The dirty secret of Digital Adders...

  • Thread starter Thread starter donpipon
  • Start date Start date
D

donpipon

New member
Perhaps this is a well known issue amongst engineering community... but some home-studio owners might not be completely aware of...

I've recently discovered (because I'm a newbie after all...) an amazing bug when mixing down several tracks of digital audio on a desktop-computer-based DAW:

Unless youre using the latest HD accel core... the ALU system on most (each and every...????) audio cards, is rather shitty.... I had doubts in the past but now that I've made some experience I have absolute certain of this..


What does this mean??

that when youre mixing multiple tracks on a mixing software on PC (and it doesn't matter if you have the best of the quedruple-octuple-core-superpowermac...) the summing of this audio tracks to a single 2-track master fader produces a hard, brickwall digital compressing effect, very audible and harmful for the final audio quality. ALWAYS. The only thing you can do to avoid this is to perform a mixdown peaking at -10dB (or less!!) on the final .WAV file .AIFF or whatever...

Do the test for your own!!! take any of your mix on any multitrack software and perform 2 mixdowns: one peaking near the ceiling (-1, -2 dB) and other at -6 or -10, and hear the difference!!! the reason is simple: Theres no concept of headroom on digital audio, like there is on analog consoles. yes even at 24 or 32 bits resolution... the problem is the same: the budget on digital adders is always poor on desktop computers sound cards. And if you're planning to take your mix to a professional mastering studio, the final level of that mixdown can really make the difference!

Ive been recently to some mastering studios to clear any doubts: they ask for -6dB peaking digital mixdowns.

Did you know this secret?
 
donpipon said:
take any of your mix on any multitrack software and perform 2 mixdowns: one peaking near the ceiling (-1, -2 dB) and other at -6 or -10, and hear the difference!!!
Click to expand...

That's not true for any DAW I know of, and it's not a valid test either. The only way to compare those two audio files is to reduce the louder mix to the same level as the soft one. Then you can not only listen more fairly, but you can also do a null test to see if both files are the same.

I set up a very similar test just yesterday in SONAR, to demonstrate at my Audio Myths workshop for the upcoming AES convention in New York. In this case I set up two tracks with the exact same mixed tune on each, and placed an EQ plug-in with a simple 1 KHz cut on both tracks.

One track's volume was set to 0, and the other was boosted by 18 dB which you might think would clip the output bus or at least clip within the EQ. On the boosted track I lowered the EQ's output volume by 18 dB to compensate.

If "digital adding" were defective, soloing one track versus the other would sound different. Not only do both tracks sound identical, the results null perfectly proving both tracks output the exact same data.

--Ethan
 
Ethan Winer said:
it's not a valid test either. The only way to compare those two audio files is to reduce the louder mix to the same level as the soft one.
--Ethan
Click to expand...

That's so obvious I didn't even think of making it clear... of course you have to level both mixes!! leveling if the Nº1 rule. and I talk about leveling with test tones, not with your ear...

And.. Please describe me in detail which are the "DAW's you know", or at least the ones you have performed this tests on.


I can swear you.... you can hear a diference!! on any set of monitors, loud, soft, close field, large 3-way main, from cheap loudspeakers to state-of-the-art studio monitors.. the diference is AUDIBLE.


Off topic: My favourite "Audio mith" is this one:
"0dB peaking on digital audio is not THAT necesary..."

what do you think?
 
donpipon

Take the audio files you are comparing, align them perfectly, and then reverse the phase on one of them. You will literally hear the difference between the two. If there is no difference the two are identical. (the null Ethan spoke of)
 
dean1964 said:
donpipon

Take the audio files you are comparing, align them perfectly, and then reverse the phase on one of them. You will literally hear the difference between the two. If there is no difference the two are identical. (the null Ethan spoke of)
Click to expand...

I think none of you understood my point... I talk about reducing the individual fader gain of Each and every Track, (better if there are tracks grouped in submixes) until the Master output peaking tops on -10dB...
NOT reducing the MASTER 2-TRACK GAIN!!
audio files will be significantly different... because the mix IS diferent.

And the more levels of "sub-mixing" you do... the more accurate the digital summing will be (for example: group "rhythm guitars" then group "lead guitars", then group this 2 groups together on a single stereo "guitars", and so on... ) DO the test, hear the diference and then tell me what you think.

This ONLY works on DIGITAL AUDIO, inside a DESKTOP PC (no outboard D/A conversion) with EXTERNAL USB/IEEE audio cards and some PCI cards e.g. MOTU .

