Copy link to clipboard
Copied
What is the proper workflow to process field recorded band music audio from start to finish?
I was researching compression and came up conflicting information.
First, I was advised years ago not to normalize the audio. I read conflicting posts for and against
After the research, I came up with this. Please look over the steps and advise if I should make any corrections.
The music audio is recorded at 48000 hz 24 bit
Open the audio file in audition from the recorder
Convert to 48000hz 32 bit
normalize to -.1 db.
Since I never normalized in the past
should I reduce random music spikes first before normalizing?
Process clapping spikes before normalizing or wait till later. I am concerned that the clapping spikes would affect the normalizing results
should I not normalize at the beginning or wait towards the end of processing prior to the compression step
Eyeball and reduce volume to about -6db
Perform EQ adjustments
Clap reduction processing
Dynamics processing
Volume maximizing
convert to 16 bit
Dither
Thanks Rick
The reason that it's always best to normalize a signal to 0dB before doing any form of compression at all - even limiting - is that 0dB is the reference point for all the settings - it's that simple, and I believe that it's pretty much universal. What that means is that if you apply any compression or limiting effect at all to a non-normalized file, the effect you get will be completely unpredictable, i.e. RANDOM! If you decide to apply compression after peak limiting (not unusual) then you need
...Copy link to clipboard
Copied
The reason that it's always best to normalize a signal to 0dB before doing any form of compression at all - even limiting - is that 0dB is the reference point for all the settings - it's that simple, and I believe that it's pretty much universal. What that means is that if you apply any compression or limiting effect at all to a non-normalized file, the effect you get will be completely unpredictable, i.e. RANDOM! If you decide to apply compression after peak limiting (not unusual) then you need to normalize again - for just the same reason; you won't know what point your compression is being applied to otherwise.
As far as the workflow overall is concerned, then all you need to do is look at a conventional mixer, and the order in which effects are applied: gain control first (effectively the same as normalizing), then there's an insert point where a compressor of some sort goes, then the EQ. If you compress after the EQ, then you will have altered the effect of it - possibly quite dramatically - so you always compress first, and alter the EQ response afterwards.
Copy link to clipboard
Copied
err, normalizing to 0db gives no headroom for eq or some effects that might raise the gain as a byproduct, unless you only cut. imho eq before compression. Here's the steps I do it. Everyone has different opinions on this so...
if you read the whitepapers and flowcharts, you'll see this pretty much matches what the engineers do as well. so does RX btw.
This is the order I do my audio for film. I put effects in order of non destructive. You can re-normalize 50x a day and that effect won't distort the raw data so don't worry about a wrong setting.
1. dc offset - offset power errors affects dynamic range so this goes first, a critical pre-alignment step
2. match volume - normalize 'peak amplitude' everything to -6db to see serious errors easier like clipping, offsets, phasing etc.
3. declip - run declip effect, remove buzzing with dynamics processing, notch filter the rest.
4. equalize - run parametric equalizer to reduce wind, hiss via freq cut, give vocal clarity 'D' curve
5. phase correct - once tracks layed, detect any phasing errors
6. noise removal - declick, de-reverb, remove hiss, wind, learn sound model/noise reduction, all else healing brush
7. studio reverb - to add vocal weight
8. RMS -23db audio - to roughly match vocals begin basic audio editing with levels(vocals center and 12db higher than music, etc.)
9. compression - multiband compress sfx: 4:1, vocals 2:1 db to give pleasing db range. always 2nd to last step because any other processes affect dynamic range. It should soft limit for you. hard limits are audible and not recommended.
10. quality control - match loudness ITU-R BS.1770-3 -23 LUFS film, -16LUFS youtube - for standard
Copy link to clipboard
Copied
chrisw44157881 wrote
err, normalizing to 0db gives no headroom for eq or some effects that might raise the gain as a byproduct, unless you only cut. imho eq before compression. Here's the steps I do it. Everyone has different opinions on this so...
if you read the whitepapers and flowcharts, you'll see this pretty much matches what the engineers do as well. so does RX btw.
And where exactly did I suggest normalizing for EQ? No, that's right, I didn't. I only mentioned compression. And what I'm saying is not an 'opinion' but straightforward fact. If you EQ before you compress, then the compression will alter the effects of your EQ. That is a fact. That is also why hardware mixers have their insert points before the EQ. That's where compression goes. Oh, and the RX charts don't suggest anything of the sort, as it doesn't have a compressor.
