Old 09-16-2009, 05:18 PM   #1
Blastrio
Human being with feelings
 
Blastrio's Avatar
 
Join Date: Sep 2009
Location: Montreal, Qc
Posts: 450
Default Gain Staging and Headroom

Hey people!
I'm a bit confused with 24 bit headroom clipping etc. If I understand, in a mix engine, if you record and mix at 32 bit, it is almost impossible to clip a signal. The problem is that when you render your mix down to 16 bit the clipping will occur. There is still a lot of headroom at 24 bit, which is the bitrate at which I record and mix everything (44.1 khz), so my multi-step question is:

1- Is it really important to keep the levels under 0dB in the track meters (I don't hear any clipping even if they exceed it)?

2- Is gain staging really important (keeping the levels from clipping at the plugins ins/outs in the plugin chain(same here, I don't hear any clipping))?

3- Is it better to keep all the track levels lower and then pushing up the overall level with a limiter, or keep the levels higher and go easier on the limiting?

I seem to notice a bit of difference in mixes done in both ways in step 3. The mixes done with the track levels hotter seem to be a bit harsher, less smooth, than the ones done with lower levels and boosted with a limiter.
The thing is I don't know if it's a placebo effect, or if the limiter's own sound has something to do with it.

Thanx in advance

Blastrio
Blastrio is offline   Reply With Quote
Old 09-16-2009, 06:28 PM   #2
scottedog
Human being with feelings
 
scottedog's Avatar
 
Join Date: Sep 2007
Location: On an island in the Pacific.......in Canada.
Posts: 704
Default

Yes it's important to keep your levels under 0dbfs otherwise you'll get digital clipping. If your meters ever hit the red clip indicator that's bad.
Definitely keep recording at 24bit and don't forget to dither when you bounce down to 16bit.
I try to keep my master peaking around -6db in my mix. Bring up the level in the master so it limits at -0.2dbfs.
Also gain staging is very important, don't be hammering your plugs with high levels.
24bit has a low noise floor so you don't need to max your levels.
Ummm, that's it.
__________________
Scott.

ScottedogStudios on ReverbNation
scottedog is offline   Reply With Quote
Old 09-16-2009, 11:50 PM   #3
EricM
Human being with feelings
 
EricM's Avatar
 
Join Date: Jul 2009
Location: Ljubljana, Slovenia
Posts: 3,801
Default

The fact is that Reaper has a 64 bit engine, so every signal path
has enough resolution that you will never clip, even if the indicator
on channel says otherwise (gets red).

The only part you have to make sure not to clip is Master fader,
as there the conversion to either 24 or 16 bit will happen. And
of course whenever you use a plug-in that process at less than
32 bit resolution (but those are rare nowadays....)
EricM is offline   Reply With Quote
Old 09-17-2009, 04:35 AM   #4
antiClick
Human being with feelings
 
antiClick's Avatar
 
Join Date: Mar 2007
Location: Mediterrenean Sea
Posts: 977
Default

You just have to make sure there is no red in the armed tracks when recording (signal is not clipping in the AD stage of your audio interface).

When mixing, it won't matter because the precission of the audio engine is much higher than 24bits. It won't even matter on most pluggins.

Then before rendering just make sure the master channel doesn't exceed -0.1db


That's it
antiClick is offline   Reply With Quote
Old 09-17-2009, 03:21 PM   #5
Blastrio
Human being with feelings
 
Blastrio's Avatar
 
Join Date: Sep 2009
Location: Montreal, Qc
Posts: 450
Default

Thanx for the replies!
I seem to be getting different opinions here. I know that it's important not to clip the levels at the input when recording and on the master track, but it's at the gain staging (between plugins) and on the individual tracks that I'm not too sure. So does that mean that I don't need to worry about peak levels in gain staging and individual tracks?

Blastrio
Blastrio is offline   Reply With Quote
Old 09-17-2009, 03:41 PM   #6
Coerce
Human being with feelings
 
Coerce's Avatar
 
Join Date: Mar 2008
Location: Kent, England
Posts: 374
Default

To me gain staging is watching everything in the entire chain.

I try to keep everything around -12, the meters on the input on my adc and then the meter on the track meter in reaper. Also the whole mix in the main outputs.

