Is there a difference in final quality if you bounce before?

Is there a difference in overall project output sound quality if you bounce all your tracks before you master your project? Or would quality stay the same if say, you kept all of your tracks frozen? I get anxiety about bouncing my tracks sometimes because sometimes I will want to make small adjustments here and there up til the final final product.. I wasn't sure if having your tracks frozen and mastering it yourself on top of that would make a difference in overall project sound quality, in regards to computing power that it takes..

 

I just want to make sure that I'm not losing any sound quality by mastering a track myself with all the channels frozen, like if my computer won't be able to handle all that at once or something?

It's not clear wether you are talking about mixing or mastering? In any case, bouncing results in quantization errors, which is a form of distortion, while freezing does not. But FWIW neither bouncing nor freezing should be necessary prior to mixing down a project.

In any case, bouncing results in quantization errors, which is a form of distortion....

 

There's something I don't get here David.

Quantization error (or quantization distortion) should already happen when recording an analog signal (say a guitar) into Logic isn't it, since an analog-to-digital conversion is done during the process, right? So bouncing an already converted analog signal should not produce any further Quantization error, right?

 

Thanks

Quantization error (or quantization distortion) should already happen when recording an analog signal (say a guitar) into Logic isn't it, since an analog-to-digital conversion is done during the process, right?

Right.

 

So bouncing an already converted analog signal should not produce any further Quantization error, right?

That's only right if you bounce the exact same guitar recording without doing ANY processing in Logic, not even changing its level - which would give the same result as if you duplicated the audio file in the Finder.

 

But the instant you do any processing in Logic (for example mixing that guitar track with a vocal track, or adding a plug-in to the guitar track, or adjusting its level with the volume fader), you are working with a 32 bit floating point signal, and bouncing that signal to a 24 bit fixed point audio file results in new quantization errors.

But the instant you do any processing in Logic (for example mixing that guitar track with a vocal track, or adding a plug-in to the guitar track, or adjusting its level with the volume fader), you are working with a 32 bit floating point signal, and bouncing that signal to a 24 bit fixed point audio file results in new quantization errors.

 

Ah, good! Thanks for this very clear and interesting explanation David! Now I'm gonna do my homework to understand better the floating point vs fixed point thing!

You're welcome!

While David is correct and I would hate to disagree, I feel strongly that I should add a point to the quantization discussion that hasn't yet been discussed.

 

At 24-bit, the quality loss is theoretically there, but insignificant for any practical purpose. The noise would be 144dB down from full scale. No recording or playback equipment that I know of even has a 144dB dynamic range; if I recall, the best 24-bit converters are only like 120ish dB. Even if it did, our *ears* don't have that much usable range - that's more than the range between "inaudible" and "lifetime deafness in seconds."

 

So personally, if it's more convenient workflow-wise to bounce, I'll do it and not worry about any quality issues. Turning up or down one of your faders 0.1dB will make more of a difference to the sound than the effect of a bounce.

 

After all, Bohemian Rhapsody was bounced numerous times in the analog domain and it still became a hit. Or another analogy: real studios use analog patchbays even though the increased cable length increases noise, because it's worth the workflow benefits.

Interesting points M.M.

Nowadays more than ever, we can see what we can't hear. But, naturally, what really matters is what we can hear not what we hear looks like on a waveform. So If we like what we hear, all the rest doesn't matter at all. The beauty of things is subjective.

So personally, if it's more convenient workflow-wise to bounce, I'll do it and not worry about any quality issues. Turning up or down one of your faders 0.1dB will make more of a difference to the sound than the effect of a bounce.

 

+1

So DAWs that render to 32bit files don't introduce quantization distortion?

32-bit floating will still introduce quantization errors but it's a different behaviour than fixed point.

You will get a constant SNR of -156 dB below the actual sample value instead of below 0.0 dBFS.

So DAWs that render to 32bit files don't introduce quantization distortion?

Freezing a 32 bit floating point signal into a 32 bit floating point audio file does not introduce quantization distortion.

I don't know who to listen to!!!

 

 

Still, I doubt you could hear -156dB above the thermal noise of a playback system, so I won't worry about it

You can listen to both of us.

David is talking about saving a fp32 process as a file.

I'm talking about the actual process.

You can listen to both of us.

David is talking about saving a fp32 process as a file.

I'm talking about the actual process.

 

Huh?

 

Huh?

Eric said quantization errors do occur when making an operation in 32 bit floating point. I'm saying that by freezing a track, therefore saving to a file the 32 bit floating point signal, you're not introducing any more distortion than you already have when you're not freezing the track. You're just putting the results of the operation (along with its quantization errors) in an audio file so you don't have to calculate that operation in real time.

Huh?

Eric said quantization errors do occur when making an operation in 32 bit floating point. I'm saying that by freezing a track, therefore saving to a file the 32 bit floating point signal, you're not introducing any more distortion than you already have when you're not freezing the track. You're just putting the results of the operation (along with its quantization errors) in an audio file so you don't have to calculate that operation in real time.

 

Ah!

 

I get it now. Thanks!