Thank you @Ladyfingers & @LDoBe for the references to EAC and foobar2000, hadn’t heard of those before. I rip & burn from linux, but I have friends who have asked for an MS-Windows tool, so this gives me something to recommend to them.
Starting late 1999 or early 2000*, I ripped all my (700+) CDs and (100+) audio cassettes to 44.1 kHz, stereo FLAC files. At the time everyone told me I was crazy and that MP3 was the only format I should be ripping to… buwahahahahaha! I had a big crate of 9GB SCSI disks I got for free when Intermedia bought Digex, so I bolted them together into RAID 1+0 arrays (which ended up with absurd numbers of fans, because the drives were prone to overheating) to get the space.
I decided I needed to do FLACs so I could losslessly generate audio CDs from storage, but it also makes it easy to transcode to whatever any other specific device wants. And I have a SlimServer and a DNLA service running on the basement server that houses the media, so it’s accessible anywhere I can get a network connection.
Today the stuff all fits on a stack of SSDs (still RAID 1+0, still salvaged) that I can hold on the palm of my hand.
*Edit: date is wrong. I got the drives in 2000, in the Digex sale aftermath, but didn’t start ripping CDs until 2002 or thereabouts, and built the cassette ripper contraption a couple years later when I finished all the CDs.
I did this as well and with a recent move I lost a CD (yea I’m old and still buy and listen to CDs) but because I had a mp3 I was able to listen to it. Now if I could get my bluetooth receiver working I could of just played from my phone.
I’m pretty sceptical about any claims about being able to tell the difference between FLAC and mp3 (as long as we’re talking about 320/v0, but probably even 192). The only FLACs I have are from when I can’t find a decent mp3 of the release, or occasionally if someone just gives me some stuff. I’ve certainly never noticed any difference, as far as I’m concerned there is zero difference in quality. I suppose it’s possible there are some golden ears out there to whom it would make a difference, but I’m pretty sceptical about that as well.
Well, MP3 audio quality also varies quite a bit depending on which codec you use.
For instance, Fraunhofer IIS’s reference MP3 encoder is atrocious. Sounds like crap. Splashy, swirly cymbals etc.
Then there’s like, middle-of-the-road quality MP3 codecs. Like Thompson’s encoder. It has the neat trick of also being able to do SBR, which means you can chop your file size in half without much loss in perceptual quality. Catch is that you have to have the Thompson decoder otherwise it sounds like a telephone.
And finally, the “industry standard” LAME. It’s the best sounding MP3 encoder by far, and is used in practically everything, since it’s both open-source, and is well-optimized for encoding speed. It’s the most feature-complete MP3 encoder as well.
So I’d say, if we’re talking modern LAME, then I’d say at 320kbps/v0, it’d be transparent under ABX tests. Or very close. But in the past, older codecs sounded much worse.
yeah, that’s all true. if you’re sourcing your mp3s from random file sharing places, and have built up a library over a long period of time, then you’ll probably have some poor quality rips in there (I’ve still got a bunch of things in 128, more in 160/192, mostly they’re fine to be honest - though there was a time when encoder glitches were pretty common, whenever I come across one of them I’ll probably try and find a better version, but the majority of my stuff is in v0 or 320). lame has been stable for a long time now though, so if you’re getting your stuff today from good sources (I mostly buy stuff from bandcamp and bleep), then there’s no need for FLAC imho.
You don’t play the FLAC on a portable device. Well, you can if you are Neil Young, but most people don’t see the point in that.
You encode your MP3s from FLAC, then if you need a lower quality MP3 or an AAC/Vorbis file you also encode that from FLAC instead of transcoding it from the MP3.
This is why I recommend FLAC. Maybe things would have been different if Xiph had managed to get bitrate peeling working on Vorbis, but they didn’t. Not that it would have worked for other formats, which would have still meant transcoding.
It was an established scene standard by 2000, it was good enough for most people (a few killer samples aside), and other encoders and decoders still had flaws to be fixed (Is the Vorbis decoder still as power hungry as it was?)
LAME is still improving the MP3 encoder. It will be interesting to see what they can do when the final parts of the MP3 standard come out of patent (in 2019, if I remember correctly).
It leaves them unplayable. It’s why I encode everything to FLAC before putting the CDs into storage.
I run it on WINE in Linux, because no other Linux audio player did what I wanted.
MP3 is totally not state of the art. I recently participated in a double blind test of MP3, AAC and AAC-HE at multiple bit rates. MP3 needs somewhere between 1.5x and 2x the bit rate before a significant number of people found they sounded like the AACs. The AAC and AAC-HE numbers were much closer, but AAC-HE had a small edge.
