I am recording old cassette tapes into Audacity, which I then process a little (with say normalise or amplify). Sometimes I want to save my project before finishing the processing. My question is why would I save as an audacity project (.aup) if .wav is a lossless format? Saving work in .aup project directories/files seem unecessary to the my uninitiated mind. In an attempt to understand can I ask the following:
Presumably I can save a new recording as .AUP and reload it into Audacity as many times as I want with no loss due to saving.
If I export a recording as .wav, then reload into audacity and re-export as .wav again, is there a loss? Can I do that infinitely with no loss?
What if I alternate between saving the project as .aup, then reloading it, and then exporting as .wav, then reloading it, then saving as .aup project etc. Is there any loss by switching between these two lossless save/export formats each time I want to save my work to be continued later?
Thanks, sorry for the simple questions, I just have a hunch that .wav is not quite lossless in some of these scenarios during processing phase before a final cut, otherwise why mess about with saving audacity .aup projects?
Presumably I can save a new recording as .AUP and reload it into Audacity as many times as I want with no loss due to saving.
You’re not saving an AUP. You’re saving a Project and the AUP file is merely the manager of the project. The show stuff is in the _DATA folder. Keep them together. Audacity Projects do not save UNDO.
Can I do that infinitely with no loss?
No. Audacity works internally at the insanely high quality 32-floating format. There is always some conversion damage, but it’s reeeeely tiny. Going into Audacity is really lossless, but coming back out isn’t.
What if I alternate…
Any time Audacity makes a 16-bit sound file, there is that tiny damage. You can eliminate all the damages by doing everything in 32-floating.
We recommend exporting sound files periodically as you work to guard against the computer going into the mud and taking all your work with it. You can also Save Projects as different names as you go.
Also please note that Audacity has a mode where it doesn’t make personal copies of external sound files. That can be dangerous and I think our default is to include them. It used to be default not to and people would “clean up” all those external music files and the Project would drop dead.
Thanks for your swift answers. Do I conclude correctly then that saving work in progress as an audacity project is less lossy than saving work in a .wav file? If I repeatedly save work in a .wav file does it lose more each time I reload and export it (compared to if I had saved it using a project)? Or is it only the first time I export the .wav after recording? Perhaps I am asking if the audacity project saving keeps the 32 bit format when saved and reloaded? And if the .wav is constantly being converted between 16bit and 32 bit each time it is exported and reloaded into Audacity again?
Sorry if these are stupid questions.
Going from 16 to 32 is lossless. The quality of the target is very much higher than the source, but the return trip you’re going “downhill” to a lesser quality format, so Audacity has to “guess” at some of the data values.
And again, we’re talking vanishingly small errors in Music CD quality sound – 44100, 16bit Stereo PCM.
Do all the production in 32-floating including the protection exports if it bothers you.
If your goal is a Music CD, then that’s going to be 44100, 16-bit Stereo. There is no other option for Music CD. If you’re headed for MP3, the compression quality is going to be the determining factor, not the WAV file that made it.
And all of these formats are pure gold compared to any sound you’re likely to get from a cassette.
OK thanks, I appreciate the general points you are making but I would really like to understand specifically what is the purpose and advantage (why / if it is actually superior) to save ‘work in progress’ using audacity projects rather exporting to and from wav files, in between edits.
Is it because the files saved in audacity projects are 32 bit whereas the wav files are 16 and therefore need to be converted between 16 - 32 each time they are exported and reloaded into the tool?
Thanks Trebor - those links have helped quite a bit. Without labouring the point I am concluding that audacity projects are not only processed in memory as 32 bits, but audacity projects are also saved to file in 32 bits, unlike wav, and this is where the dithering problem for wav comes in.
Sorry to all for being a pain, but this is mostly all new to me, and I like to understand things myself rather than just do what I’m told
This leads me to (hopefully) my final question: Is there any reason NOT to use 32 bit wav as an export format for work in progress, other than file size?
If this can be successively loaded and exported with no accumulation of noise, or any other degradation, then it seems (at from a quality perspective) audacity projects do not add anything useful, but rather complicate matters (seeing as they proliferate their own directory structures and are not readable or as portable as wav files, and are more susceptible to corruption).
File size is the main “down side” of 32 bit float.
The (only?) other downside is that few audio programs support 32 bit float format, but that is irrelevant for “work in progress” because Audacity does support 32 bit float format.
If you are working with a simple one track project, I’d recommend going with WAV format (for the reasons that you state).
For more complex projects, the Audacity Project format (AUP file and _data folder) has the advantage that it can support multiple tracks, label tracks, track pan and gain and envelopes.
Just before we drift from this too much, the reason a tiny dither signal is added is to keep the bit conversion distortion from going up. When you process from higher bit density to a lower one, the musical accuracy goes down. You could have musical tones that just don’t convert well – or convert with rolling distortion. It’s not just “stop using dither and all your problems are over.” They’re just different, more serious problems. Dither means it’s difficult for any conversion error to exist long enough to be heard.
32 bit floating is not nearly as common a file type as 16-bit is. If you sent a 32-float sound file to your cousin in Philadelphia and they didn’t have Audacity, it would be something of a crap shoot whether or not they could play it.
Music CD format is 44100, 16-bit, Stereo and basic digital television is 48000, 16-bit, Stereo. 16-bit is not a dreadful bit depth to use, but you may decide to go higher for production and personal use. I don’t know anybody who shoots songs and music in less than 24, but many of them have no idea why. I believe Windows has trouble with 24 bit, so that can get you into trouble, too.
