When to normalize?

  • Thread starter Thread starter dafduc
  • Start date Start date
dafduc

dafduc

New member
I just finished a choir recording project - vox plus pipe organ - and really found myself with a distortion problem.

I usually like to get everything close to the max (I leave 1-3 db headroom) so I'm hearing consistent levels and roughly similar speaker response to the final product.

I think I let myself get blindsided by the wide dynamic range on this - anyhow, wound up having to keep reducing levels through the chain, and had a really muddy product at the end, plus I had let some distortion creep in. Few things worse than digital distortion on a choir...

So my questions - when do you guys normalize? More than once? What kind of headroom do you like to have to start with? Does the genre of music change your answer? would you handle a live recording differently?

Thanks...
 
1. I do not normalize.
2. see 1.
3. I like to track around -3dB
4. No
5. N/A

Sounds like you need compression in general, more specifically, a limiter. Some people like to track w/ compression, I prefer to compress after tracking. I track to PC, so my path is

Mic > Preamp > Limiter (compressor set at 1:inf, fast attack) > PC

I use the limiter to limit transient peaks, allowing more head room to maximize the signal without transients clipping the the signal.

Now if you say the vocals have a great dynamic range, just make sure that the loudest sustained passages (musical peaks, not transient peaks) peak under the threshold of the limiter so you don't affect the music content of the signal too much.

I'm not sure how other feel about normalizing, but I have no use for it.

http://www.prorec.com/prorec/articles.nsf/files/F0057F0FA68D59DF8625664B00130E57

Check out the normalizing section at the bottom of the page.

I bet vox + pipe organ sounds awesome...
 
Thx for the link.

Compression is supposed to be a giant taboo for classical music. Or maybe I misunderstood the rules?

Anyhow, I always saw norming as simply a smart volume control (I probably made things weirder by trying to do RMS).

I'll put an mp3 up if you're interested (I'll probably have to do 64k mono, though, to get the size down...)
 
If you use a compressor as a limiter described earlier, you are not compressing the music (prob the no-no to which you are referring), but the occassional (hopefully infrequent) freak spikes that clip your A/D converters.

https://homerecording.com/bbs/showthread.php?threadid=38797&highlight=limit*+transient*+tracking

That thread may help a bit. Basically, you are following the purist philosophy of trakcing - no effects. A limiter used in the function described is not as much an effect as it is a way to maximize the signal w/o clipping or affecting the musical content, thus minimizing noise.

Do some deeper research into limiting and compression (I do not compress when tracking! The effect is irreversible.)

One more thing to check out: The Behringer compressor manuals are a great introduction to compression and may explain the different applications of compression better.
http://www.behringer-download.de/MDX1400/MDX1400_ENG_Rev_E.pdf
Check out pages 11-24
 
couldn't have said it better myself. The only time I would ever compress is as a last resort to patch-up a mistake (cold levels). It's still going to sound aweful, but you'll be able to hear it (and all of it's horrible noise floor). I use a hardware limiter while tracking - I'm sure most people probably do.

There are also some great resources on this board for understanding compression / limiting / expanding. It's very important knowlege, and worth the research effort.
 
I've found this board to be an invaluable resource, along with the internet in general.

Thanks to the operators, moderators, and contributors of audio wisdom.
 
Here's the choir and the organ (a Casavant 17-rank):

Gloria Patri, from the Magnificat (Vivaldi)

96K stereo mp3, about 1.4 meg. Highs seem to have been boosted my the mp3 encoding (compared to the wav, which is 26mb, so I won't be uploading it :D ), but the wav was a bit bright, too. My bad.

Anyhow, thanks for the info, guys. I'm still not sure I understand the objection to normalizing - if all it is is a volume control pegged to a fixed reference point (which is what I think the linked ProRec article was saying), how am I better off using a volume control than normalization? BTW, I defeated the built-in compression, if that makes a difference.

Again I avoid limiting and compression (for choir work, that is) because they squash my dynamic range and raise my noise floor. MY concern was that, by normalizing too early in the chain, I didn't leave enough headroom and introduced distortion with subsequent processing (resampling, Acoustic Mirror, EQ).

For s&g, I might take it all the way through the chain again, save the norm for last (or replace it with a volume boost) if I have time.
 
the kind of normalizing it sounds like you are doing is RMS normalization...are you using sound forge?

a properly designed PEAK normalization will never distort, if you dont normalize past around -0.3. Using RMS to anything higher than -12 db will almost certainly distort.

The objection many have to normalizing is because of math errors introduced, BUT, if you are mixing digitally, normalizing is often the exact same thing as simply changing gain on the digital ( or computer) mixing console, so youd get math errors either way, but dont always need to be afraid of them
 
dafduc said:
Again I avoid limiting and compression (for choir work, that is) because they squash my dynamic range and raise my noise floor.
Click to expand...

