DAW (Software) Headroom...how does it work ???

[mark4man]

OK...I'm not hearing clipping & I've got the output overloaded by 4 & 1/2 dB or more. Now...I've always been under the assumption that...for any DAW/audio editor...because of the floating point audio engine...clipping will not occur as long as the signal strength doesn’t exceed the headroom of the AI/sound card (& not, the multi-track software)???

The 32-Bit FP signal is truncated to 24 at the output to be sent to the AI/sound card...but what comes off is at the bottom of the range, correct? The peak of the signal always floats below 0dB, with the truncation at the bottom of the scale varying up & down as the entire range floats. So, when the master faders are adjusted so as to allow the output to technically clip (& if we rendered the file...it would), but if we're playing back we're just turning up the floating point signal to the D/A...& if the hardware can handle it...that's fine.

Am I on the right track, here?

Thanks,

mark4man

 

[kylen]

...The peak of the signal always floats below 0dB...

On my system - Sonar5 32bit float native file --> 64bit float busses --> +3.9 dbfs peak output of track directly to sound card at 24bit truncated by the audio app --> soundcard clips --> audio distorts.

If I let the soundcard clip the audio ( > 0dbfs or thereabouts) it's refered to as an 'over' meaning greater than 0dbfs - the digital ceiling for all intents. I don't know which 'endian' we're talking off the top of my head concerning the digital 'word' but anything past 0dbfs gets clipped in my house!

 

[mshilarious]

Am I on the right track, here?

Thanks,

mark4man

For that to work, the app would have to look ahead at all the overs to attenuate the whole signal accordingly.

If you clip your master bus, you've clipped the playback, and the audio if you render/save. Clipping will not occur in the internal engine, but once you leave the app (to a soundcard, etc) . . . it will clip.

 

[mark4man]

On my system - Sonar5 32bit float native file --> 64bit float busses --> +3.9 dbfs peak output of track directly to sound card at 24bit truncated by the audio app --> soundcard clips --> audio distorts.

For me...SONAR5 is more forgiving than WaveLab...but the whole idea behind floating point math is to provide infinite headroom inside the DAW (or at least...one of the main ideas, anyway.) I'm not hearing distortion...it's just not there. See...I know the headroom is there...I just don't know why. It could be one of two things. It could have something to do w/ the clipping level of analog amplifiers in the gear being 26dB over 0 VU (using a K-20 scale)...putting that ceiling @ +6dB. Or it could be that, what leaves the DAW will theoretically never distort (since it's FP based); & that by increasing that level in the DAW, you're just making it louder. But the second idea doesn't comport with the fact that the signal is converted to fixed integer at the output.

I've read posts where the user brings a full scale, undistorted track into the workstation...turns up the track gain by 6dB & lowers the headphones by the same degree & the signal sounds the same...no distortion. Then they do the same thing with track level...& then w/ the master fader...& get the same result...which is...infinite digital headroom, until the headroom in the audio interface starts to go.

This is driving me nuts.

Clipping will not occur in the internal engine, but once you leave the app (to a soundcard, etc) . . . it will clip.

Then why ain't I hearin' it, bro ??? There has to be something more to this (something more than...never, ever go over 0dB or you'll be sent to bed w/ no supper.)

mark4man

 

[TimOBrien]

What you think you HEAR means diddly.

Render out a .wav file, then open it up in an audio editor (Audacity is free) and look at the peaks.

 

[BigRay]

What you think you HEAR means diddly.

that is a bit counter intuitive, isnt it??

HEaring is everything. of course if your ears suck all bets are off.

 

[mark4man]

What you think you HEAR means diddly

Hearing is EVERYTHING (well...almost everything)...if I'm not hearing it on my studio's system (Lynx Aurora8/AES16 > KRK 8" ref monitors)...(& my hearing isn't shot)...I don't think it's there. My ME used the "clip the converters" method on my CD to up the avg. RMS...& I've played the ref on every imaginable system & it sounds fine.

Render out a .wav file, then open it up in an audio editor (Audacity is free) and look at the peaks.

That I understand...I'm talking about what leaves the DAW for the interface.

mark4man

 

[RedStone]

I've been using REAPER lately - which is basically impossible to clip internally - you can pin every single track to +24dB, so that every track is widly distorted when the master fader is at unity gain.

BUT, lower the master down a ways (to lets say, -30dB) and the clipping stops ... voila! no more distortion ...

You wouldn't want to do that as a regular practice since not all plugins work with 64-bit internal processing