Weekly Edition
Return to the Press page
| |||||||||||||
Interview with Richard Hulse of Radio New Zealand, on the decision to offer Ogg Vorbis (Groklaw)
Groklaw has an interview
with Richard Hulse about Radio New Zealand's decision to offer Ogg
Vorbis as one of
their audio formats. "One of the great things about Public Radio is
that you can take a longer view of things. You can put something in place
for the future, looking to changes that you see on the horizon. So at this
stage I am not concerned about the number of downloads - new services are
not always popular when you first introduce them. For example, we had only
300 subscribers to our podcast feeds in the second month of offering the
service. Two years on there are over 20,000 people who download about
250,000 items every month. If you played all that audio end-to-end 24/7 it
would run for nearly 18 months. [ ... ] I do hope that people use the Ogg
files because apart from the freedom
aspect, the quality is better than MP3 for the same data rate. I'll be
happy if we get up to a couple of percent by the end of the year."
(Log in to post comments)
Interview with Richard Hulse of Radio New Zealand, on the decision to offer Ogg Vorbis (Groklaw) Posted Aug 25, 2008 16:44 UTC (Mon) by pr1268 (subscriber, #24648) [Link] I do hope that people use the Ogg files because apart from the freedom aspect, the quality is better than MP3 for the same data rate. I can vouch for this claim--I recently converted my music collection from 320 kbit MP3 to Ogg vorbis (quality=8, ABR=256 kbit or so), and the whole collection of Ogg files use 80% of disk space that the MP3s used to take. I further analyzed the files using an audio file editor and noted no loss of fidelity on the Ogg files--frequency response all the way to 22 KHz (which is what audio CD can produce).
Interview with Richard Hulse of Radio New Zealand, on the decision to offer Ogg Vorbis (Groklaw) Posted Aug 25, 2008 19:46 UTC (Mon) by vmole (guest, #111) [Link] Uhh, it's nice that it saved the disk space, but converting from one lossy format to another is a usually a killer on audio quality. If you're up for an experiment, try re-ripping one of your CDs to ogg (same parameters), and then compare the MP3->OGG file to the new CD->WAV->OGG version. Compare not with a frequency analyzer, but by actually listening (which is the whole point of loss formats). If you can't tell the difference: great! You win! If you can hear a difference, you lose: you'll now feel obliged to re-rip your entire collection. But, this time, rip to flac. That way you can generate the lossy-formats for your portable, and never have to re-rip again.
Ogg vs. MP3 Posted Aug 25, 2008 22:11 UTC (Mon) by pr1268 (subscriber, #24648) [Link] I actually did all that you mentioned... While I agree that lossy audio formats are just that (lossy), I did hesitate actually making the change for fear of ruining some of the subjective listening quality of my songs. But, after long nights listening to every genre in my collection (spanning Beethoven to Guns N' Roses with some Snoop Dogg and Norah Jones mixed in), I decided that Ogg Vorbis at quality=8 is equally indistinguishable from uncompressed WAV (ripped from a storebought audio CD) as MP3 at 320 Kbit converted from the same WAV. Although, I will admit that all lossy formats, even audio CDs (yes, they are a lossy format also) have a somewhat clinical, mechanized, overprocessed sound to them, but that's a whole other debate (and one that's been raging since the early 1980s).
Ogg vs. MP3 Posted Aug 25, 2008 22:54 UTC (Mon) by vmole (guest, #111) [Link] Well, as I said, if the MP3->OGG conversion make you happy, you win. And yeah, my LP of _Appetite for Destruction_ on a Rega 300 blows away the CD...but I listen to the CD more often. I've gotten too damn lazy...
Ogg vs. MP3 Posted Aug 25, 2008 23:13 UTC (Mon) by robert_s (subscriber, #42402) [Link]
"(yes, they are a lossy format also)"
Nonsense. To argue this, you would have to argue that all sampled-data formats are lossy. As there are no non-lossy ways of representing real-world data, this argument would mean the word 'lossy' were totally redundant and meaningless.