Effective limiting should not squash.
 
Just buy a 100 000$ 2" reel to reel, and a 10 000$ limiter and your problems are history.
 
Thanks for the nice comments.

Sound Forge 6.0, -16db RMS, switch to 0 peak norm if clip (choices are compress or 0 peak norm). Given your note about -0.3 or lower to avoid clipping, that 0.0 peak might have done it.

Here's the whole chain:

pair Sound Room MC-012's x-y's -> M-Audio AudioBuddy -> VS-840 stereo 16 bit /44.1 khz.

VS-840 S/PDIF -> Delta-66 -> SoundForge 6.0.

Adjust bit depth to 24.
Resample to 96khz.
Norm to -16 RMS (0 peak norm if clip)
SF Noise Reduction
Smooth -1 (a questionable decision, thought sound was too hyped)
Acoustic Mirror (add church space)
Readjust bit depth to 16
Reresample to 44.1k
SF Click/Crackle removal (for reresample errors)
SF Clipped peak restore (should've done at 96k - or redone chain to lose clips)
EQ (you heard the 3rd take - still overhyped to my ears)

Doing it again, I'd probably save the norm and noise reduction to the end, do a -1 peak norm instead of an RMS, skip the smoothing, and be way more careful with the EQ. Live-n-learn...

Happy New Year, everyone!
 
Does anyone have comments on 44.1kHz <> 96kHz conversion vs. 44.1kHz <> 88.2 kHz conversion?

44.1kHz <> 88.2 kHz conversion might give you better sounding results, but I think that is much debated.

Happy Occidental New Year
 
I can't see why you would even use the Roland recorder in the chain. While the Delta 66 isn't exactly 'high end', the recordings I have heard using them for A/D conversions certainly have had a much more realistic sound than any VS units I have heard!!!

I am not sure WHY you feel the need to "normalize" your audio! If you are mixing in a digital environment, it is wise to keep things away from digital peak levels. So much of the DSP out there does not handle audio approaching the 0db levels as well as it should, and leaving headroom is greatly desirable.

In addition, after reading all the processes you did to the audio, man!!! You sure have a very high faith in the integrity of DSP!!! Sample rate conversions, bit depth conversions, "smoothing", "acoustic spaces", etc..... I can't possibly see benefits in most of the processes you described!

In this kind of situation, I would have skipped the Roland altogether. I would have found the best stereo mic postion possible for the choir, and the same with the pipe organ. I would have INCLUDED the "church space" as part of the sound I recorded. Your Delta card has 4 inputs, and I would have gone into them at the highest bit depth and sample rate it will support. I would have adjusted RECORD LEVELS to be as close to the MIX LEVELS as was possible. I would have adjusted mic's to capture as closely the sound I wanted. I would have skipped almost EVERY process you did to that stuff. Up sampling and up bit depth conversions aren't doing you a whole lot of good in this situation, mainly because you will be going right on back down to that in the end.

I would suspect that it was all the format conversions, and all the funky DSP crapola you did that created MANY places that distortion could be introduced. I NEVER normalize tracks. NEVER. I never have found a need to do so. I see absolutely NO point in normalizing digital audio. Decracklers and Denoisers are great is there is something that went terribly wrong in a recording, but I certainly would try to avoid using devices that FORCE me to use these audio degrading processes!

Choir music depends upon hi fidelity for it to come out right. The LESS you do to the audio the better. Capturing the audio the way you want it to sound is the best way to keep it sounding good. This is true in most genre's of music, but is a MUST in this genre of music! Point blank, if you are not in the ball park of what you want in the sound, you have messed up in the capture of the sound, and no amount of DSP after the fact is going to save it. In fact, DSP is going to muck things up even worse in ambiant recordings like choir music.

Not trying to be hard on you, it is just that you possibly need a little kick in the butt in terms of HOW you are working on this project. Simple is better, and it seems that you have introduced WAY too much processing for your own good on this project. Most of that DSP is just not needed!

Ed
 
dafduc said:

......
Adjust bit depth to 24.
Resample to 96khz.
Norm to -16 RMS (0 peak norm if clip)
SF Noise Reduction
Smooth -1 (a questionable decision, thought sound was too hyped)
Acoustic Mirror (add church space)
Readjust bit depth to 16
Reresample to 44.1k
SF Click/Crackle removal (for reresample errors)
SF Clipped peak restore (should've done at 96k - or redone chain to lose clips)
EQ (you heard the 3rd take - still overhyped to my ears)
......
Click to expand...

Holly Sh##!
What is all this...stuff, for?
:eek:
What Sonusman said...
:)
 
Here's the reasoning behind the process:

1) This was a live concert. We were not recording in my studio. The VS-840 is my field recorder, so when I go into the field, that's what I record with. There's no way I'm dragging my PC along! My next rig might be a laptop-based studio, but for now, it's VS-840 for the field, PC for the studio.