Yes there is a considerable degree of ignorance out there, and some of it is in patently incorrect flowcharts. Please don't add to it here.
Copy link to clipboard
Copied
Steps 1,3,5,6 and 7 should be completely unnecessary with properly recorded audio IMHO.
Copy link to clipboard
Copied
Thanks for the input
The recordings are not soundboard but from a portable recorder and external mics.
My goal is to get a crappy sounding audience recording to sound closer to pro recorded audio
Software used are
Audition cc latest version
Har-bal 3.7 for equalizing and eq match
Ozone 8
RX6
Original workflow (close to suggested by Izotope literature and tutorials. They never mention Normalization which created my conflict while researching)
Match Volume to about -6 db
Equalize
Attenuate clap spikes if and where necessary
Compression if necessary
Quality control.
Maximize volume
convert to 16 bit
Dither
proposed workflow with compression (only if compression is necessary. Otherwise follow original workflow omitting compression step and without normalizing)
Normalize to 0 DB
Compression
Equalize
Attenuate clap spikes if and where necessary
Quality control.
Maximize volume
convert to 16 bit
Dither
please review and suggest any changes. Then I should be good to go.
Thanks
Rick
Copy link to clipboard
Copied
Rick0725 wrote
proposed workflow with compression (only if compression is necessary. Otherwise follow original workflow omitting compression step and without normalizing)
Normalize to 0 DB
Compression
Equalize
Attenuate clap spikes if and where necessary
Quality control.
Maximize volume
convert to 16 bit
Dither
This is pretty much it, yes, except that the compression stages are potentially the wrong way around! If you apply compression to the signal, the spikes will be reduced, but you won't really be able to control how much, or how complete this is. What you actually need to do rather depends upon the nature of the applause; if the spikes are only 2-3dB above the general level, then it's hardly worth bothering with them, and your workflow is fine. If they are much higher (say 10dB) then the thing to do is to remove those spikes first by limiting - it will be relatively easy to see where the threshold point is - and then normalize again, so that you can set the threshold for whatever compression you're going to use. Oh, and quality control happens at every stage!
If you are not conversant with the concept of 'limiting', please say.
Copy link to clipboard
Copied
Should perhaps add that I wouldn't normally do this with applause unless there was something seriously wrong with it - I just manipulate the level of it so that it's appropriate.
Copy link to clipboard
Copied
Where do I Normalize to 0 DB is the question ???
The main point of the discussion was where to stick the normalization step since I never normalized before. Instead of normalizing at the first step, I volume matched to -6 db. before I began processing.
Initial Volume adjustment
Equalize
Attenuate clap spikes if and where necessary (Think it is best to address the spikes before compression since the spike level is typically way above the music level peaks and would make normalizing less "accurate" for compression)
place Normalize to 0 DB here prior to compression? (or do not normalize at all)
Compression
Quality control.
Maximize volume
convert to 16 bit
Dither
It seems logical now to normalize based on the discussions just before the compression step. Or should I not bother to normalize here is the ultimate question.
please review and advise. Want to get this normalization step right confidently. I am concerned that I did things wrong for years. Plus everything on the internet is so inconsistent with too many options. I just want to get the basics right. .
Thanks
Rick
Copy link to clipboard
Copied
Rick0725 wrote
Where do I Normalize to 0 DB is the question ???
This is easy, and in my first reply: Every time you go to use a compressor, or limiter you normalize first - otherwise the controls mean nothing. The only times you can 'get away' with it (and the reason that hardware compressors have them) is if you have a reduction meter showing you how effective your settings are - and even that doesn't solve all of the problems - especially if you haven't got much signal feeding it anyway, as everything can get very quiet. So you add make-up gain to compensate, and the chances are then that you'll end up with signals all over the place. It's definitely better to avoid the confusion, especially as it's so easy...
So it makes sense to keep the signal normalized before you compress or limit, every time.
The other thing to note is that there's absolutely no penalty in this, assuming that you're working in 32-bit FP. But if you are working with 16-bit integer signals, it's almost essential that you normalize first, otherwise if you're not careful, you can end up with massive bit-depth deterioration. The way that works is as follows; you forget to normalize, compress a waveform, somebody says it's lunchtime so without thinking, you save the result, only the level's at -30dB. When you re-open your 16-bit file, you discover that increasing the level again brings the noise floor up, because you've only got an 11-bit signal there, not a 16-bit one. Floating Point maths doesn't suffer from this at all - it has a massive dynamic range, and even if you save a signal 90dB below 0dB, you can get it back without loss.