Lossless audio formats Posted Aug 25, 2008 23:53 UTC (Mon) by pr1268 (subscriber, #24648) [Link] Um, all sampled-data formats are lossy. Continuous, non-sampled, analog recordings are closer to "lossless" than digital audio (although they introduce other artifacts and anomalies). In order for sampled audio to be lossless, it would need to have a sampling period of zero and infinite resolution, both mathematical impossibilities. What FLAC does, IIRC, is to make a lossless clone of an uncompressed WAV file, and using lossless compression techniques (similar to LZ77 or Burroughs-Wheeler transforms) to shrink the file size. Other "lossless" audio compression algorithms do similar, but all these do is avoid introducing any more artifacts/anomalies than what the source WAV file originally had. Not that I'm arguing against digital audio formats, even lossy ones, because in this day and age it's too darned inconvenient to whip out the reel-to-reel and play back at 30ips. ;-)
Lossless audio formats Posted Aug 26, 2008 0:47 UTC (Tue) by Mithrandir (subscriber, #3031) [Link] Any recording is lossy, sure. It's impossible to record the exact sound-wave field of an original audio source. But you start to stretch the bounds of sanity when you go down this route. We need to draw some kind of line in the sand if we're to use the term "lossless". Sure, any recording is not going to be exact. But there are so many other influences in the recording pipeline that may be more important, like the placement of the microphones, the frequency response of the recording electronics, the aliasing caused by digital sampling, the frequency response and placement of the speakers or headphones that you use to listen to the resulting recording etc. etc. etc. The warmer, more enjoyable tone that people often refer to when listening to vinyl is actually caused by distortion of the original recording, by a non-flat frequency response. The clicks and pops caused by microscopic dust particles can be either enjoyable and soothing or absolutely, intolerably annoying depending on your point of view. Whether the recording is closer to the original is such a subjective argument as to render it almost meaningless for the purposes of debate. I mean on what measure are we comparing them? Distortion of the frequencies normally audible to the human ear? CDs win. Distortion of the frequencies not theoretically recordable with a CD's sampling rate? Records win. Continuity of fidelity over time and with consecutive plays? CDs win. With a tactile, satisfying user experience? Records win. Whatever. I love this anecdote that my dad told me once; he was a solid state physicist and electronics buff, and he couldn't abide people who talked stuff up, especially needlessly expensive stuff. He went over to an audiophile's house one time, and he was showing my dad his amazing audio set, with its state of the art amazingness etc., enthusing about how exact the sound was, how perfect the rendition of the original, how the thousands of dollars spent was totally worthwhile. My dad politely pointed out that the road noise coming in from outside caused a bigger dip in signal to noise ratio than did the cheaper electronics in his own set at home. For all intents and purposes FLAC is lossless. The original signal is retrievable, and is exactly identical to the uncompressed one. Its strength lies in being specifically tailored to compressing audio signals, and as such it typically compresses up to 50%, compared with general purpose compression like ZIP with typical compression of 20%. If you rip straight from CD, you can record a .cue file which allows an exact reproduction of the original CD. Try that with Vinyl. I guess, if I actually had a point to this comment, it would be something like: sure, no recording is precisely lossless. But CD is so close to it that 99.99% of people wouldn't be able to tell the difference, especially when listened to on typical audio electronics. Have a look at Nyquist's Theorem. If you want to go buy all your music on DVD, that's fine, the recording industry will love you. I hope you can hear above 22kHz though. The popularity of MP3 and OGG attest to the fact that they're good enough. The unwashed masses just. don't. care. They've voted with their feet. It's fun to argue about it though, isn't it? ;)