Peaks will jump past that but on average leaving more headroom stops any nasty digital clipping. You can always turn the main output up in the mastering stage.

When I do vocals I do it differently leaving about 24db's of headroom. Seems to work nicely.
Coerce is offline   Reply With Quote
Old 09-17-2009, 04:46 PM   #7
Lawrence
Human being with feelings
 
Join Date: Mar 2007
Posts: 21,551
Default

Gain staging in digital daws (once you are past the ADC and before you get to a DAC) is pretty much irrelevant and nothing like analog gain staging.

Those good habits from analog don't do any harm though. Neither does going well over 0dBFS in between hardware.

Someone should just print a list of all 3rd party plugs that will clip internally at 0dBFS and make it a sticky.

Quote:
Originally Posted by antiClick View Post
Then before rendering just make sure the master channel doesn't exceed -0.1db
That's it
The common understanding is to go a little bit lower than that due to intersample peaks that don't meter as clips. -3 or so is more the norm for maximum peak level on the master bus.

Not that anyone could actually detect a few clipped samples out of thousands. But it's good practice.

Last edited by Lawrence; 09-17-2009 at 05:01 PM.
Lawrence is offline   Reply With Quote
Old 09-17-2009, 06:21 PM   #8
Blastrio
Human being with feelings
 
Blastrio's Avatar
 
Join Date: Sep 2009
Location: Montreal, Qc
Posts: 450
Default

Quote:
Originally Posted by Coerce View Post
To me gain staging is watching everything in the entire chain.

I try to keep everything around -12, the meters on the input on my adc and then the meter on the track meter in reaper. Also the whole mix in the main outputs.

Peaks will jump past that but on average leaving more headroom stops any nasty digital clipping. You can always turn the main output up in the mastering stage.

When I do vocals I do it differently leaving about 24db's of headroom. Seems to work nicely.
Damn, that seems pretty low. But how do you get your vocals to stand out if they're 12 dB lower than anything else?
Blastrio is offline   Reply With Quote
Old 09-17-2009, 06:22 PM   #9
Blastrio
Human being with feelings
 
Blastrio's Avatar
 
Join Date: Sep 2009
Location: Montreal, Qc
Posts: 450
Default

Quote:
Originally Posted by Lawrence View Post
-3 or so is more the norm for maximum peak level on the master bus.
Did you mean -.03? -3 seems a bit too much headroom no?
Blastrio is offline   Reply With Quote
Old 09-18-2009, 10:32 AM   #10
audioguy_on_ca
Human being with feelings
 
Join Date: Apr 2008
Posts: 259
Default AES -18

The Audio Engineering Society recommends that -18dBFS be equivalent to +4dBu, or 0 dBVU.

If you track with that in mind, meaning keeping the average level of the track at -18 on reaper's track meters (as long as you don't clip on peaks), you'll be capturing 23 bit files...with a possible 138 dB dynamic range, from noisefloor to peak.

The same thing applies when mixing, except you'll need to insert a trim/gain reduction plug-in on most if not all tracks to tweak things back into a range that will allow additive or subtractive processing...and still allow that 6dB of headroom on the mix buss for when you master to -0.01dBFS, and take full advantage of 24 bit digital.

Proper application of this, as well as good capture/engineering technique when tracking will make your mixes sound bloody HUGE and Clean and Punchy and PHAT, and all those good things people miss from the good old analog days. You'll find yourself using less EQ, fooling with compression a lot less.
audioguy_on_ca is offline   Reply With Quote
Old 09-18-2009, 11:04 AM   #11
Bubbagump
Human being with feelings
 
Bubbagump's Avatar
 
Join Date: Nov 2006
Location: Columbus, Ohio
Posts: 2,028
Default

Quote:
Originally Posted by Blastrio View Post
Did you mean -.03? -3 seems a bit too much headroom no?
-0.03 is fine for a master peak level... but I would be perfectly happy sending off a premaster at -3dbFS.
Bubbagump is offline   Reply With Quote
Old 09-18-2009, 11:37 AM   #12
Lawrence
Human being with feelings
 
Join Date: Mar 2007
Posts: 21,551
Default

Quote:
Originally Posted by Blastrio View Post
Did you mean -.03? -3 seems a bit too much headroom no?
Typically mix engineers leave at least that much headroom for the ME. Unless you're trying to get every last bit of volume for the loudness war it's pretty much irrelevant, and even then since peak level has little to do with RMS level and/or how loud the thing sounds. You can "L2" a thing that peaks at -6 that sounds loud as heck.

