Re: [ARSCLIST] CD versus Download was "All hail the analogue revolution..."

[Don Cox <doncox@xxxxxxxxxxxxxx>]

On 27/09/06, Tom Fine wrote:
> Hi Don:
> 
> Two points:
> 
> 1. I agree with you about lossy formats in general. Storage space is
> cheap. Bandwidth is cheap. Preserve without lossy compression. As to
> what format to make widely available, that's another matter. But
> preservation should never be in a lossy format. It's penny-wise and
> pound-foolish to the extreme.
> 
> 2. That said, something with the limited frequency response and
> dynamic range of, for instance old radio or many 78's can sound great
> in pretty squashed MP3 and MP4 files, and with as low as 128K
> streaming WinMedia. It does depend on how the engineer EQ's the
> material before squashing. There's a whole article about this in the
> latest Electronic Musician. Different encoders do better with
> different types and different squashed formats sound better with
> different material.
> 
> Here's a for instance. I can't get much FM here but I love Riverwalk
> Jazz (www.riverwalk.org). The 128K WinMedia stream is good enough that
> I can stream it to my laptop, put it thru the console to do some EQ
> tweaking to take out the harshness in the upper mid and boost the bass
> a little bit and then record it to CD. The results actually sound
> better than recordings others have done for me off-air because most FM
> stations do horrible processing (over-compression, super-hyped bass,
> etc -- stuff you wouldn't expect on a public radio station but seems
> to be the norm in all too many places). So my point is that even
> high-fidelity modern stereo material can be made available in lossy
> formats and still sound OK. The opposite is the awful typical 32K
> streams coming from radio station websites. Unlistenable! With old
> radio material, as another for instance, I've done OK crunching to 64K
> MP3 mono (which actually has the resolution of 128K stereo). I find I
> do best if I hard-roll the treble before doing it because otherwise
> things like transcription surface noise result in digi-swishies.
> Crunching OTR to that level means a 20gig iPod can literally hold
> hundreds of hours of programs. I've found that with 78's material,
> especially a well-restored CD issue like what Doug Pomeroy does for
> Bluebird/BMG or my friend Art Shifrin has done, it's gotta be at least
> 128K mono to capture everything so it sounds relatively close to the
> source. For something like you can buy from the guy down in Maryland,
> which are just dubs of decent-condition public-release records, then
> the same rules as OTR seem to apply.
> 
> One final point on squashed formats. This has been said many times in
> many places but bears repeating yet again. Each encoder is different.
> Something like LAME is a workaround of the German "official" MP3
> encoding technology. I find it does not do a very good job in some
> cases, does a respectable job in others, takes too long to do the job
> in all cases. The German-licensed encoders like MusicMatch and iTunes
> are fast and do a very good job in most cases. I find MusicMatch is
> easier to tweak, and iTunes software stupidly has issues with CD Text
> on home-brew CD's whereas MusicMatch looks for CD Text first before it
> queries the often-inaccurate or sloppy online databases like Gracenote
> (the better one) or Freedb (which iTunes uses because Apple won't pay
> the Gracenote royalty, as I understand it). When Sony bought
> Soundforge, they ponied up to the Germans and thus their versions of
> the software have had full MP3 capabilities. I find that Soundforge
> makes good sounding files but they are very large with the same
> parameters as compared to iTunes or MusicMatch. Why??? The Windows
> Media Player later versions makes the best WMA files, but I can't see
> too many situations to use that format except if you are working with
> a Windows-based server that streams media. I don't do Real -- I think
> it's an awful format for a variety of reasons. I also don't do
> Quicktime/AAC because I think it's a fringe format and Quicktime
> software is very invasive if a user doesn't install it carefully on a
> Windows system (so is Real, BTW). So I don't want to encourage those
> proprietary formats -- same counts for WMA, but something available to
> 95% of desktops as part of the OS is hardly "proprietary."

All these complications are another reason to avoid lossy formats.

The main reason, though, is that they distort the sound in ways that the
brain cannot interpret. 
 
> One format that does save on file size but does not lose any
> information is Apple Lossless Format. It's been standard on iTunes for
> a couple of versions now and the newer iPods and Nanos can play it. To
> my ears, it's identical to putting a CD in the drive and playing it
> through iTunes software.

It must be, if it is lossless and the samples are 44.1/16.

>  I can't understand why Apple doesn't make
> more use of this format -- for instance putting a security wrapper
> around it and selling iTunes downloads for maybe a 10% premium in
> "full fidelity" -- or competing with Wal-Mart's rising download
> business by selling squashed AAC iTunes for 75 cents each and
> full-fidelity ALF for the standard $1. As I've said before, I'd gladly
> pay $1 for a full-fidelity legal download but I'm not wasting money on
> blatantly inferior-sounding formats. Anyway, ALF files tend to be
> about 75-80% the size of the original WAV. 

I wonder if it is worth the trouble for such a small saving. 

> The big problem with ALF is
> that so far it only works with iTunes and a few obscure
> freeware/shareware programs. I am hoping it will grow in popularity.
> Make no mistake -- I do not think it or any other data-compression
> (lossy or not) is a good idea for archiving; again, storage is cheap
> and bandwidth is cheap compared to re-doing something because data is
> lost or corrupted. But as a consumer format, I see ALF as a win-win.
> It's certainly what I'd use in my ultimate-networked-jukebox server
> that exists in my mind but has yet to be built here.

Most image formats (GIF, TIFF, IFF-ILBM, etc) use lossless compression.
So long as the decoding algorithms are well known, it is not a problem.

Regards
-- 
Don Cox
doncox@xxxxxxxxxxxxxx