The No.1 Website for Pro Audio
 All  This Thread  Reviews  Gear Database  Gear for sale     Latest  Trending
Audio clipping & 32 Bit Floating Point Processing
Old 13th May 2010
  #1
Here for the gear
 

Audio clipping & 32 Bit Floating Point Processing

Hello all,

I have heard that there is something called 32 bit floating point processing that I assume gives a program like Logic's internal environment "unclippable" headroom, which is why I presume there are no real clipping indicators in Logic? (e.g. a red light that stays on when any of our tracks have clipped.) Anyway, my question is if we accidentally do hit 0db while recording a vocal track, for example, is there still cause for concern because we have overloaded the A/D converters for that brief moment regardless of 32 bit FP?

I have been in different studios where I have seen engineers pushing the recording levels to the max, even occasionally clipping into the red according to their DAW's level indicators. This however doesn't seem to concern them at all. So then I was curious as to why.

I guess I'm not too sure how 32 bit floating point processing applies to setting our recording levels, or even watching levels during the mixing stage for that matter. Regardless of the benefits of 32 bit FP processing, should we all still be tenacious in avoiding 0dbs whether tracking or mixing?

Thanks for any advice!
Old 13th May 2010
  #2
Lives for gear
you don't actually record in floating point

the A/D conversion remains integer

the floating point benefits kick in with post processing

so even if you save a file as 32 bit floating point during the A/D process you can still clip with ease

while mixing/editing it is possible to exceed 0dB without damaging the end product . . . at some point prior to creating a distribution file you just pull levels, on track or mix, down to where ever you want
Old 13th May 2010
  #3
I think we might need to get a little clarity in terms here...

Clipping means damaging the signal by truncating all values above a certain level. In a system that has an absolute headroom, such as AD or DA processes, this results, crudely speaking, in wave forms flattened razor sharp at the top... get enough of those -- and it doesn't take many -- and you can end up with some nasty and generally quite unwanted distortion.

So, in a digital system where such truncation occurs above 0 dBFS (full scale), we can say, reasonably enough, that it will clip above 0 dBFS.

But in a system with elevated headroom (where 0 dB is calibrated below the top of the full scale range or in an analog system which goes into a gradual saturation eventually resulting in clipping), clipping does not necessarily occur simply from crossing the nominal 0 dB point.


At any rate, oretez is right in suggesting that, in our conventional AD and DA processes where absolute top of scale is correlated to 0dBFS, the math is, indeed fixed point and, so, the fixed ceiling of 0 dBFS will cause clipping if one tries to 'exceed' it. There is also the less-understood issue of intersample peaks. (You can read a good explainer about them here, and DL a free intersample peak-aware meter plug in from the nice folks at SSL.)


And he's also correct in suggesting that in floating point DSP and DAW systems, once the signal is in the digital doman and converted into a floating point format that can 'float' it's accuracy range across a more or less infinite range of abstract values, 'clipping' is essentially irrelevant. (Assuming proper design, of course.)

The fly in that ointment is the possiblity that a plug in may not use internal floating point or may be designed with a 'sweet spot' or other optimal range of operation.

For that reason, many advocate staying on the 'safe side' and adhering to 'normal' gain staging concepts even inside the relative freedom afforded by FP systems.
Old 14th May 2010
  #4
16942
Guest
Quote:
Originally Posted by theblue1 View Post
The fly in that ointment is the possiblity that a plug in may not use internal floating point or may be designed with a 'sweet spot' or other optimal range of operation.

For that reason, many advocate staying on the 'safe side' and adhering to 'normal' gain staging concepts even inside the relative freedom afforded by FP systems.

Which is exactly what I advise in my book.
Old 14th May 2010
  #5
Lives for gear
 
dualflip's Avatar
 

Quote:
Originally Posted by PunkSmurf View Post
Hello all,

I have heard that there is something called 32 bit floating point processing that I assume gives a program like Logic's internal environment "unclippable" headroom, which is why I presume there are no real clipping indicators in Logic? (e.g. a red light that stays on when any of our tracks have clipped.) Anyway, my question is if we accidentally do hit 0db while recording a vocal track, for example, is there still cause for concern because we have overloaded the A/D converters for that brief moment regardless of 32 bit FP?

I have been in different studios where I have seen engineers pushing the recording levels to the max, even occasionally clipping into the red according to their DAW's level indicators. This however doesn't seem to concern them at all. So then I was curious as to why.

I guess I'm not too sure how 32 bit floating point processing applies to setting our recording levels, or even watching levels during the mixing stage for that matter. Regardless of the benefits of 32 bit FP processing, should we all still be tenacious in avoiding 0dbs whether tracking or mixing?

Thanks for any advice!
Its actually easy:

If the clip indicator lights up while recording, you have a clip, simple as that.

If you are mixing, and the channel clip indicators light up but the master fader clip indicator doesnt light up, theres no clip at all!, this of course doesnt include clips that may happen in the plugins, so be careful with that cuz some plugins go back to 24 bits after processing. Meaning, make sure that no plug-in is clipping, if no plug in is clipping but channel fader is clipping, and the master fader is NOT clipping, then theres no clip.

Theory says that the master fader is the only place were the signal can clip (as i said this doesnt apply to recording) since all the internal processing is done at 32bits (depending on the software) so at the output of the master fader is were it gets back to 24 bits.
Old 25th February 2012
  #6
Old 19th September 2012
  #7
Gear Head
 

Can someone explain this for me, just for the hell of the experiment today in cubase I set my project to 32bit and imported a Commercial CD track and converted it to 32bit on its way in too (not that I mix in 32bit) anyways i noticed that the track was clipping.... but I know for a fact this track don't clip in 24bit so I set my project back to 24bit and sure enough it does not clip. so why does 32bit clip an already mastered track? what's happening there. Sorry I tried to research this one but couldn't put 2&2 together.
Old 19th September 2012
  #8
The 24 bit file can't show clipping because a fixed file can't go above full scale, but the contained signal could actually be clipped. The converted floating point file is probably referenced to some level above full scale and you can simply lower it without degradation. Try testing with a 24 bit bounce of a 1k test tone at e.g. -12 dBFS, convert it to float, and see what happens to the level when played back at unity.

BTW you are mixing in 32 bit floating point in Cubase, though your source files may be 24 bit. They expand to 32 bit floating point when processed unless you're using a fixed point plug-in from e.g. Waves. At the summing or post fader stage they'll expand to 32 bit float in any case.

Anyway, to the gentleman poster of this thread I suggest you read this thread (and with that name, please upload a forum avatar):
Guide to Levels in Digital Audio
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Forum Jump
Forum Jump