For reasons of space and utility I’ve decided to digitize my 500-odd record albums and dispose of them. I’m going to be archiving them in uncompressed form (big drives aren’t all that expensive). What do people think makes sense as a sampling rate?
Storage space may be cheap, but I don’t want to be utterly profligate about it. Is 48 kHz at 24 bits good enough? Should I be considering 96 kHz? 32 bits? Or is it overkill? Is 16 bits plenty?
Understand that my sound system is merely decent but never’s likely to be great. I’m going to be transcribing the albums with a Shure M91ED cartridge. I doubt I will ever spend even $500 on a pair of speakers. So, at what point is it likely that I’m simply not going to be able to hear the difference?
I’m asking this as a practical matter. I understand the theory and practice of sampling quite well; part of my living comes from consulting wherein I advise people on such matters. But that’s all from a scientific/technical point of view. When it comes to personal audio, I really don’t know what’s adequate.
~ pax Ctein
[ please excuse any word salad. MacSpeech in training! ]
Personally I record and edit with Audacity set to 44.1kHz 32-bit floating - and when I export completed projects I export to 44.1kHz 16-bit PCM stereo WAV fles (the CD Red Book Standard). This produduces good CDs for me and good AAC files when I rip the CDs into iTunes.
Using 32 bit for recording and editing provides extra headroom for multiple application of effects- it makes the Audacity projects twice as large as using 16-bit. But that is not a space issue for me as I delete the Audacity projects as soon as I have created and backed up my WAV files.
If you are never going to spend more than $500 on speakers (or using studio quality headphones) then I don’t think you will ever hear the difference between using 44.1kHz and 48kHz. 24 bits may give you a problem burning CDs.
Thanks for the info. I, too, doubt that I’ll hear the difference between 44.1 and 48kHz, but it’s also less than 10% difference in storage space, so it’s cheap insurance on a little extra headroom. 32 bit vs 24 bit vs 16 bit interest me more. I’ve read some persuasive-sounding arguments that I really don’t need more than 16 bit for straight transcription of my vinyl… but I’d trust the hands-on experience resident here more.
In the same vein, while I know that technically there are compelling arguments for 96kHz, I’d be really surprised if my system would make them perceivable. But, again, I’m here to learn, not to assume.
I should probably mention that I will NOT be transcribing the vinyl to CD. In fact, my next project after converting the vinyl is to copy my 500-odd CD’s to hard drive and dispose of the CDs, too!
(I’m doing that second because it’s so much faster and easier than converting the vinyl that I’m afraid if I do the CDs first, I may not ever get around to doing the vinyl. )
pax / Ctein
Ctein - I do hope that in addition to copying your transcriptions to a hard drive that you are also making a second backup copy on a separate hard drive (Jack Schofield who writes for The Guardian in the UK has as his Second Law of Computing “data doesn’t really exist unless you have at least two copies of it” - and consider Taylor’s rejoinder “two copies of the same data are always slightly different” - food for thought … )
I keep one copy of my WAVs on a USB RAID disk configured as two 500gb shadowed volumes - and a second copy on a separate pair of 250gb USB disks - and make CD-Rs for each album as a further backup. You will do a lot of work getting those WAVs - it would a shame to lose them ever …
I pretty well agree wit waxcylinder on all points:
32 bit is great for processing as quantizing errors are virtually non-existent.
I find 16 bit quite adequate for the final export, and indistinguishable from 24 or 32 bit.
I usually use 44.1 kHz since the final destination is usually CD, but I think there is a marginal improvement in quality at 48kHz. When listening with my best speakers, 48 kHz seems to have a little more clarity and space - very subtle, and could be my imagination, but unless I do a double blind test and convince myself otherwise, I would say that 48 kHz is better than 44.1, and worth the extra file size. 96 kHz is overkill in my opinion.
Backup - YES !!!
Hard drives are very reliable, but they do go wrong, and can go wrong suddenly and dramatically. Worst case scenario - you turn on your computer and the hard drive is dead (not ill, or poorly or in a coma, but dead and gone forever - it does happen).
Totally agree about backups. I’ve got 2 TB in various external archiving and backup drives. Also regularly burn archival DVDs that get moved offsite. My typical photographic file is the same size as a full audio album.
32 bit is actually pointless unless you plan on doing LOTS of manipulation; it’s just that Audacity 1.2.x crashed and burned every time I try to set it to 24 bits. I’ve moved on to 1.3.5 to see if it’s more stable (it also supports saving directly as FLAC, which saves me a step).