2) This was a volunteer gig - although I'm past the what-does-this-button do stage, I'm very much a newbie at this, so these are my chance to learn and grow (and you guys are part of the growing process - so BRING IT ON! I'm learning from you all, even on the what-the-hell's-the-matter-with-you posts).

3) There were performance and space problems - Choir didn't blend too well, organiste and choristers missed some notes, sanctuary had a sort of gymnasium sound to it, and a high noise floor. Because the final product was a vanity recording for the choir members, I sacrificed historical accuracy for a bit of sweetening, though I tried to minimize the amount of processing used to accomplish that. Our director has the historically accurate recordings - hopefully, he'll kick our tenoro profundo in the ass... talk about your "army of one!"

4) The upsample/downsample was at the suggestion of a lot of the articles I've read - and I'll agree the post-downsample results are disappointing. It was glorious at 96K though! Given the clicks that got introduced, I'll definitely try 88.2 before I do 96 again - thx for the tip.

5) The rest of the processing was there to solve particular problems that I heard. I'll admit that my "solutions" may have introduced new problems, particular the "smoothing". Because I have the raw recordings, I can go back again to try to improve it, but I probably won't have time - classes start Tuesday, and we're back to overtime at work again.

I appreciate the input.
 
I can definitely understand the use of the VS as a field recorder. It would be a cold day in hell before I dragged my studio rig out anywhere.

I always thought the VS units were pretty good though, less the somewhat weak pre's which explains the AudioBuddy (I've worked a bit with an older 1680). But doesn't the 840 record in 24 bit?

Good choice of mics IMO (wish I had a pair matched by the Sound Room)

Here's how I might have done it:

- 012's > Audio Buddy > hardware limiter > VS-840 (hot as possible to kill the noise floor an avoid normalizaion)

- VS-840 S/PDIF -> Delta-66 -> SoundForge 6.0. 24 bit if available
===OR===
- VS-840 ANALOG -> Delta-66 -> SoundForge 6.0. to 24/88.2

- no longer necessary to adjust bit depth, resample, normalize, run Noise Reduction, process clicks/crackles or clipped peaks.

- Acoustic Mirror (no arguments here)

- EQ (you heard the 3rd take - still overhyped to my ears)

- Mixdown to 16/44.1

To be honest, I didn't think your mix sounded bad, but I'm sure it could sound better.

Another quickie tip - if you encode MP3's with VBR instead of a fixed khz, you end up with much, much better sounding MP3's (and many times, they're smaller than encoding at 128kbps.)
 
Hey, I have been doing live stereo recordings for a while now, about 3 years. My honest opinion is this: you setup and get the best sound you can. Period. Sometimes it may not sound as good to you, but 95% of the time the performers/conductor think it is great. Remember you are not going to achieve that movie sound.

Also try and remember what the rooms sounds like in your head. If you thought it was an absolutely amazing performance and souded incredible, the recording should reflect that. Also remember that reverb makes or breaks the recording. So mic placement is a huge issue. The trick is getting enough blend without to much reverb. Try to get at least one rehearsal or get the group to do one song to let you experiment with placement.

Vocal groups generally sound good with a long reverb. Bands and orchestras do not because they get all muddy. They sound best with moderate to less reverb because of their nature.

By the way, your recording had a nice balance to it between the organ and choir. It sounds like maybe you could have put the mics back a little further and it would have probably come together better.

Encode your MP3's at a higher bitrate, like 128 or 160

Beezoboy
 
Thx!

One last point I forgot - I was IN the choir, so had to record on set-and-forget. And the director asked me to do this after the last location rehearsal, so no chance for a dry run.

Good news is that we use that space a lot. Perhaps less good news is that they're doing construction now, could be a very different sonic space when they're done.

Newer 840s might do 24-bit, but not my dinosaur. I don't think. Guess I don't know what those "quality" settings represent (maybe a better question in the Roland forum) - I use standard rather than high (or is it "normal 2" rather than "normal 1"?). Maybe it IS bit depth? Anyhow, I lose too much record time when I go to the highest settings. Maybe I need to get the 250mb upgrade, but I hate throwing money after old equipment.

Oh, and yes, mics will go back farther next time. Think that might bring my noise floor up a bit more?
 
So you are converting from 44.1 to 96khz? That is a big no no. There is no reason to upsample because there isn't any high frequency info in the original recording. You can never get higher quality or resolution than the original tracks.

Don't convert the sample rate until the end of the project if needed.

Don't use noise removal plugins unless the noise is absolutely, completely distracting.

Always record at the highest possible quality.

Every time you do anything to your tracks you risk reducing the quality.
 
You must log in or register to reply here.
Back
Top