PS: personally, I heard best transient response, broader freq. ends (low and high) wider stereo and clearer dynamics. (like the whole mix was more "raw" or something)
 
donpipon said:
Unless youre using the latest HD accel core... the ALU system on most (each and every...????) audio cards, is rather shitty....
Click to expand...
Mixing software does not perform any mixing or summing on the sound card, but rather uses the mobo CPU to do the mix math. If you want to test that, use an external USB-based interface to get sound in and out of your PC and remove the sound card entirely. You'll still be able to mix just fine and just exactly like you could with the sound card in. In fact, you can even remove the USB device and the sound card and the mixing will still take place and take place identically. You won't be able to hear anything unto you enable an output, of course, but the math will still take place.

Now, if you're using your sound card for MIDI generation or VSTis, then that's a different ball game altogether, and yeah, a sound card with a supplimental DSP chip can help with the sound quality of those instruments. But that's not a function of the mixer summing itself.

Second, you probably should not be creating mixes that peak anywhere near 0dBFS on the raw mix to begin with; this is often a leading indicator that you're running something too hot upstream or in your mix levels. This is probably not directly affecting your results as described here, though it can be a warning of upstream clipping or preamp overdrive.

And the reason mastering studios ask for lower levels is so they have some headroom with which to work. Give them stuff that peaks at 0dBFS and they have no room to maneuver. Plus, they also know that high levels typically (or at least often) means the beginning client ran the gain structure too hot somewhere along the line like described above.

What's possible you may be hearing (though you'd probably have to have a golden ear for this) is the D/A converter in the sound card. there has been some excellent testing done that shows that some poorer quality D/A converters can distort some between ~ -3dBFS and 0dBFS.

Keep your levels down, yes. But not because your sound card is summing badly, but because the use of quality gain structure through your recording and mixing chain will in and of itself provide better S/N ratio, better dynamics and less distortion of both the analog and digital type.

G.
 
100% false on a properly designed 32 bit float DAW (which is the majority of modern DAWs, meaning this millennium), and yes, I have tested mine with mix bus levels at silly highs, such as +48dBFS mix with a -48dB reduction on the master fader. It really is as simple as addition (mixing)and/or multiplication (gain) followed by division. In fact I can even write you a plugin that mixes if you like; the operation is addition.

On fixed-point DAWs it is possible to clip the mix bus at very high levels, so don't do that. But I am not aware of a modern DAW that is clipping with the mix bus at -1dBFS peak.

You do lose some precision with a float bus, but it still returns 24 bit precision which is likely far better than your analog dynamic range.

I suspect if you have a problem it would be attributable to some plugin in your chain that does not properly handle levels over 0dBFS (plugs should hand off 32 bit float data to each other). And of course if you clip the output of your master fader, nothing will help you. But you didn't say you did that.
 
Analog is better than digital.
Plasma is better than LCD.
Glass is better than wood.
Corvette is better than a traktor.

My god is better than your god.
 
WhiteStrat said:
I have a problem with that one.
Click to expand...

Have both

RedCorvette.jpg
 
donpipon said:
The only thing you can do to avoid this is to perform a mixdown peaking at -10dB (or less!!) on the final .WAV file .AIFF or whatever...
Click to expand...

A peak of no more than -10dbFS sounds perfectly reasonable to me. :confused:


donpipon said:
Off topic: My favourite "Audio mith" is this one:
"0dB peaking on digital audio is not THAT necesary..."

what do you think?
Click to expand...

Screw 'not that neccessary', peaking digital audio at 0db is a terrible idea. Period.
 
legionserial said:
A peak of no more than -10dbFS sounds perfectly reasonable to me. :confused:




Screw 'not that neccessary', peaking digital audio at 0db is a terrible idea. Period.
Click to expand...

He's talking mixdown, not tracking. You hear a lot of CDs peak at -10dBFS?
 
SouthSIDE Glen said:
Second, you probably should not be creating mixes that peak anywhere near 0dBFS on the raw mix to begin with; this is often a leading indicator that you're running something too hot upstream or in your mix levels. This is probably not directly affecting your results as described here, though it can be a warning of upstream clipping or preamp overdrive.
Click to expand...

Don't worry, I'm not that stupid....


SouthSIDE Glen said:
Keep your levels down, yes. But not because your sound card is summing badly, but because the use of quality gain structure through your recording and mixing chain will in and of itself provide better S/N ratio, better dynamics and less distortion of both the analog and digital type.
Click to expand...

Finally!!! something that sounds reasonable! thanks!!

now, whatever the reason might be... this "compression" effect -within your DAW- is REAL and not subtle at all, in fact is very audible.
and also... not real, or much less... on analog mixing,
This sounds so stupid and obvious... yes.. but... does everyone has this in mind in every mix down??

and A/D, D/A conversion, like someone posts above, is also VERY audible for any trained ear, at least... between a cheap outboard (no offense to presonus :cool:) and good quality converters (LUCID, APOGEE, etc)
 
You must log in or register to reply here.
Back
Top