Lossless audio formats Posted Aug 26, 2008 1:29 UTC (Tue) by Sutoka (guest, #43890) [Link] Try that with Vinyl.The Xiph.Org guys recently announced that was something they planned on changing. A lot of people believe Vinyl gives a better experience (like you said) than CD/FLAC/any digital, so Xiph.Org is planning to announce a new project that'll accurately reproduce that experience but with much better convenience. Apparently the initial code base is going to be based on FLAC, but as it's meant for analog audio, they're planning on naming it the Analog Free Lossless Audio Codec, or AFLAC for short. Apparently they're already planning on using this as their logo. In case you just heard a a 'whooshing' noise, that was a joke :P
Lossless audio formats Posted Aug 26, 2008 1:52 UTC (Tue) by pr1268 (subscriber, #24648) [Link] Any recording is lossy, sure. It's impossible to record the exact sound-wave field of an original audio source. But you start to stretch the bounds of sanity when you go down this route. Agreed. And I hope no one thinks I'm going insane for arguing whether CDs are lossless. I don't mean to criticize or dismiss audio CDs, but rather simply point out that even they are not a true lossless audio format. But they're incredibly close! My dad politely pointed out that the road noise coming in from outside caused a bigger dip in signal to noise ratio than did the cheaper electronics in his own set at home. Hehe, also agreed. But what a downer of an argument! I've a factory CD player in my little pickup truck, but when I get up to 65 MPH or so, the road/wind noise reduces my S/N ratio to about 20-25 dB (on a quiet road). Kinda defeats the purpose of all that available dynamic range. :-/ I guess, if I actually had a point to this comment, it would be something like: sure, no recording is precisely lossless. But CD is so close to it that 99.99% of people wouldn't be able to tell the difference, especially when listened to on typical audio electronics. Have a look at Nyquist's Theorem. If you want to go buy all your music on DVD, that's fine, the recording industry will love you. I hope you can hear above 22kHz though. Yes, I've studied Nyquist's theorem (and read the related parts of Claude Shannon's famous AT&T paper on information entropy). I'm recently enlightened from a computer multimedia studies class in grad school, and thus I'm probably showing off my newfound knowledge more than I'm trying to open a can of worms. :)
Lossless audio formats Posted Aug 26, 2008 6:39 UTC (Tue) by nix (subscriber, #2304) [Link]
People keep on saying FLAC compresses at about 50%, but that's just not
true. It manages 50% for noisy stuff like a lot of popular music or full orchestral works. For me, it manages *70%* consistently for, say, piano sonatas or something like that. Get *that* out of zip. :)
Lossless audio formats Posted Aug 26, 2008 7:17 UTC (Tue) by Los__D (guest, #15263) [Link]
FLAC is a lossless representation of certain digital audio formats.
MP3/Ogg is a lossy representation of certain digital audio formats.
Any recorded audio is a lossy representation of "real" audio (Loss going on through the entire chain; microphone, wires, electronics, sampling, etc.). Hell, any listening is lossy.
In my mind, when talking about anything lossless, it just mean that _this_ particular step does not remove anything, not if it is lossless compared to the original sound of the vocal cords moving.
In that case, there isn't any limit to "how lossy" it is before calling it lossless. Either it is lossless, or it isn't.
Lossless audio formats Posted Aug 26, 2008 14:28 UTC (Tue) by zotz (guest, #26117) [Link]
"In my mind, when talking about anything lossless, it just mean that _this_ particular step does not remove anything, not if it is lossless compared to the original sound of the vocal cords moving."
Bingo!
drew
Lossless audio formats Posted Aug 26, 2008 10:15 UTC (Tue) by epa (subscriber, #39769) [Link]
I think you are mixing two things.
There is the difference between the original sound and what is sampled. No CD can ever exactly reproduce the original sound waves, and no image file can exactly reproduce the light waves. A sampling process turns the real world into digital data and loses some information. But this is not what 'lossy' refers to.
Being lossy or lossless refers to a transformation of data to data. If the transformation is reversible to get back the exact original data then it is lossless. If not, it is lossy (though the difference may be almost indistinguishable to the human ear or eyeball).
You are right that some loss of information is involved when recording a CD - and more loss of information if you did say a 10KHz, 8-bit mono sample - but to use the word 'lossy' for this is making the word so broad as to become effectively meaningless. It's better to keep lossy and lossless to describe algorithms that work on data, where there is a real distinction between the two.