Peaking at -0.03 would be susceptible to intersample peaks that may occasionally clip the converters during reconstruction. It's more a "working practice" than anything you'd notice or hear. But there is no practical reason to even approach 0 that closely... it won't sound any better.

These discussions always waver around between hobbist/semi-pro/pro so some of the stuff gets smeared a little bit. Most professional engineers are leaving a few db of headroom for those two reasons I mentioned above, ME headroom and intersample clipping.

For most hobbyists and many people mixing at home with no intention of pro mastering, even an occasional clip at 0 on the master is pretty much irrelevant as evidenced by the fact that you don't really hear the occasional clips, you just happen to see the red light come on. For stuff not going to mastering I put on an L2 at -1 and don't worry about it.

Quote:
2. You Don't Need To Normalize Or Get It As Loud As Possible - Keep in mind most Mastering Engineers like some head room. Even having your mix peaking 2db below zero is great. I regularly turn in mixes with peaks at -12db since that's where I find my gain staging sounds good on the console I work on. The mastering engineer will have great gear to gain stage it as long as you go to a pro.
http://musformation.com/2009/06/tips...mastering.html

Quote:
Although its good practice to get a strong signal, dont let the audio clip at all! If your hitting 0db at mixdown, your pushing the levels too much. Getting your tracks to peak at a maximum of around -3dB gives the mastering engineer headroom to work with.
http://www.earwaxaudio.com/mastering-advice.php

That kind of advice is pretty common from ME's. It's minutia that was settled (in the pro realm) many years ago but it regularly comes up in Internet discussions from people anyway. In most of those cases it's far from important, a few samples clipping.

Last edited by Lawrence; 09-18-2009 at 12:12 PM.
Lawrence is offline   Reply With Quote
Old 09-18-2009, 12:19 PM   #13
stupeT
Human being with feelings
 
stupeT's Avatar
 
Join Date: Jan 2009
Location: frankonia
Posts: 1,996
Default

Quote:
Originally Posted by audioguy_on_ca View Post
The Audio Engineering Society recommends that -18dBFS be equivalent to +4dBu, or 0 dBVU.

If you track with that in mind, meaning keeping the average level of the track at -18 on reaper's track meters (as long as you don't clip on peaks), you'll be capturing 23 bit files...with a possible 138 dB dynamic range, from noisefloor to peak.
The most significant bit (say 24 in that context, even though it should be called 23, starting numbering from 0, as usual in IT...) is set to "on" when the level exceeds -6dB. The levels in the area around -0.03 dB are actually encoded in the least significant bits with all higher bits already set to 1.

Even if you'd only use 16 bit from the 24 bit converter you would still benefit from the 24 bit format in the sense of lower distortion, but as you said with lower dynamic range.
__________________
------------------------------------------
Don't read this sentence to it's end, please.
stupeT is offline   Reply With Quote
Old 09-18-2009, 01:20 PM   #14
schwa
Administrator
 
schwa's Avatar
 
Join Date: Mar 2007
Location: NY
Posts: 15,747
Default

Like Lawrence said, clipping at the track output level isn't meaningful in a 64 bit DAW. If you want to pay attention to track levels and keep them below the red that's fine, but nothing gets ruined or degraded if you run tracks at +100 dB or whatever.
schwa is offline   Reply With Quote
Old 09-18-2009, 01:28 PM   #15
stupeT
Human being with feelings
 
stupeT's Avatar
 
Join Date: Jan 2009
Location: frankonia
Posts: 1,996
Default

Quote:
Originally Posted by schwa View Post
Like Lawrence said, clipping at the track output level isn't meaningful in a 64 bit DAW. If you want to pay attention to track levels and keep them below the red that's fine, but nothing gets ruined or degraded if you run tracks at +100 dB or whatever.
Sounds like a synonym for FREEDOM
__________________
------------------------------------------
Don't read this sentence to it's end, please.
stupeT is offline   Reply With Quote
Old 09-18-2009, 01:40 PM   #16
Lawrence
Human being with feelings
 
Join Date: Mar 2007
Posts: 21,551
Default

Quote:
Originally Posted by stupeT View Post
Sounds like a synonym for FREEDOM
Yeah, I find working without the distraction of clip lights on audio channels a tiny bit of freedom.

