I’m following the guides to digitize some cassette tapes, and they both recommend exporting a “raw master” as a WAV file using 32-bit float format. However, I’d rather not use all that storage space if I don’t need to, and FLAC doesn’t support floating point (and even 32-bit integer support isn’t so great).
Is WavPack 32-bit float sufficient for this type of backup? As far as I can tell, if I unpack it to WAV, the file is identical, but I’m just wondering if there’s anything else I’m overlooking that makes WavPack unsuitable for this purpose?
I actually wrote those tutorials some years ago with that 32-bit float recommendation (just to get the highest available quality) …
however when I was working myself digitizing my LPs and tapes I made do with 16-bit WAV files for my raw masters. And for my backups of the finished edited files too.
Thanks for writing those guides, they’re very helpful.
I like the concept of the unprocessed master backup, so you can go back to the original if needed (since Audacity isn’t non-destructive).
I’d rather use a lossless compressed format with higher resolution, than a lower resolution non-compressed format. It seems like it would be the best of both worlds. The only issue I see with WavPack so far is that it’s not as widely supported as other formats, especially in ffmpeg.
I understand that, as a format used for export to the final product, however from what I understand you want higher resolution during editing to avoid issues like clipping, dithering, etc. I’m sure it’s a moot point for cassette tapes but I do like the idea of keeping an unprocessed master at “archive” quality. Especially if WavPack allows you to retain the quality while saving storage space with the compression.
For example I use a couple of Crucial X9 Pro 1TB SSD drives for my other hobby which is underwater photography/videography (one for live use the other for backup). They make even larger capacity sizes too - but usefully that are physically very small 6.5x5x0.9cm.
Sadly Crucial have withdrawn from the retail market (remaining stock has gone up a LOT in price) - but other SSD drives are available at a lower price point.
My backed-up WAV files from my LP/tape digitization are all backed up on 3 x 1TB hard-disk Toshiba Canvio (two here in the UK and one in Switzerland) - I don’t need high speed for my backups.
Clipping generally happens during recording and it’s the ADC (analog-to-digital converter) that clips. So, recording in 32-bits floating point doesn’t help with that.
Leave plenty of headroom while recording to prevent clipping. Unlike analog tape where you need to overcome tape noise, you don’t need a “hot” signal. On the other hand, digital is less forgiving if you go “into the red”. You don’t have to leave do this, but pros typically record at -12 to -18dB (at 24-bits).
By default, Audacity converts everything to 32-bit floating point anyway. That helps, if for example, you boost the bass and the peaks go over 0dB. (Mixing can also push you over 0dB.) But if you lower the volume before exporting you’ll be OK and you shouldn’t keep a “permanent copy” that goes over 0dB so there’s no reason to use floating point as your everyday or archive format. …The main advantage of floating point is that it virtually has no upper (or lower) limits. You can go to something like +700dBFS without clipping, or down to about -700dB.
Pros don’t keep floating point masters or archives (typically they use 24/96). But their software processes in floating point (as does Audacity).
Dither is added noise that’s supposed to sound better than quantization noise. But cassettes (or most analog) has more noise than the dither so it can be considered “self-dithered”. A tiny amount of inaudible dither isn’t going to matter, although it’s probably theoretically better NOT to add more noise.
As a practical matter, you can’t hear quantization noise or dither at 16-bits or better (except under “artificial conditions” ) so it’s not big deal. You CAN hear quantization noise at 8-bits. Quantization noise sounds a lot like analog noise and it’s most noticeable with quiet parts, but unlike analog noise, it goes-away with silence. A lot of people don’t know what “low resolution” audio sounds like, but you get quantization noise. And at low sample rates you lose the highs.
The REAL sound quality issues are normally on the analog side but for your archive you should avoid MP3 or other lossy compression.
