DAW mix bus overloading?

[Kiwiburger]

OK - inspired by Phil's suggestion in a previous thread, I've just done this basic torture test in Cubase SX.

 

Basically - I seem to be in the minority here. I don't believe you can "overload" a 32 bit floating point digital mix bus. Others have disagreed.

 

I created a 5 second 50 hertz pure sinewave, peaking at 0dBfs. Unfortunately this was a 16 bit file - but I imported this into Cubase SX (converting to 32 bit) and it sounded and looked like a pure 50 HZ hum.

 

Next - I cloned this to create 32 identical sine waves.

Then - for good measure - I maxed out each and every fader. In Cubase SX, for some reason this gives a 6.02 dB boost to each channel. I also maxed out the master fader (another 6.02 boost) to add insult to injury.

 

The resulting clipped waveform going to my 24 bit converter was a fruity square-wave-sounding buzz.

 

I exported this as a 16 bit 44.1 kHz stereo file, and it sounded just as fruity.

 

Then ... I exported a 32 bit floating 44.1 kHz stereo file - which also sounded exceedingly clipped (due to my 24 bit converters).

 

However ... I reimported this file back into Cubase SX. For comparison, I also reimported the original sine wave.

 

Now I roughly calculate that the Cubase SX mix bus has added about 200 dB of abusive gain to this pure sine wave (i'm not really sure how to calculate what i've done - but 32 x 6 = 192 for a start).

 

The waveform was looking rather square on top - but I decided to level the playing field and perform another abuse on this summed waveform. I normalised it to 0dB.

 

Now - I expected that the poor sinewave to have been completely munted by this stage, and it was sounding fairly buzzy.

 

But - I realised that conversion to stereo had added another 6dB, so it was actually still clipping after the normalisation process.

 

I manually adjusted the levels of both the original, and the munted tracks so that they were both peaking at exactly -3dBfs.

 

Then - I flipped the phase, and guess what I heard ...

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

absolute silence.

 

 

Now somebody please explain to me very slowly how can I get my master bus to overload. I just can't seem to do it ...

[UstadKhanAli]

Kiwiburger or others: do you happen to know what kind of a mix buss Pro Tools 5.1 has? Just curious.

 

Also, I have noticed that if I slam the levels into the red on the Master Fader, I can plainly see that the peaks are leveled off on AIFF file.

 

But if I lower the Master Fader, the clipping goes away on the AIFF file. Kiwiburger, this is what I believe you mentioned in the other thread.

 

Just the same, I lower the volume of all the faders in a manner that Where mentioned to me a while back since I don't really understand the math behind all this.

 

Thanks.

[Kiwiburger]

I'm no expert, but I consider 32 bit 'floating' to be a bit like scientific notation on a calculator.

 

If you keep on multiplying numbers on a calculator, you would normally run out of digits. Like if you have 8 LED's you can't normally exceed 99999999. But with scientific notation you can exceed this number and still represent it using the 8 LEDs.

 

So by summing 32 tracks of full scale sine wave (doubling the volume each time - i.e. adding 6dB gain) and also adding extra gain, what I created was a huge sine wave that must have peaked over 200dB if you could actually graph it.

 

The graphs in DAWs just don't go that high, which is why you can't see what you are doing. But mathematically, the data was just a huge pure sine wave.

 

The confusing thing is when you send that to your A/D converters, which aren't designed for floating point maths. They just truncate it to 24 bits, hence the square wave. But the audio file it'self is not clipped or distorted. In theory there are some minute rounding errors, but they are insignificant.

 

I still don't believe you can "overdrive" a floating point digital summing bus. But you can certainly overdrive a D/A converter. The coloration on my AD2496 was very obvious above -6dBfs.

 

 

[Kiwiburger]

As I understand it, Protools has a 48 bit fixed mix bus. This is probably better than floating, providing you don't actually clip it.

 

As good as floating point is (and I have no complaints), in theory the noise floor is constantly pumping. If you use fixed numbers that are large enough to sum your biggest number of tracks, then I don't believe there is ever any problem.

 

I think Traction has a 64 bit mix bus, which seems like a really good idea to me.

 

So with ANY of these DAWs, I still don't think you can "overdrive" them. Maybe you could clip, but until you actually did this I don't think there would be any audible signal degredation.

 

When your hear the sound getting fuzzier, that's all happening in the A/D converter - as far as I can tell.

 

If you mix totally in the box, as I do, there is nothing to worry about, provided you don't clip.

 

Obviously when tracking, the coloration of the A/D converter isn't desirable, which is why you shouldn't use the top 6dB going in.

[Kiwiburger]

I know there are mathematicians who understand this very well, and will shoot down my lame laymans interpretation of this. All I care about is making music with the tools I have - and right or wrong, my ears and eyes tell me I have nothing to worry about with the mix bus in Cubase SX.

[Danny (NS::U)]

I once tried an experiment with a signal generator, busses, and the output of Cubase SL. No matter what I did, pulling the master fader down "declipped" the signal.