The bottom line here for the relatively straightforward tape transfers is:
do all your recording and editing in 32-bit 44.1KHz stereo
If you need to store your work while you are working on a particular project - the Save an Audacity Project - this wiil be lossless and you won’t get dither applied (as you have probably worked out by now).
When it comes to final production then Export to 16-bit 44.1 kHz PCM stereo WAV (can be used to make CDS and will play on most players). Dithering is applied by default here - you can turn it off in Preferences - I don’t, by choice I use “triangular” dithering. This is the only stage in this simple process when you will get dithering.
Personally I normally never bother saving Audacity projects for this type of task, I just record into a fresh empty project do all the processing and export 16-bit WAVS at the end and then close the project without saving.
The other key part of my workflow is to make backup copies of these production WAVs (before I load them into iTunes and convert them to AAC for my iPod).
If you are a purist and very thorough (and you have plenty of disk space) you may want to make the optional step of immediately after capture Export a 32-bit WAV of the raw unprocessed recording (and back it up). That way you can always go back to it later for re-processing if required.
A couple of other bits of advice relevant to tape transfers:
Do clean the tape deck transport mechanism, the capstan and the rubber pinch roller that presses the tape to the capstan - over the years they tend to accumulate oxide dust that is shed from tape usage. Don’t use alcohol-based solvents on the rubber parts!
Befeore recording, wind the tape completelt end-to-end and nack again - this evens out the tension in a tape, particularly if it has been stored for a long period.
Thank you Steve, Koz. and Wax. for your further answers. The fog surrounding the purpose of audacity projects files, and the drawbacks of 16 bit wav has cleared significantly from my mind, and for that I am very grateful. Based on your information, I will replace all use of 16 bit wav format (which I was using for masters, and backups) with 32 bit wav format. Disk space is no concern and I mostly do not need to manipulate multi-track projects, so with masters and W.I.P. in 32 bit wav, I can cut ‘disposable’ mp3 and 16 bit CD versions for real world listening, and recut them indefinitely into the future from the master 32 bit wav, without loosing quality regardless of how many times I import or export, and I don’t need to retain any audacity project files to achieve that 100% lossless quality.
One question on saving 32 bit wav - I noticed that when I export to “'other uncompressed files” and look in the options dialogue, with ‘Header:’ field set to “WAV (Microsoft)”, there are at least 2 different 32 bit options I can choose for the Encoding: field -
32 bit float
Signed 32 bit PCM
…which of these two would you recommend, or does it make no difference to quality during import/export cycles?
Can I assume with either of these 32 bit wav encodings I will get totally lossless import/exports and there will be no dither applied (or do I have to change a setting to switch off dither?). If so, which would you guess is the most likely to be readable by IT sytems into the future ? Thank you for all your information and patience.
Audacity works in 32-float. Straight PCM (pulse code modulation) offers a fixed collection of digital bits and sound levels and it’s up to you to fit your show into that. Float, in general, means the system will assign bits and levels to you wherever they’re needed. Koz
So, back to my quesion then - is Wav 32 float format superior to MS PCM 32bit wav for saving audacity projects (if they are still going to be worked on again and will be opened, processed, and re-exported repeatedly)? Presumably either produces the same size files, but do you loose information on export to PCM compared to float due to the fact it is not a floating format? What if you open export to float, open again, export to pcm etc. repeatedly between the two, will the two translate into each other perfectly no matter how many times the export is performed?
Also, my other question - is dither automatically off when exporting to a 32 bit wav file, or do I need to switch it off manually in the settings.
Logic would suggest it’s automatic, because there would be nothing to dither if audacity works in 32 bit and the export is to a 32 bit format.
“32 bit float” is the format that Audacity uses internally. It is an extremely high quality format.
If you export as “32 bit float” (WAV or AIFF) then the exported file will be identical to the Audacity track.
You do not need to turn off dither.
Export > Import > Export > Import > Export > Import > … as many times as you like in “32 bit float” format and there is no change to the audio data. The copy is perfect.
Dither is only applied when exporting to a lower bit format than the original. As Audacity works in 32 bit float, that means that dither is applied when exporting to a “lower” bit format. The maximum amount of dither (shaped dither) that is applied is:
32 bit float > 16 bit = less than -70 dB peak (-82 dB rms - “sounds like” about -90 dB)
32 bit float > 24 bit = less than 138 dB (-145 dB rms)
32 bit float > 32 bit (integer) = none, but there will be rounding errors less than -192 dB peak (-216 dB rms)
As a rough guide, if you are playing your music (normalized close to 0 dB) at a volume equivalent to heavy truck traffic, anything below -90 dB is completely and totally inaudible.
Because Audacity uses “shaped” dither by default, the dither noise is primarily in the very high frequency range where low level hearing is less sensitive, so you get about another 10 dB, giving the “sounds like” figure.
Dither noise does “add up”, so if you repeatedly convert from 32 bit float to 16 bit with dither, then to 32 bit float and repeat several times, the dither noise will gradually increase and become audible even at normal listening levels. Ideally you should only convert down to 16 bit once. In practice you can do it a few times with no audible loss of quality, and with a low quality recording (anything on cassette or vinyl) you can probably covert dozens of times with no audible loss because the noise level inherent in the original recording will swamp the dither noise.