"Give me free." - Amistad
Lawrence is offline   Reply With Quote
Old 09-18-2009, 04:43 PM   #17
Blastrio
Human being with feelings
 
Blastrio's Avatar
 
Join Date: Sep 2009
Location: Montreal, Qc
Posts: 450
Default -.3

Quote:
Originally Posted by Lawrence View Post
Peaking at -0.03 would be susceptible to intersample peaks that may occasionally clip the converters during reconstruction. It's more a "working practice" than anything you'd notice or hear. But there is no practical reason to even approach 0 that closely... it won't sound any better.
Ooops... I meant -.3, not -.03. But anyway, from the answers you guys give me, I don't think that would change anything.
The thing I was most worried about was the gain staging.
So I think I'll keep the master reasonably under 0, and squeeze some dBs out of the mix with a limiter. By the way, I don't know if you guys tried PSP's Xenon. It's a lot more transparent than the L2 and A LOT cheaper. Or even Voxengo's Elephant, but I'm digressing here.
Thanx a lot for the input guys. I think I had quicker answers here in a couple days than searching around the web for weeks.
Actually the Reaper community was one of the reasons I chose to switch to Reaper. I don't foresee looking back anytime soon.

Blastrio
Blastrio is offline   Reply With Quote
Old 09-19-2009, 10:54 AM   #18
audioguy_on_ca
Human being with feelings
 
Join Date: Apr 2008
Posts: 259
Default

Quote:
Originally Posted by stupeT View Post
The most significant bit (say 24 in that context, even though it should be called 23, starting numbering from 0, as usual in IT...) is set to "on" when the level exceeds -6dB. The levels in the area around -0.03 dB are actually encoded in the least significant bits with all higher bits already set to 1.
You're correct on both points; "bit 23, the 24th bit" and it being "on" when the level exceeds -6dBFS.
audioguy_on_ca is offline   Reply With Quote
Old 09-22-2009, 03:38 AM   #19
stupeT
Human being with feelings
 
stupeT's Avatar
 
Join Date: Jan 2009
Location: frankonia
Posts: 1,996
Default

Quote:
Originally Posted by audioguy_on_ca View Post
You're correct on both points; "bit 23, the 24th bit" and it being "on" when the level exceeds -6dBFS.
Quite counter-intuitive for non-techies, isn't it?
__________________
------------------------------------------
Don't read this sentence to it's end, please.
stupeT is offline   Reply With Quote
Old 09-23-2009, 12:24 AM   #20
EricM
Human being with feelings
 
EricM's Avatar
 
Join Date: Jul 2009
Location: Ljubljana, Slovenia
Posts: 3,801
Default

Quote:
Originally Posted by audioguy_on_ca View Post
You're correct on both points; "bit 23, the 24th bit" and it being "on" when the level exceeds -6dBFS.
This got me thinking... I thought the last bit (call it 23 if counting from 0)
is assigned value 1 when the signal exceeds -3.01 dBFS, that is when the
quantization process decides to round the last bit's value from 0 to 1.
-6.02 dBFS peak is exactly 23 bits (counting from 1) and the last bit is 0, no?
EricM is offline   Reply With Quote
Old 09-23-2009, 02:12 AM   #21
stupeT
Human being with feelings
 
stupeT's Avatar
 
Join Date: Jan 2009
Location: frankonia
Posts: 1,996
Default

Quote:
Originally Posted by EricM View Post
This got me thinking... I thought the last bit (call it 23 if counting from 0)
is assigned value 1 when the signal exceeds -3.01 dBFS, that is when the
quantization process decides to round the last bit's value from 0 to 1.
-6.02 dBFS peak is exactly 23 bits (counting from 1) and the last bit is 0, no?
Every bit is worth round about 6dB, right? This is the same as making input voltage the half. So the highest bit (23) is responsible for "more than 50% input voltage). The next (22) is half of that half again hence a quarter in total... and so on. So bit 0 is worth some nano volts, which is in the range of the ananlog noise of the converter. These bits with low numbers are encoding the 0.x dB stuff...
__________________
------------------------------------------
Don't read this sentence to it's end, please.
stupeT is offline   Reply With Quote
Old 09-23-2009, 02:48 AM   #22
politcat
Human being with feelings
 
