For edits, why work with sample format at 32 bit?

Hi all,

I apologize for harping on this particular thing but I would just like to clarify - why is it important to work in 32 bit mode when making edits? To my understanding, 32 bit mode is default because Audacity is made to avoid as much audio quality loss (in the exporting process) as possible. I was advised to work in 32 bit mode to use the “add label at selection” feature of Audacity and I thought to ask why this was necessary as (and please correct me if I am wrong), using that feature does not affect the db level of a given project.

Thank you for any answers.

It’s not 32-bit. It’s 32-floating and it’s not an export problem. If Audacity didn’t do that, the first time you applied a filter or effect that caused the blue waveforms to go over zero or 100%, that’s the end of the show. Distortion City.

32-floating doesn’t overload. You can apply an overloading effect to your show, bring the waves back down later with another effect, and you’re good.

Audiobook Mastering makes good use of this. When you apply RMS-Normalize, it’s a good bet some of the tips of the waves will sail over zero and normally be damaged or lost. The next step in mastering is Soft Limiter which gently brings the waveforms back down to normal with little or no audible distortion.

However. The instant you try to export a WAV or other conventional sound file, you better make sure everything is technically perfect with no overload anywhere. Most sound formats do not support going over zero.

View > Show Clipping and look for red marks in your show. Those are overload points.

Koz

Thank you. If you see a PM from me about this same thing, please disregard it. I will definitely use the show clipping feature to be sure but I don’t think I’m going to be causing any overload as I am just adding labels. I understand that in general, should I want to work more in-depth, it would only make sense to use the 32-floating mode. I am going to use it for everything, though.

I know I am being OCD but at the same time I am just trying to keep my CD rips in tact as degrading them in any way would defeat the purpose of using a secure ripper etc.

I don’t think I’m going to be causing any overload as I am just adding labels.

It’s easier to do than you think. Import a good quality, uncompressed WAV file that gets close to overload, but doesn’t hit it. Export as a medium quality MP3. MP3 creates cleverly hidden sound distortion (on purpose) and that distortion can change the volume of the work pushing it over 100%.

What is your actual job? If you open an MP3 and just export to a new one, the quality of the new one will be lesser than you started with. Never do production in MP3.

Koz

I am not making edits for a job. I am editing files ripped from a CD with dbpoweramp. I rip them to AIFF. I would never use an MP3 as I know using lossy formats creates all kinds of annoyingness. I have been trying to figure out why my rips end up having incorrect tracking starting and ending times. I honestly think it has to do with the read offset of my CD drive but no one in the CD ripping world seems to agree, even though it is logical to think that a drive with a positive read offset number of samples (reads the CD too early) would then affect what is audible in my rips. :confused: I am supposed to not be able to hear the effect of the offset, but I do. I can see how some of the flaws can go unnoticed, as they can be quite minuscule, but they are audible when they shouldn’t be.

The best way I know to fix this is by using the label function of Audacity to find the correct point in a track to use as its beginning or ending (depends on the song). I align any number of tracks end to end and then zoom in until I find the points that the tracks are supposed to be separated at.

If when I export the edited files, they are somehow degraded quality-wise, I might as well not bother editing them because making 1:1 copies of what is on a CD is the reason why I even spend money on CDs. I know certain levels of degradation are not easy to hear, but knowing that there is degradation kinda makes me go nuts.

You know CDs don’t have song titles. right? When you play a CD, the player tries to go on-line to figure out what the song titles are. If I take a CD out to my pickup with no internet connection, the songs are Track 1, Track 2, Track 3, etc.

There’s another trick to this, too. Almost every time you Export, Audacity adds a dithering (noise) signal to the show. That’s to keep the downconverting errors (32 > 16) from lining up and becoming audible. If you Import a song, don’t do anything to it, and then Export to the same format, I believe you can turn dithering off. It’s never going to do anything. However. If you change formats, I don’t know what the rules are.

Technically, I believe both WAV and AIFF are 44100, 16-bit. With AIFF, as you noticed, having song titles and other info. Maybe someone else knows.

Koz

I know CDs don’t have track titles. The song titles aren’t an issue because dbpoweramp fetches the song titles from the internet (similar to what you are saying).

Yes, AIFFs are 44.1/16. I know that I can turn dither off. The root of my confusion has to do with that, actually. If dither is off and I’m editing instead of processing, how is it that exporting to the same format can create overloading? Would I even need to worry about my sample format set to 32 bit-floating if I am not doing anything that causes overloading? I know it seems like the answer is in my question but I ask because I was told by Steve to add the labels to my files with 32-float set as default.

Or am I mislead and the adding label feature is actually a form of processing/content modification?

Would I even need to worry about my sample format set to 32 bit-floating

I think you’re changing more stuff than you need to. If you don’t edit the actual music, just turn dither off. 32 Float is the Audacity default and it becomes completely transparent under those conditions. 16 > 32 > 16.

Koz

Okay. I know that when the files are exported, they are 16-bit anyway, so that’s that. Thank you for your help.

If I export a 16 bit file in 32-float mode with dither turned off, won’t the file be affected quality-wise?