MP3 is not state of the art, but it is definitely “good enough” if you don’t care about using half again to twice the storage for the same quality. Given the price of storage, that is an entirely valid trade off. For streaming bandwidth avilablitly and cost might make it a less justifiable trade off.
Yeah, I already said that I was referring to the client-side when I said it’s not in need of improvement. On the server side you need to weigh up increased computational costs against storage and bandwidth requirements though, so it’s not so obvious that 50-100% storage improvements would actually be worthwhile (I know that AAC decoders used to be about half as performant as mp3, but they were very unoptimized then, so maybe the picture has changed today). If you can achieve it without overly increasing the computational load then great, but generally there has to be a trade off somewhere when it comes to performance, and storage and bandwidth are definitely less scarce resources than compute resources.
I know that certain batches of commercially produced CDs have had issues - there was one European plant in … '89, IIRC, that had issues with its discs a few years later. (My copy of Monty Python’s “The Final Rip Off” came from them, and the metal tarnished and essentially ceased to be reflective, which was a common problem from that plant.)
Also, in 1992, the run of TMBG’s “Apollo 18” had a number of discs that had the metal layer just kinda evaporate. (Kinda like the pattern you get when you microwave a CD, but just from sitting around.) Those discs are mostly playable still, but it’s kind of shocking to see.
It’s certainly not common, but various kinds of systemic disc degradation are definitely a thing.
I used to be a Laserdisc enthusiast, so I’m well aware of this phenomenon. I just wasn’t aware of it happening in any large scale on commercially produced CDs.
I do! And I am pretty much like everybody else. (cue laffs). I experimented a bit with MP3 and LAME in the late 90s, and instead settled upon ogg, aiff, or aifc for my encodes at the time. Transcoding between between uncompressed AIFF and WAV is most just a header and endian swap, too quick and easy to worry about.
I started using FLAC in the early 2000s for my home setup, and when I bought a portable player got an iRiver H120 which I upgraded with Rockbox. The OS can handle ogg, mp3, aac, flac, wv, wav, aiff and other formats. For playback I typically use large-driver, large-diaphragm, closed-frame over the ear headphones like Sony MDR7506. Unless I am at a building site or on a noisy subway at the time, it is worthwhile. When in poor listening environments, my experience has been that better fidelity of encoding any playback actually improves the experience, rather than ceasing to matter. I would no sooner give up my FLACs than rely upon using cassette tapes unless I truly needed to.
When I had a project or a friend who needed something to be uploaded in MP3, I have done so. That’s the only time I bother with lossy encoding, because storage is cheap. If I had to choose, I have found ogg and AAC more efficient with regards to storage/fidelity.
What might be adequate quality for passive playback has often IMO been dreadful when it come to remix and reuse of lossy encodes. Especially for inharmonic material.
What makes you think that? There are sample libraries that use FLAC, and Native Instruments have their own lossless format for Kontakt, NCW. Storage is pretty cheap though, so for most people keeping their own content in WAV is usually fine, but I can imagine situations where it might be worthwhile (if someone wanted to keep everything on the one laptop and not have worry about external drives for example).
Is it? I’ll do some serious digging into this later, but a rushed and in-exhaustive google search later I’ve got this:
Compression
The conversion of analogue audio into digitised form determines the available digital audio
quality forever.
The use of digital compression at the time of conversion must be understood in terms of the
compromises between saving expense in data storage, the sound quality that is made available and
the possibly disastrous longer-term implications of some decoders not remaining available over
time.
It is clear that the IT data storage cost presents a variable with descending tendency whilst the
audio quality represents a constant value. In light of this simple consideration the following very
important declaration can be made:
The audio archive should, whenever possible, avoid the use of audio compression and it
should preferably be organised in PCM audio format
There is no possible reason why an audio archive should be at the mercy of the continuing
availability of a decompression codec with which to access its contents.
I don’t mean its not good enough for its purposes, I only mean that using the format for archival purposes doesn’t make sense if you have to encode and then decode to retrieve. There’s an inherent risk in adding points of failure in order to save disk space.
This refers to data transfer for reproduction purposes and short term retention, all fine purposes, but not long time archiving which is the thing I’d find surprising.
I’m not saying FLAC isn’t fit for purpose, I’m saying that in the context of archiving original recordings it doesn’t make sense to store a compressed version as it basically destroys the original and adds a dependency to data retrieval.
If you know what you’re listening for, you can definitely hear tiny differences, especialy in cymbals. You certainly won’t identify the fomat if you’re just blindly listening to something, though.