politcat's Avatar
 
Join Date: Jul 2006
Location: stuck in transition
Posts: 1,870
Default

Quote:
Originally Posted by Lawrence View Post
Someone should just print a list of all 3rd party plugs that will clip internally at 0dBFS and make it a sticky.
is there a quick way to find out?
politcat is offline   Reply With Quote
Old 09-23-2009, 02:51 AM   #23
norbury brook
Human being with feelings
 
norbury brook's Avatar
 
Join Date: Mar 2007
Location: London UK
Posts: 3,378
Default

I have a feeling a lot of DSP card based plugs run in fixed point,which would mean they'd clip.

A good point though regarding a list :-)


MC
norbury brook is offline   Reply With Quote
Old 09-23-2009, 03:12 AM   #24
stupeT
Human being with feelings
 
stupeT's Avatar
 
Join Date: Jan 2009
Location: frankonia
Posts: 1,996
Default

Quote:
Originally Posted by politcat View Post
is there a quick way to find out?
Well, digi-clipping is so ugly, so obvious. Use your ears..
__________________
------------------------------------------
Don't read this sentence to it's end, please.
stupeT is offline   Reply With Quote
Old 09-23-2009, 03:59 AM   #25
politcat
Human being with feelings
 
politcat's Avatar
 
Join Date: Jul 2006
Location: stuck in transition
Posts: 1,870
Default

Quote:
Originally Posted by stupeT View Post
Well, digi-clipping is so ugly, so obvious. Use your ears..
searching out plugin specs (not all manuals have them) can be time consuming so I was wondering if there's a way (a utility?) to find out if a plugin is 64 bit or not

Last edited by politcat; 09-23-2009 at 04:03 AM.
politcat is offline   Reply With Quote
Old 09-23-2009, 09:16 AM   #26
schwa
Administrator
 
schwa's Avatar
 
Join Date: Mar 2007
Location: NY
Posts: 15,747
Default

This free plugin can tell you what resolution a plugin is processing at: http://www.stillwellaudio.com/?page_id=33

schwa is offline   Reply With Quote
Old 09-23-2009, 09:34 AM   #27
Lawrence
Human being with feelings
 
Join Date: Mar 2007
Posts: 21,551
Default

Quote:
Originally Posted by norbury brook View Post
I have a feeling a lot of DSP card based plugs run in fixed point,which would mean they'd clip.

MC
Not necessarily. PTHD for instance runs at 48-bit fixed point / double precision and while it's internal clipping point is nowhere near where it would be in floating point (as it relates to levels on the daw meter) it's still far above the levels where anyone would reasonably push a signal. Not sure exactly but +40 or thereabouts over 0 on the meter I think. Someone else can give a precise level but it's not 0 on a 24-bit metered scale.

The only thing that clips at 0 in a daw is something doing raw 24-bit processing internally which afaik, is no daw or hardware DSP these days. If TDM plugs hold true to that internal depth they shouldn't clip at digital 0 internally either. Of course I don't have PTHD so I cannot test that for myself.

If they process at 24-bit fixed then yeah, they'd clip at 0. But don't confuse the 24-bit fixed scale of the meters with even a more dynamic internal fixed scale that goes beyond what the 24-bit meter is showing. The meter is set to 24-bit fixed to represent the limits of converters, all daws internally go far beyond that... even fixed point daws and I suspect all DSP hardware like UAD (floating point) or Duende (?).

Quote:
A good point though regarding a list :-)
Yeah. Put Waves Linear at the top of it.

Edit =========

You can see below why internal clipping (on a 24-bit meter's 0 point) shouldn't occur even in fixed point beyond 24-bit. Here it doubles the dynamic range to 288db for dsp processing.



So even in PTHD fixed point there is almost 48db of headroom above 0 before clipping occurs. The arrow marks 0 on the 24-bit scale, or the daw meters.



I hope that clears up some common misconceptions.