Anyway, consensus seems to be 48kHz and 24-bits.
pax / Ctein
Across the forum there have been various reports of instability with using 24 bit, particularly on some Mac computers. The choice of using 32 bit can have noticeable benefits over 16 bit even with fairly modest amounts of processing (depending on what you are doing), so if your computer can handle 32 bit OK, then it is a better choice than 16 bit for your Audacity project.
I doubt that you will notice any difference in your final export between 16 bit and 24 bit - but test it out for yourself - export a project at both 16 bit and again at 24 bit, then get someone else to rename the files so that you do not know which is which, and do a blind listening test.
Another tip - don’t normalise to 0dB, not even for the final export - leave a little headroom. There are both practical and theoretical reasons why normalizing to 0 dB can cause distortion and offers no benefit over leaving say 1dB of headroom.
Well, FLAC is my ultimate format, and it doesn’t know from floating point, so I guess it doesn’t matter if I capture 32 vs 24. Even with v1.3.5, 24-bit is a little unstable (and I seem to have a tmp-directory-not-found problem to work out). So I’ll try capturing in 32 bit and seeing if the 24-bit flac save works.
In theory, absent much data massaging, one can’t hear the difference between 16 and 24 bit, even with a really good system. In practice, 24 bits doesn’t hurt and gives you room to mess around. I’m generally recording with the absolute peak reading down 4-6 dB (just in case I missed a few transients when setting the volume). This might cause me qualms at 16 bits; at 24 it’s not an issue.
Meanwhile, back to figuring out the directory weirdness. If anyone cares, I’m running the beta on a late 2007 MacBook Pro (Santa Rosa chipset), 2.2 GHz Core Duo, 4 GB RAM, Mac OS 10.4.11, saving to external FireWire 400 drive.
pax / Ctein
There is a considerable debate raging about trying to capture directly to a FireWire drive. I would personally be capturing to my internal System Drive and then Exporting live captures to the FireWire drive for safekeeping. If you involve external drives in your Saved Project (not exported sound file), you are setting yourself up for major headaches down the road.
The AUP is not a sound file. Audacity Projects are clouds of files and don’t move easily.
Macs have troubles with 24-bit. No idea why. 32-floating for production and 16 bit for burning to a music CD if the CD authoring program can’t handle 32-floating.
Thanks for the heads-up on that. Is the same problem known to exist for eSATA drives? Also, might this have something to do with the tmp directory problem I’m seeing?
At the present time, I’m not dealing with AUP files-- just doing raw captures of the vinyl and immediate conversion to FLAC. But down the line I expect I’ll be pulling FLACs back in, splitting them into tracks, converting them to MP3’s etc. IOW, using the AUP functions. I’ll keep this in mind for when that time comes.
pax / Ctein
Follow-up to my previous post:
Moving the temp directory to a partition on my internal drive AND/OR renaming it “audacity_temp” solved the tmp-directory-not-found problem with v1.3.5.
I believe the problem was with the directory name rather than the location-- Audacity, by default, names the directory “.audacity_temp”; this is poor practice on a Mac, as the system ‘reserves’ “.***” file names for system files. Starting a file name with a period usually doesn’t cause problems… but it can. I think this is one of those cases.
Dunno if the code monkeys for v1.3.5/Mac read this forum, but if they do, here’s your heads-up to use a different default name.
If not, I’ll repost this in the ‘unstable’ forum.
pax / Ctein
If you can record at higher sampling rates and your music will not be exported to cd then i suggest you do. The higher sampling rate the better your signal to noise ratio will become. And as vinyl often has a degree of noise (i don’t mean hum) regardless of how good your setup is then the higher sampling rates offer greater overall sound quality
I don’t follow. In my experience with optical sampling (similar though not identical problems) sampling rate only affects S/N if the sampling rate is low enough that the noise is not fully resolved (this is almost always the case if someone is scanning film-- it requires upwards of 8K ppi to well-resolve film grain). It seems to me that with sound, at 48KHz and up, you’re fully resolving the noise the audio system is passing thru. Higher sampling rates may give you better fidelity (if-and-only-if your audio system’s good enough) because you avoid transient phase and aliasing errors, but most of us are not fortunate enough to own phase-accurate audio. But that’s not an S/N issue.
So, please elaborate. What’s the S/N issue I’m missing?
pax / Ctein
I don’t think your missing anything. Increasing the bit depth will increase the theoretical maximum SNR. Increasing the sample rate will increase the bandwidth (higher frequency response).