Lossless audio formats Posted Aug 26, 2008 14:30 UTC (Tue) by zotz (guest, #26117) [Link]
"Being lossy or lossless refers to a transformation of data to data. If the transformation is reversible to get back the exact original data then it is lossless. If not, it is lossy (though the difference may be almost indistinguishable to the human ear or eyeball)."
Bingo.
drew
Lossless audio formats Posted Aug 26, 2008 15:36 UTC (Tue) by johnkarp (subscriber, #39285) [Link]
Signal sampling *isn't* lossy for information below the Nyquist frequency: perfect reconstruction is possible. Thats what the Nyquist-Shannon sampling theorem shows.
And since audio has a finite bandwidth and bounded resolution, its therefore quite possible to losslessly represent an audio signal using samples.
Also, I don't see how you can claim non-sampled recordings are closer to lossless than sampled ones. The only effective distinction between the two is that the frequency response of non-sampled signal trails off to zero at higher frequencies, whereas sampled signals have a uniform response that cuts off quickly at a particular frequency. The bandwidth of one isn't necessarily greater than the other, it all depends on the particular technology involved.
Lossless audio formats Posted Aug 26, 2008 16:08 UTC (Tue) by epa (subscriber, #39769) [Link] And since audio has a finite bandwidth and bounded resolution, its therefore quite possible to losslessly represent an audio signal using samples.Does it really have a bounded resolution? Do air molecules move in discrete jumps? I am not a physicist so I don't know the answer, but I kind of assumed that in the real world the resolution is infinite. Of course the human ear has only limited capacity, but that comes back to the same discussion as before: some of the information from the original sound has been lost, but it is not perceptible to ordinary listeners.
Lossless audio formats Posted Aug 26, 2008 16:46 UTC (Tue) by johnkarp (subscriber, #39285) [Link]
Not exactly, its more that particles, even macroscopic ones, don't *have* exact positions. Only probability distributions describing where you'll measure them to be.
Also, I was being somewhat redundant: since the bandwidth is proportional to resolution, if the bandwidth is finite, the resolution must be too.
Lossless audio formats Posted Aug 26, 2008 16:57 UTC (Tue) by dmaxwell (guest, #14010) [Link] I am not a physicist so I don't know the answer, but I kind of assumed that in the real world the resolution is infinite.That is not a good assumption. Signal to noise ratio and slew rate determine the "resolution" of an analog format like tape or vinyl. To put it another way, signals like what is coming off a mike will do things that either don't wind up on the vinyl or imperfectly on the vinyl. What Nyquist and Shannon do for us is tell us the sample rate we need to reproduce signals at given level of fidelity expressed in dB. If the sample rate is high enough to meet or exceed the vinyl then one cannot fairly say that the vinyl has higher resolution. One reason vinyl often sounds better is that CDs are "hot mixed" by egomanic producers who want to have the loudest record in the jukebox. Google up the "loudness war" for more. A vinyl stylus will literally jump out of the groove if hot mixed the way many CDs are. Hot mixing increases average loudness at the expense of dynamic range. It sounds terrible. It also turns out that the capabilities of human hearing can be objectively quantified and yes CDs exceed that for 99% of individuals. The intersection of "golden ear audiophiles" and that remaining 1% is very very low. If we add the set of people "knowledgeable about the physics of sound, engineering of electronics, and understanding of information theory" then the intersection is all but non-existent.
