When I export tracks to 32 bit wav files, then import the files into Izotope or back into Audacity, they’re totally clipped.
Please see attached screen shot, showing the original track above, and the exported then imported track below.
What am I doing wrong?
I haven’t changed my workflow in Audacity, but now I am using a 32 bit recorder, and over-recording tracks for sure. I normalize on import though, and haven’t had this problem until recently.
Thanks in advance, I’m completely stuck at this point.
Which 32? There’s 32-bit Signed PCM and 32-bit Floating. Audacity uses 32-bit Floating internally to avoid overload damage and clipping during filters, effects, and processing.
32 Signed may be an invention of Promotion and Publicity. Even studios don’t go that high. They usually settle on 24-Signed. 16-bit Signed is the format for Audio CDs and most people.
What’s the goal? Why did you run into his problem?
Tho goal is location recording 2 stereo tracks on a Sound Devices Mix Pre 6 in 32 float so I don’t have to worry about over-recording. Then import and normalizing, mixing and normalizing again prior to exporting for spectral repair in Izotope rx8.
I can’t reproduce this by simply importing a stereo file (into a 32-bit project) then exporting to 32-bit WAV. The imported WAV is identical to the stereo track in the project.
What do you mean by “normalizing, mixing and normalizing again”? What are you mixing?
Sorry to be slow, I’m not getting notifications on this thread somehow.
What I mean by “normalizing, mixing and normalizing again”:
I have 2 stereo tracks recorded in 32 bit float @ 48k. Both over-recorded, so looking all red on import.
Normalize both stereo tracks.
Mix and render to new track, with one stereo track at about -6db.
Normalize the mixed track.
The mix track looks fine in Audacity. If I export as a 32 bit float wav then immediately import, the track is totally blown out as shown in the attachment.
Now happening on tracks that weren’t ever over, or mixed. Just imported, worked on (trimming, label track) then exported.
Ugh.
Also, just realized that notifcations were going to junk folder. Thanks Bill for trying to help, I ought to be more responsive now
Does it sound blown out? I think you may be dancing right around the volume where Audacity turns the waveform red. All blue and then with a very tiny increase in volume, everything turns red, but the sound quality doesn’t change.
Audacity doesn’t really sense clipping. It assumes clipping because normal waveforms don’t go right up to 100% without going over. So that’s what the warning really means. “These waves are dangerously close to overload damage.”
Where are you getting the original work from? Why are you recording it intentionally hot? I think that’s dangerous because if you are causing overload on the analog side of the capture system, that’s permanent sound damage.
I know of one lavalier microphone system that records work at two different volumes to greatly reduce the chances of a bad recording.
Sincere thanks to all who’ve replied to this thread…
It looked blown out, but I didn’t listen to it. After normalizing, big chunks of the track had a flat top.
Really useful to learn how Audacity regards waves, that’s super interesting.
I ended up going back to the source files and redoing things in 2.4.2, with good results.
Koz, I use a Sound Devices Mix-Pre 6II to record live music using 2-4 Schoeps mics plus a SBD feed.The deck is designed with huge headroom on the preamps so that one can make 32 bit location recordings and just over-record to ensure good levels in post. I never get 2 takes, so want to make sure levels are high enough. Never had this issue w 2.4.2. I’ve never heard any analog clipping using this technique the past couple years.
Ran into this again tonight, using 3.0.0 because 2.4.2 is broken in ways that cause headaches.
I record 2 stereo tracks in the field using 32bit float @ 48khz.
Import as 4 mono tracks, make into 2 stereo tracks.
Mix the 2 tracks and normalize the mix to -0.1.
The track looks like pic1.
Export the track as a 32 bit float wav.
(In real workflow, import to izotope rx for spectral repair, but for purposes of this thread, skip that step.)
Import it back into audacity and it appears much louder, looks like it clipped or compressed. Sounds louder, see pic 2.
