On Peak vs RMS-
An RMS measure approximates perceived sound level.  It reflects a specifically formulated "averaged" value across peaks and troughs over short duration which more closely corresponds to our perception of loudness. RMS is an intentionally "slow" measurement, where as Peak is an "instantaneous" measurement.   Rapid transient peaks produce brief high level values but do not sound "loud".

With simple periodic waveforms such as a straight sine wave, the relationship between RMS value and Peak value can be calculated and is predictable.

But audio produces complex arbitrary waveforms and is not predictable in the same way, so with music the relationship between peak and RMS values is not constant or known without measuring.   If you choose to normalize based on RMS value and choose an overly high normalization target value, the waveform will be amplified such that the highest RMS value equals your normalization target, but the peak values will exceed 0dbFS. 

To avoid that, if the goal is getting close to but not exceeding 0dbFS, normalization based on peak value is more appropriate than RMS.  In that case the waveforms will be amplified such that the highest momentary peak matches the normalization target but RMS may vary between channels.  If the goal is getting all the sources to similar loudness levels, then RMS normalization is appropriate, but each file can then have different maximum peak values and one does not know exactly what the maximum value might be.

You may be well aware of all this already. I just wanted to clarify these fundamentals.


How it applies-
We normalize or adjust gain manually to optimize the level of a file so that the signal fits comfortably within the dynamic range of the medium.  This clearly applies to the output format where we manipulate the master-bus output level to get the highest peaks relatively close to but not exceeding 0dBfs.  Yet prior to that we may need to manipulate levels for additional reasons: Getting each raw file to a reasonable working level within the mixing/editing software, getting approximately similar levels across all channels prior to mixing them, and making mixing decisions which will precisely determine the level relationships between channels, which we wish to preserve from that point onward.

If you are recording in the traditional way such that the highest peaks range between, say -12dBfs and -3dBfs, you don't really need to normalize the raw files prior to mixing them.  But if you had very low recording levels in one or more channels, or if you are using a 32-bit-floating point recorder where you aren't manually setting input levels, you'll want to normalize to some reasonable starting-point level prior to mixing.  That simply gets them all in a good working range within the mixing software.  Once you start mixing, do so by ear - checking the balance of each stereo source in isolation and adjusting that if necessary, then adjusting the balance between each of those pairs and individual channels.  At this point you want to preserve the overall level relationships between channels based on your mix choices.  Once the mix is done, you might peak normalize the 2-channel output (or manually adjust gain, or limit, or some combination of these things) to make sure the peaks are sufficiently close to 0dBfs yet do not exceed it.  But you don't want to normalize each channel individually at that point as doing so will change the mix level relationships, and will affect mix features such as threshold settings in an unwanted way.

Specifics-
Quote
I changed to normalizing to a custom dB or normalizing to RMS, peak RMS, or overall peak RMS.

Custom dB presumably normalizes based on the highest instantaneous Peak value found within the range (typically the length of the entire file).
RMS presumably normalizes to the highest RMS value found within the range (typically the length of the entire file).
I don't know what peak RMS or overall peak RMS mean or refer to.  Those are confusing and potentially misleading terms to be sure.

Quote
..any clue why my normalization process(es) is/are taking my RMS and VU values over 0dB even though I assume they would not?

If you normalize two or more raw files to RMS=0dBFS prior to mixing them, the RMS sum of those files will exceed 0dBFS.  And even if RMS normalizing to a few dB below 0dBFS, the RMS of the sum may still exceed full scale.  Yet even if levels above full scale are occuring within the mixing program, it's not necessarily a problem at that point as long as long as arrangements are made so as to peak below 0dBFS prior to output.  This is because all modern mixing programs use 32-bit float internal calculations which allows them to handle levels above than 0dB internally without clipping.  It's really only when writing an output file that the signal must be properly "fit to the output file limits" again.  However, it's still best to maintain good levels within the software so that various thresholds and other tools such as a limited amplification ranges or fader movement ranges all work correctly.

 Even though modern tools make it less critical to keep an eye on than it once was, good gain staging remains fundamental though all phases of recording and production.
« Last Edit: February 05, 2020, 06:23:16 PM by Gutbucket »
musical volition > vibrations > voltages > numeric values > voltages > vibrations> virtual teleportation time-machine experience
Better recording made easy - >>Improved PAS table<< | Made excellent- >>click here to download the Oddball Microphone Technique illustrated PDF booklet<< (note: This is a 1st draft, now several years old and in need of revision!  Stay tuned)

Offline Twenty8

  • Trade Count: (13)
  • Taperssection Member
  • ***
  • Posts: 482
  • Heading for a better thesis
Re: Normalization issue
« Reply #4 on: February 05, 2020, 05:08:57 PM »
After reading these first three responses, and before I respond to each, I just want to quote the following and clarify:

Then I changed to normalizing to a custom dB or normalizing to RMS, peak RMS, or overall peak RMS.
I generally select all 4 channels to normalize at the same time during this process.