Lossless audio formats Posted Aug 26, 2008 17:48 UTC (Tue) by pr1268 (subscriber, #24648) [Link] One reason vinyl often sounds better is that CDs are "hot mixed" by egomanic producers who want to have the loudest record in the jukebox. Yes, that's true, but back in the 1980s most CDs were pressed with soft-volume masters created for vinyl phono equalization. While the LP would have effects created on the playback end reversed ("undone") by this equalization, the CDs didn't and thus they sounded like crap (IMO). But I agree that the "loudness wars" have had a detrimental effect on the audio quality of music CDs for the past 12 years or so. My impression is that most CDs mastered from 1988 to 1996 (or thereabouts) sounded best, as this time period was after the mastering engineers wised up to CD equalization and before the loudness wars started. If the sample rate is high enough to meet or exceed the vinyl then one cannot fairly say that the vinyl has higher resolution. As the earlier comment said, analog recordings will have their frequency response tail off slowly at high frequencies, but I don't think that vinyl LPs are limited by a Nyquist frequency--I read somewhere a while back that some high-quality ("extra virgin") vinyl could reproduce audio frequencies all the way to 60 kHz and beyond. Of course, this is all moot considering that:
Lossless audio formats Posted Aug 26, 2008 19:10 UTC (Tue) by vmole (guest, #111) [Link] Yes, that's true, but back in the 1980s most CDs were pressed with soft-volume masters created for vinyl phono equalization. While the LP would have effects created on the playback end reversed ("undone") by this equalization, the CDs didn't and thus they sounded like crap (IMO). "Most" seems unlikely, since it's an obvious mistake. Anyway, even the DDD CDs sounded like crap, often even worse than the AAD CDs. I think a lot of it was that the D/A and A/D converters sucked for audio. Remember the modded CD players of the 80s/90s? Remember the cult of the Burr-Brown D/A converters? They really were much better than the typical $0.05 piece in a cheap CD player.
CD audio issues Posted Aug 26, 2008 20:34 UTC (Tue) by pr1268 (subscriber, #24648) [Link] Anyway, even the DDD CDs sounded like crap, often even worse than the AAD CDs. I agree, but I think a lot of the crappy sound also came from engineers' lack of experience using digital recording/mixing/mastering equipment. Plus, I once heard a recording engineer comment that he felt "artistically and functionally limited" when mixing an album purely in the digital domain (which may explain the appearance of DAD discs in the early 90s). There were a few excellent-sounding DDD discs from the early 80s; Donald Fagen's The Nightfly (1982) was/is one of the nicer-sounding all-digital albums from that era. Remember the modded CD players of the 80s/90s? Remember the cult of the Burr-Brown D/A converters? They really were much better than the typical $0.05 piece in a cheap CD player. Can't say that I have--I would love to have heard a disc played on such a player. Understand that I was a whole lot younger back then and not nearly the hacker I am today. :-)
Lossless audio formats Posted Aug 27, 2008 3:07 UTC (Wed) by pr1268 (subscriber, #24648) [Link] "Most" seems unlikely, since it's an obvious mistake. I stand by my original claim about most early/mid-1980s CDs mastered with vinyl phono equalization (based on an article I read in Stereo Review magazine around 1995), but the "mistake" wasn't so obvious back then. Most CDs were rushed to market using the most convenient master handy--which was often the RIAA Phono-Equalized master. But, in retrospect, I think that the rapid commercial success of the audio CD sort of necessitated getting discs to market quickly. Not that I'm rationalizing the recording industry's decision to do so...
Lossless audio formats Posted Aug 27, 2008 22:04 UTC (Wed) by jrigg (subscriber, #30848) [Link]
RIAA phono equalisation is applied on the way into the cutting lathe, not on the master tape (which has its own IEC or NAB equalisation applied during recording and removed during playback). Some CDs were recorded from master tapes that were optimised for vinyl reproduction, but that's nothing to do with the RIAA equalisation that takes place in the cutting lathe preamplifier.
Aargh! Another digital vs. analogue argument! Posted Aug 27, 2008 18:41 UTC (Wed) by jrigg (subscriber, #30848) [Link]
These discussions always seem to turn into a debate about vinyl vs. CD, digital vs. analogue etc.
As a recording engineer, here's my perspective.
All recording methods are lossy.
The move to better sounding lossy formats is good IMO. I look forward to the day when there is enough bandwidth and storage space available to allow the use of lossless compression (or even no compression at all) as standard.