 

If you've got a big enough accumulator at the end, you can keep adding numbers for a long time. The real problem occurs when there's another accumulator even farther down (your D/A) that can't handle numbers that big.

 

I would wager that, these days, most DAW software is written with ginormous buss accumulators. If you're clipping an output, pull that master down a couple of dB and be happy again.

 

You cannot get away with this on an analog console. (If you clip that summing buss, that fader's just making that distorted signal a bit less loud.)

[Zooey]

 

Originally posted by Kiwiburger As I understand it, Protools has a 48 bit fixed mix bus. This is probably better than floating, providing you don't actually clip it.

 

 

Pro Tools LE (Mbox, 001, 002, 002R) is 32 bit floating point and always has been.

 

Pro Tools HD has a 48 bit fixed mix buss with a 56 bit accumulator. It does not clip under any real world mixing scenario (though it is theoretically capable of clipping). Everyone I have talked to who claims to have clipped it turns out to be clipping the DA converter instead.

[Lee Knight]

 

Originally posted by Kiwiburger So with ANY of these DAWs, I still don't think you can "overdrive" them. Maybe you could clip, but until you actually did this I don't think there would be any audible signal degredation. When your hear the sound getting fuzzier, that's all happening in the A/D converter - as far as I can tell. If you mix totally in the box, as I do, there is nothing to worry about, provided you don't clip. Obviously when tracking, the coloration of the A/D converter isn't desirable, which is why you shouldn't use the top 6dB going in.

 

 

I'm all ears when it comes to this stuff and...

 

...in my Pro Tools LE ITB world, if you don't clip the master buss, it sounds right. Even if you bring down level with a plugin on the master buss, things get blurry.

 

You sound much more investigative than I. Cheers to you! Well done, and thanks.

 

I'm fine if I keep my master buss from overloading, including any master plugin attenuation, that I avoid big time.

 

Mixes are bigger it seems, when I keep them down a tad.

[Kiwiburger]

I think there are still good reasons to treat DAW tracks as though they are analog - mainly the nature of certain plugin designs. A lot of plugins are designed to emulate analog - certainly anything with saturation or non-linear coding - so they might not behave as expected above 0dBfs.

 

Obviously interfacing with analog gear the D/A and A/D converters have to be given a level with optimum headroom.

 

But I personally think that there is no logical reason to be scared of making use of the available headroom in DAW tracks or master bus. Graphical waveform displays and meters and faders are designed for maximum focus on the last 6dB range. Being afraid to use this means you are aren't making full use of the visual and ergonomic range of meters and faders.

 

Obviously, if you make use of that range, you will have to pull the master fader down. I don't see a problem with this - and I don't think you have to insist that it be left at zero.

 

I think analog summing boxes are being sold for the wrong reasons. One reason given is that you can now make use of the full range of the track fader. I think that's a false reason.

 

If anything - using an external analog bus will introduce much more distortion. But maybe distortion is a good thing - that's a whole different issue.

 

In the box summing is good enough for me.

[Lawrence Farr]

 

Originally posted by Kiwiburger The confusing thing is when you send that to your A/D converters, which aren't designed for floating point maths. They just truncate it to 24 bits...

 

 

The daw's output busses are 24-bits. There has been some discussion about whether to let the daw truncate it or whether to dither. I always dither the 32-bit float audio to 24-bits (without noise shaping) unless I'm running multiple stems from the daw to a digital console. That would take too many plugs.

 

Even while monitoring an ITB mix I have a 24-bit dither across the LR bus in SX.

 

Lawrence

[where02190]

 

Originally posted by Lee Knight I'm all ears when it comes to this stuff and... ...in my Pro Tools LE ITB world, if you don't clip the master buss, it sounds right. Even if you bring down level with a plugin on the master buss, things get blurry. You sound much more investigative than I. Cheers to you! Well done, and thanks. I'm fine if I keep my master buss from overloading, including any master plugin attenuation, that I avoid big time. Mixes are bigger it seems, when I keep them down a tad.

 

 

Theory is one thing, but like Lee, I'll trust my ears over the math anyday. Listeners don't calculate music, they listen to it, and if they don't like what they hear, conciously or unconciously, they won't continue to listen to it. Like Lee, my experience is that if you watch your levels going to the masterbuss, and keep them in check, mixes sound smoother, fuller and more pleasing than ignoring this.

[Kiwiburger]

When listening, you have no option but to hear through D/A converters and analog monitors. I don't think anyone is arguing that you can definately overload or overdrive analog stuff.

 

I think the problem is when this analog thinking is applied to decisions made entirely in the digital domain.

 

For example - some people are very strong on the idea that you must keep your DAW channel faders low (avoiding the top 6dB range, where your graphs and faders have more visual and ergonomic. I'm suggesting this may not be necessary - and the sinewave test convinced me there is no reason to work with this handicap.

 

Provided plugins don't have headroom issues, and provided the final A/D converter is not overdriven, I don't believe there is any audible difference in sound.

 

In short: I still don't believe you can overdrive a digital summing bus. Just analog stuff, or digital emulations of analog stuff.