whether wav or compressed offer redundancy is irrelevant.  error correction is specific to the communication or storage medium and belongs at a higher layer, in the transmission/storage protocol, not the file format.

hard drives have a built in error correction layer suited to hard drives, par2 is suited to most file-system-on-optical-disc storage, CDs have reed solomon, etc.  the codec doesn't need to know.

error recovery (how well a codec can recover from errors and keep decoding) is different, that belongs in the codec.  wav uncompressed PCM does better with single bit errors than a compressed codec where you might lose more than one sample.  but as pdq mentioned it's really bad with a single missing byte, much worse than flac.

The insecure uncompressed audio CD offers surprising redundancy compared to compressed mediums like MiniDisc. A CD with errors ir very much listenable and the errors may even go unnoticed, while an error in compressed datastream will result in dropout of the entire frame or loud artifacts.

I don't care if if I can't tell the difference between lossy and lossless. Why should I take the chance with lossy? Why waste time trying to see if I can tell the difference when storage is so cheap these days?

As Atkinson points out, lossy has its virtues with portable media where space is still at a premium. But with large hard drives so cheap, flac is my way to go at home with my main listening. I don't have to worry that I may be missing something.

I think that, more and more, the question of going with lossy or lossless is becoming a moot point.

Personally speaking, once I hit the 2 TB+ lossless archive mark -- each set of files backed up twice on separate drives, I found the HD storage issue to be unnecessarily problematic and more than slightly ridiculous. I could either continue to litter my desk with even more firewire drives -- at a couple of hundred bucks a throw; needing two or more new HDs for immediate backup purposes doesn't exactly amount to trivial pocket change for the average person, or I could find a more rational/manageable and (yes) secure solution.

I converted the lot to AAC, then backed up the lossless files onto DVD-R (twice). Result: a drastic simplification of immediate access, less time now spent checking and maintaining HDs, less money spent on said HDs, etc. I am satisfied that VBR AAC is perfectly acceptable for serious listening. I don't spend my time time "worrying" that I may be "missing something" while listening to AAC compressed music. Instead I spend my time engrossed in the music. What can I say? I guess some foolhardy people like to live dangerously, "taking the chance" on lossy and all; catastrophic consequences be damned. Call me Ishmael.

Wait, what? Aren't DVDs more expensive per GB than hard disks? And more likely to decay? And take up more physical space?

Wait, what? Aren't DVDs more expensive per GB than hard disks? And more likely to decay? And take up more physical space?

Don't forget to account the price of Jewel cases or whatever you are using to store the DVDs.

Quote from: j7n on 2008-03-14 16:01:04

Don't forget to account the price of Jewel cases or whatever you are using to store the DVDs.

For offline, long term storage of DVD-Rs that are unlikely to be needed unless it is an emergency, I suggest using the 100-disc spindle the discs came in for storage, because a) it was good enough for transport and sitting on the shelf and b) it's free. 

-brendan

Wow.

...and I thought we were off-topic with the talk about redundancy. 

I think stereophile is out of touch with the times and even some of its fan base. How do they really know the percentage of readers that are already using mp3-aac in some situation If they keep this aggressive stance they will only drive more and more readers away.

The flaws of mp3 @ V4 or similar - and I know that Atkinson & co would burn me at the stake for saying it : the difference if any would be small and for a spilt second.

Also this thing about judging quality from spectograms if childish. Its like looking at two cars and concluding the winner will have better specs on paper, but not paying much attention to the test drive.

If I look at the specs of two identical model cars - exactly the same in every respect - model, model year, tires, etc. except that one has a higher horsepower engine, I'd be pretty certain that the higher powered engine will get from 0 - 60 faster.

I have been admonished for injecting flippancy into a discussion about the hard-to-take-very-seriously subject of that esteemed bastion of rigorous scientific investigation, Stereophile magazine, so I'll bow out of this thread before I start giggling uncontrollably over their 3-figure AC cable "recommendations."

I've been thinking about a response for a while but haven't had the time to post until now. Hopefully JA is still watching this thread.

First, some HA guys mentioned that the lossy spectrum looks like dithering was not used during decode (and that the decoding bit depth was 16 bits rather than 24 bits). If true, that may be a somewhat significant flaw, and some of the spectral artifacts may be attributed to that rather than to any fundamental flaw of lossy encoding. Can JA confirm if this is the case?

I was going to comment about the flawed use of spectrum charts to analyze encoders, but it looks like people have already covered that (and JA has responded to it). I agree with the naysayers - lossy encoders are supposed to have bad-looking spectral charts - and just because a particular distortion is most prominent on the chart doesn't mean that's the distortion that will be particularly audible. Historically, at any given bitrate, the encoders that have the best-looking spectral charts do not sound the best - in fact they usually sound worse. Tuning an encoder for a better-looking spectral chart usually results in a worse-sounding encoder. In other words, frequency spectrum comparisons have no predictive power in evaluating psychoacoustic encoders.

All of that was mostly discussed earlier, and I don't have a whole lot to add to that, except to comment that JA doesn't really have a choice to use them; that's the only evidence he has, because he's already decided to ignore better evidence. The whole point of psychoacoustic encoding is perceptual transparency. The only way to prove that is empirically, through blind testing. But if you don't buy into the blind testing methodology, then you can't trust that the encoder is perceptually transparent. Most of the Stereophile audience, for whatever reason, does not accept DBTs as a valid system evaluation. JA's caught a lot of flak over using frequency spectra to evaluate encoder quality, and how logically flawed that is for accurate quality evaluations. And that flak is well deserved. But really, he's not about to validate the DBT methodology that goes into every lossy encoder in existence, so he doesn't have a choice except to use *****ty charts to prove his point.

But by dismissing all lossy encoders outright, JA is doing a major disservice to audiophiles. As many people have commented, a lot of us would love to use lossless but are still constrained to use lossy encoding for various reasons, mostly relating to a lack of disk space. At the very least, bandwidth still costs money, and people are still looking for the best lossy encoding techniques for online distribution. That we are looking for the best sound quality given those constraints - rather than the absolute best sound quality - means that we'll continue to look elsewhere for our information. And once you start asking those questions about encoding, you realize that it's a lot more complicated than "never use 128kbps" or "use LAME" - the answers require dedicated time and effort to deduce. And lots of ABX testing. (Before ABX testing was enforced, people did tune encoders by ear, and that's largely how the stigma against MP3 and 128k encoding happened. The tunings were terrible.)

To put the point far more bluntly, elitism about lossy encoding is one of the reasons why Stereophile is becoming less relevant (and HydrogenAudio is becoming more relevant) in the hi-fi world.

That said, he's generally right about lossless encoders and how vastly preferrable they are for even mainstream (let alone audiophile) use.

Another point that bears reemphasizing is that most people don't realize the very serious and audible quality loss that occurs with transcoding. A lot of people still think it's ok to transcode from 320k down to a lower bitrate, and even those of us who believe 128k is transparent (which, btw, includes me) will say that's shooting yourself in the foot. Again, all this detail gets glossed over when you ignore lossy encoding entirely, and that lack of information hurts people in the end.

I actually am one of the unlucky few who have experienced corrupted FLAC files and had to junk them entirely. But using an uncompressed format like WAV doesn't solve the problem; you'd still have a corrupted section. The real fix is to back up your data. (And consider using a relatively stable source medium, like pressed CDs.)