Last edited by Lawrence; 09-23-2009 at 10:14 AM.
Lawrence is offline   Reply With Quote
Old 09-23-2009, 11:47 AM   #28
politcat
Human being with feelings
 
politcat's Avatar
 
Join Date: Jul 2006
Location: stuck in transition
Posts: 1,870
Default

Quote:
Originally Posted by schwa View Post
This free plugin can tell you what resolution a plugin is processing at: http://www.stillwellaudio.com/?page_id=33

cool. thx!
politcat is offline   Reply With Quote
Old 09-24-2009, 02:47 AM   #29
politcat
Human being with feelings
 
politcat's Avatar
 
Join Date: Jul 2006
Location: stuck in transition
Posts: 1,870
Default

it appears the only plugin I use that's 24 bit is T-Racks3 Brickwall Limiter. Shock! same thing with the shell itself

all the other T-Racks3 plugins are 32 bit. ???


politcat is offline   Reply With Quote
Old 09-25-2009, 09:26 AM   #30
imMute
Human being with feelings
 
Join Date: May 2009
Location: Duluth, MN
Posts: 97
Default

Quote:
Originally Posted by stupeT View Post
Quite counter-intuitive for non-techies, isn't it?
There are two types of people: 1) those who start array indexes at one and 1) those who start array indexes at zero

imMute is offline   Reply With Quote
Old 09-29-2009, 05:05 AM   #31
stupeT
Human being with feelings
 
stupeT's Avatar
 
Join Date: Jan 2009
Location: frankonia
Posts: 1,996
Default

Quote:
Originally Posted by imMute View Post
There are two types of people: 1) those who start array indexes at one and 1) those who start array indexes at zero



On the other hand the bit's number calculates it's value:

Bit 0 is 2 power 0 = 1
Bit 1 is 2 power 1 = 2
Bit 2 is 2 power 2 = 4
Bit 3 is 2 power 3 = 8

And that's a good reason to start at zero...
__________________
------------------------------------------
Don't read this sentence to it's end, please.
stupeT is offline   Reply With Quote
Old 09-29-2009, 08:18 AM   #32
airon
Human being with feelings
 
airon's Avatar
 
Join Date: Aug 2006
Location: Berlin
Posts: 11,817
Default

More fun numbers.

Every time the floating point number exceeds 1.0 or goes below 0.1 the exponent shifts, giving you a change of precision on one end of the 53(52+1 actually) bit mantissa. This is a 20 dB shift in fact, and thus a little over three bits of precision.

So if you cross the 1.0 line (0dB FS in the tracks) a little over three bits of precision die a truncation death at the bottom of the 53 bit number that is your audio. Muahahaha! So no matter what happens, the precision will always be greater than the mixbus of Protools. I think. Probably.

Are you scared yet? Think it's gonna blow your chances at a hit record ? You're right !!!!! Your cat'll probably kill you in your sleep for playing with your levels in this fashion.

Amateur fears are so easy to exploit. That's how PR works for DAWs .
__________________
Using Latch Preview (Video) - Faderport 16 setup for CSI 1.1 , CSI 3.10
Website
"My ego comes pre-shrunk" - Randy Thom
airon is offline   Reply With Quote
Old 09-30-2009, 08:40 AM   #33
audioguy_on_ca
Human being with feelings
 
Join Date: Apr 2008
Posts: 259
Default

Quote:
Originally Posted by airon View Post

So if you cross the 1.0 line (0dB FS in the tracks) a little over three bits of precision die a truncation death at the bottom of the 53 bit number that is your audio. Muahahaha! So no matter what happens, the precision will always be greater than the mixbus of Protools. I think. Probably.

Are you scared yet? Think it's gonna blow your chances at a hit record ? You're right !!!!! Your cat'll probably kill you in your sleep for playing with your levels in this fashion.

Amateur fears are so easy to exploit. That's how PR works for DAWs .
Right, and people have to get used to the fact that overly hot tracks make things sound like a$$.

Just make music. record it, mix it, get it out there. if it has legs, it will run. keep it simple, let it breathe.

Check this out: https://www.youtube.com/watch?v=KO2vY1brMDU

The streaming audio is bad, but I'm sure you'll get the idea
audioguy_on_ca is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -7. The time now is 05:29 AM.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.