jmtd → log → HD audio and more archiving
Having obtained a respectable set of headphones last year, I started pondering about high definition music. Could I hear the difference between a track mixed at CD quality, 16 bit samples / 44.1kHz sample rate, and the same track mixed at a higher bit or sample rate? A number of artists have made HD audio releases available in a variety of formats, most of which are now obsolete (see: SACD, DVD-Audio) and/or hard to demux to something more modern. However some artists have made digital HD audio releases. The last two Nine Inch Nails records were released in HD and are more readily available.
These releases also had nice per-track embedded art
"Ghosts Ⅰ-Ⅳ" had a standard-quality digital release but a high quality Blu-Ray release was also available, commercially. All versions were Creative Commons licensed, so people have legimitely demuxed the Blu Ray release and made it available in other formats. One was uploaded to archive.org, although it sadly lacks precise lineage information:
"The Slip" was released in HD audio digitally. However, I noticed that nobody had uploaded the HD release to archive.org, and having developed an interest in archiving, I did:
I tried performing some listening tests on my desktop computer, which has an M-Audio "Audiophile" 24/96 PCI sound card. However, in non-blind tests, I failed to distinguish the SD and HD versions of these songs. A friend suggested that despite the card's quoted rates, the signal-to-noise ratio was such that you wouldn't get high enough fidelity to play 24/96 properly. I can only take their word for it…
My preferred music player can only decode 16/44.1 but I recently learned that my phone, an Apple iPhone, can decode up to 24 bit, 48kHz. I decided to downsample the above two releases in order to try further listening tests on my phone. I uploaded both to archive.org:
So far in non-blind tests I have not noticed a particular difference. Whether it's my equipment, my ears, these particular tracks or whether HD audio being audibly better than CD quality is a myth, I am not sure yet.
I'm planning to do some proper ABX testing of various lossy codecs and bitrates prior to a major CD ripping exercise, so I may try proper ABX testing of HD audio at the same time.
(I'm collecting together past and future log posts about archiving under the tag 'archiving').
Comments
I'd certainly be interested to hear more about how your listening tests go. I did a similar exercise a few years ago and wasn't able to distinguish any difference between CD/Flac/high bitrate MP3 in a blind test. Although I still rip my CD's in flac format because I feel better about it
And thanks for "The Slip" upload to archive.org
Read a lot about the subject lately and have to conclude 24bit does not add anything, sacd are mostly better remastered.
Nice article:
http://www.audiocircle.com/index.php?topic=70893.0
Don't worry if you hear no improvement
It is quite possible that increasing the sampling frequency is actually a bad idea:
http://people.xiph.org/~xiphmont/demo/neil-young.html
For what it's worth, Monty of xiph.org (he of Ogg Vorbis fame) says HD audio is "Very Silly Indeed":
http://people.xiph.org/~xiphmont/demo/neil-young.html
Unless you have top-of-the-line equipment, I wouldn't worry about audio rates above 16/44k1; and even then, the benefits are debatable. There is a lot of theory available, if you're really interested in the background and have a few hours to spare, then I suggest reading the following three links (in decreasing level of accessibility).
Monty's article focuses mainly on human perception:
http://people.xiph.org/~xiphmont/demo/neil-young.html
Dan Lavry's articles emphasize the technical capabilities of the equipment, and he argues that 60kHz is the optimal sampling frequency. It's not because of the bandwidth itself but because of the characteristics of shaping filters. Sadly, he doesn't really explain the filters (the wikipedia article on aliasing isn't exactly a joy to read either):
http://www.lavryengineering.com/pdfs/lavry-white-paper-the_optimal_sample_rate_for_quality_audio.pdf http://lavryengineering.com/pdfs/lavry-sampling-theory.pdf
Jon My professional opinion as an acoustic engineer is that you can probably notice a difference between 44kHz and 48kHz only if you are very young and extremely well trained to listening. Sampling rate determines where aliasing occurs, roughly at half the sampling rate. Human range is 20Hz-20kHz in theory. In practice, at 18 you are only likely to be hearing sounds at frequencies below 15kHz and it gets worse with age. Another matter is bits, which determine dynamic range. Pointless if you listen to most amplified music, it makes sense though if you listen to classical music. However you need a veeeery quiet environment to enjoy the difference. Also, it depends on the dynamic range of the recording and reproducing equipment, not just the sound card. As an example, top quality measurement mics typically have a dynamic range of about 140dB (roughly 24bits). Another matter is if you use your recordings for measurement purpose like in acoustics, where you want as much info as possible both dynamically and in terms of frequency range, no matter whether you are able to hear it or not. I suspect that a possible difference in HD recordings is created with effects in hd recordings, which are not used in SD recordings and are not based on expanded FS or bits. lots of propaganda fuss in this I'm afraid. Suggestion, check the spec of your headphones and see their dynamic range if quoted (I doubt it). You might find the culprit. Best
I am now experimenting with the DR14 T.meter to calculate the dynamic range of my flacs. I compared a couple of versions of the song Thriller and it varies across albums between 5 and 13. The SACD and normal cd version give a DR13 but most best of albums only have DR5 or DR8. SACD shows the most headroom in audacity.
http://www.dr.loudness-war.info/
Thanks everyone for your comments! I had indeed read Monty's post at Xiph, which was interesting but I got the feeling ever so slightly coloured by his preconceptions. Still useful stuff and lots of follow-on links to read. I'm looking at packaging 'squishyball' for Debian.
Pier, I think your point about the type of music is very relevant. I suspect there is no noticeable difference for these particular releases I tested, but some more accoustic, classical, folk, or whatever might be more distinguishable.
Benjamin, I'm not convinced that a better headphone would make the difference in this case. I didn't make clear in the blog post but I tried two or three different headphones, a Sony minisystem and a Philips AV receiver for my listening tests. I think if there was a distinguishable difference, I would have detected it. I might not have been able to convincingly figure out which was "better" or even more preferable, but I'm sure they would not have sounded identical. If they weren't perhaps a better set of 'phones would help to appreciate the difference more.
Final note on whether this "type" of music is suitable for 24/96: the musicians themselves are quoted in 2010 as saying (about their subsequent releases) " We recorded this at 24bit 48k because we're currently subscribing to the belief that it actually sounds better than 96k. Hence, no reason for larger files to be distributed.".