All of those options are available, but I have never normalized all 4 tracks as a group to the same RMS level.  Each track gets normalized individually based on its own RMS.
Open/Closed ~ AT U853 (subc, h, c, o):
Decks ~ MixPre 6ii, A10, M10
Pres ~ 2x Oade mod UA-5 (W+, ACM+)
Also ~ Line Audio CM3/CM4/OM1 : AT3031 : AT831 : DR2D

Offline jerryfreak

  • No PZ
  • Trade Count: (31)
  • Needs to get out more...
  • *
  • Posts: 6205
  • The plural of anecdote is not data
Re: Normalization issue
« Reply #5 on: February 05, 2020, 05:18:36 PM »
Are you using four-channel playback? If not, why bother with normalizing the four channels to begin with? Mix it down to stereo and then normalize those files (since stereo seems not to present an issue).

i agree with this, it would be unusual to want to mix 4 channels down to a stereo pair at exactly a 50/50 mix

im not a fan of 4 mic mixes in general, due to the odd phasing imo
Unable to post or PM due to arbitrary censorship of people the mod doesn't like. Please email me using the link in my profile if you need to connect

Offline Twenty8

  • Trade Count: (13)
  • Taperssection Member
  • ***
  • Posts: 482
  • Heading for a better thesis
Re: Normalization issue
« Reply #6 on: February 05, 2020, 05:21:37 PM »
Are you using four-channel playback? If not, why bother with normalizing the four channels to begin with? Mix it down to stereo and then normalize those files (since stereo seems not to present an issue).

Personally, I would not generally normalize all channels to a consistent level; those differences between channels give the recording some of its spatial aspect.

I think what you are suggesting is to mix down both left channels and both right channels of my four track recording, then normalize the result.
As I mentioned in my above post, for clarification I don't normalize all the channels to a certain level.  Each one ends up with a different increase in dB (ch1 +6.88, ch2 +8.20, ch3 +5.7, etc.)

I like this suggestion.  It seems simple enough.

normalize each track to the same RMS level, independently ... and if the mix is similar in both channels, it should have a relatively balanced sound with equal RMS in each channel

not sure about reaper, but in soundforge i can scan a track for levels

I also like this suggestion.  It's more complex, but I think its accuracy and allowable adjustment.  I have to "math", but it seems simple enough.  However, I feel like I will have to render small areas in order to check if the highest peaks fall under the red.

To the second point, I searched around to find that Reaper will allow for scanning levels.  I didn't know it before, but for Reaper users its Actions > show action list > "rms" in the filter > sws: analyze and display item peak and RMS (entire item).
SWS plug-in extensions are what I changed to for normalization (from Reaper's native normalization).
Open/Closed ~ AT U853 (subc, h, c, o):
Decks ~ MixPre 6ii, A10, M10
Pres ~ 2x Oade mod UA-5 (W+, ACM+)
Also ~ Line Audio CM3/CM4/OM1 : AT3031 : AT831 : DR2D

Offline Twenty8

  • Trade Count: (13)
  • Taperssection Member
  • ***
  • Posts: 482
  • Heading for a better thesis
Re: Normalization issue
« Reply #7 on: February 05, 2020, 05:26:34 PM »
i agree with this, it would be unusual to want to mix 4 channels down to a stereo pair at exactly a 50/50 mix

I haven't yet mixed at 50/50.  That mix generally sounds ... not to my liking.  I don't feel all four channels blend well that way.
Usually my main pair will supercede my subcards.
However this weekend, with the distance of the stacks apart from each other, the recording sounds better (IMO) with the supercards in DINa reduced several dB and the subcards out in front at NOS.
Open/Closed ~ AT U853 (subc, h, c, o):
Decks ~ MixPre 6ii, A10, M10
Pres ~ 2x Oade mod UA-5 (W+, ACM+)
Also ~ Line Audio CM3/CM4/OM1 : AT3031 : AT831 : DR2D

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (15)
  • Needs to get out more...
  • *****
  • Posts: 15683
  • Gender: Male
  • "Better to love music than respect it" ~Stravinsky
Re: Normalization issue
« Reply #8 on: February 05, 2020, 06:12:19 PM »
Primarily, normalization (peak) is used as a way to get the signal level of the output file as high as possible without exceeding 0dbFS.

The best reason I can think of to normalize each individual channel prior to mixing is to get approximately the same level in each as a starting point for mixing.  This is similar to what jerryfreak is proposing. 

You can do the same with manual amplification changes rather than normalization.  Using normalization to do so simply automatically determines the amplification needed to reach the normalization target.  If the recording had decent levels to start with, you don't necessarily need to do any initial normalization at all. BTW, I mentioned 32-bit-floating point storage in my post above because as recording to that format becomes more common, so will including normalization as a first post-production step.  Choosing RMS normalization for this step is appropriate because it brings all the files to approximately the same loudness level, and it doesn't matter if some peaks exceed 0dBFS inside the mixing software.