Aargh! Another digital vs. analogue argument! Posted Aug 27, 2008 20:22 UTC (Wed) by pr1268 (subscriber, #24648) [Link] Thank you for your thoughts. CD isn't too bad, but compared with a 48 or 96 kHz 24 bit .wav file they are noticeably degraded. As the starter and active participant of this discussion thread, I do wish to state for the record that I never meant to criticize or otherwise harshly judge the technical aspects and audio quality of audio CDs. In fact, I'm still impressed with how well the engineers at Sony and Philips did considering the late 1970s technology with which the CD was developed. The move to better sounding lossy formats is good IMO. I look forward to the day when there is enough bandwidth and storage space available to allow the use of lossless compression (or even no compression at all) as standard. Lossless compression, sure. No compression? Forget about it. Computer science and networking has always been about how to do more with less, or how to do it all with minimal use of resources. Given a choice of downloading a WAV or a FLAC file, I'll always choose the FLAC.
Aargh! Another digital vs. analogue argument! Posted Aug 27, 2008 21:50 UTC (Wed) by jrigg (subscriber, #30848) [Link]
> Lossless compression, sure. No compression? Forget about it. Computer
> science and networking has always been about how to do more with less, or > how to do it all with minimal use of resources. Given a choice of > downloading a WAV or a FLAC file, I'll always choose the FLAC.
You're right, of course. If the compression is lossless there's no point wasting bandwidth by not using it.
Aargh! Another digital vs. analogue argument! Posted Aug 27, 2008 23:44 UTC (Wed) by dlang (✭ supporter ✭, #313) [Link]
it all depends which is the scarce resource, bandwidth/space or CPU.
if you have a huge amount of bandwidth (at least compared to the signal you are dealing with), and a poor cpu (or a cpu busy with other tasks), then no compression can be a win.
however processor speeds have been outpacing everything else in the industry to the point that this is seldom the case nowdays.
Aargh! Another digital vs. analogue argument! Posted Aug 28, 2008 6:23 UTC (Thu) by dvdeug (subscriber, #10998) [Link]
The weird thing is, Blu-ray supports and many disks use LPCM, which I'm pretty sure is an uncompressed format. I would think given the existing complexity of Blu-ray codecs and despite the space available, cutting the space used on audio in half with no loss in quality would be a win.
Aargh! Another digital vs. analogue argument! Posted Aug 28, 2008 11:31 UTC (Thu) by lysse (guest, #3190) [Link]
Maybe it's a question of error correction? If you can't correctly read a few samples of uncompressed audio, you just guess at what it should be (or flatline it, even) and the listener hears a transient impulse at worst; the data either side is intact. If you can't read a word of a losslessly compressed audio frame, the whole frame (potentially 10-12ms of music - definitely audible) has effectively gone for a burton's. Given the likelihood of physical media being damaged or otherwise prone to misreads, it makes sense to go with the format which maximises your redundancy.
Uncompressed better for error correction Posted Aug 28, 2008 15:04 UTC (Thu) by pr1268 (subscriber, #24648) [Link] Good point. And, dlang made a good point above: conserving scarce resources could prescribe using uncompressed data instead of lossless compression in the interest of conserving memory and CPU. Computer science is actually a study of economics in disguise. Just a general comment, as if I haven't already strayed far enough off-topic about Ogg vs. MP3. But what a fun discussion this has been! :-)
Uncompressed better for error correction Posted Aug 28, 2008 18:03 UTC (Thu) by dvdeug (subscriber, #10998) [Link]
I don't think that's an issue in this case; are there no lossless codecs that can stay below the footprint of AC3 or MP3, which are already supported?
Aargh! Another digital vs. analogue argument! Posted Aug 29, 2008 0:10 UTC (Fri) by dlang (✭ supporter ✭, #313) [Link]
I think that compressing the data and then adding error correction to your file format would allow you to add MUCH more robustness to the file compared to the uncompressed version
|
|||||||||||||