Stereo balancing of pairs and the mix balance between them should always be determined by ear.  If using an identical 4-channel microphone setup each time, your prefered mix balance may end up being similar from recording to recording, but the only way to determine that is by listening while adjusting the balance between them.
^
Knowing the actual mix ratio used isn't important, is not information which is translatable to another recording.. and can actually be misleading!   Unless the files actually were RMS normalized prior to mixing so that they all start with equivalent loudness levels, the mix ratio between them not only determines the relative contribution of each element, but also compensates for the level differences of the raw files.  Because it is determined by those compounded variables, it becomes meaningless.
« Last Edit: February 05, 2020, 06:33:02 PM by Gutbucket »
musical volition > vibrations > voltages > numeric values > voltages > vibrations> virtual teleportation time-machine experience
Better recording made easy - >>Improved PAS table<< | Made excellent- >>click here to download the Oddball Microphone Technique illustrated PDF booklet<< (note: This is a 1st draft, now several years old and in need of revision!  Stay tuned)

Offline Gutbucket

  • record > listen > revise technique
  • Trade Count: (15)
  • Needs to get out more...
  • *****
  • Posts: 15683
  • Gender: Male
  • "Better to love music than respect it" ~Stravinsky
Re: Normalization issue
« Reply #9 on: February 05, 2020, 06:28:57 PM »
im not a fan of 4 mic mixes in general, due to the odd phasing imo

Plenty of that to be found, mostly due to the use of microphone arrangements which aren't properly designed to avoid it - such as combining two standard 2ch stereo near-spaced pair configurations, even though they work well on their own in isolation.  But that's another discussion altogether.
musical volition > vibrations > voltages > numeric values > voltages > vibrations> virtual teleportation time-machine experience
Better recording made easy - >>Improved PAS table<< | Made excellent- >>click here to download the Oddball Microphone Technique illustrated PDF booklet<< (note: This is a 1st draft, now several years old and in need of revision!  Stay tuned)

Offline Twenty8

  • Trade Count: (13)
  • Taperssection Member
  • ***
  • Posts: 482
  • Heading for a better thesis
Re: Normalization issue
« Reply #10 on: February 05, 2020, 06:53:58 PM »
If you normalize two or more raw files to RMS=0dBFS prior to mixing them, the RMS sum of those files will exceed 0dBFS.  And even if RMS normalizing to a few dB below 0dBFS, the RMS of the sum may still exceed full scale.  Yet even if levels above full scale are occuring within the mixing program, it's not necessarily a problem at that point as long as long as arrangements are made so as to peak below 0dBFS prior to output.  This is because all modern mixing programs use 32-bit float internal calculations which allows them to handle levels above than 0dB internally without clipping.  It's really only when writing an output file that the signal must be properly "fit to the output file limits" again.  However, it's still best to maintain good levels within the software so that various thresholds and other tools such as a limited amplification ranges or fader movement ranges all work correctly.

 Even though modern tools make it less critical to keep an eye on than it once was, good gain staging remains fundamental though all phases of recording and production.

The underlined statement helps me to understand more of what jerryfreak and aaronji are saying.
Take care of your mix, mix the four tracks down to stereo, then normalize the results to just under 0dBFS.

The best reason I can think of to normalize each individual channel prior to mixing is to get approximately the same level in each as a starting point for mixing.  This is similar to what jerryfreak is proposing. 
Yeah, this hits home with me.  I generally have been doing this so I could figure out envelopes when I was running a podcast.  As I am processing the recording in front of me, and am about to start over, it DID help me figure out the mix balance between the pairs and a small amount of stereo balance before I begin again.

Stereo balancing of pairs and the mix balance between them should always be determined by ear.  If using an identical 4-channel microphone setup each time, your prefered mix balance may end up being similar from recording to recording, but the only way to determine that is by listening while adjusting the balance between them.
^
Knowing the actual mix ratio used isn't important, is not information which is translatable to another recording.. and can actually be misleading!   Unless the files actually were RMS normalized prior to mixing so that they all start with equivalent loudness levels, the mix ratio between them not only determines the relative contribution of each element, but also compensates for the level differences of the raw files.
While I commonly use my Line CM3s in most recordings, my main pair changes constantly as I grow into this hobby.  I have to continually change ratios.  So much of my processing is bouncing between levels as I master, raising and lowering a pair to meet a sound I can be happy with.  I find myself several times a week, nearly daily as I tape more and more, trying to determine what sounds good between my playback and my headphones (and sometimes my truck).

As always gutbucket, a great detailed read.  Thank you.
Open/Closed ~ AT U853 (subc, h, c, o):
Decks ~ MixPre 6ii, A10, M10
Pres ~ 2x Oade mod UA-5 (W+, ACM+)
Also ~ Line Audio CM3/CM4/OM1 : AT3031 : AT831 : DR2D

 

RSS | Mobile
Page created in 0.161 seconds with 35 queries.
© 2002-2024 Taperssection.com
Powered by SMF