The Death of High Fidelity

[Brian]

In the age of MP3s, sound quality is worse than ever


David Bendeth, a producer who works with rock bands like Hawthorne Heights and Paramore, knows that the albums he makes are often played through tiny computer speakers by fans who are busy surfing the Internet. So he's not surprised when record labels ask the mastering engineers who work on his CDs to crank up the sound levels so high that even the soft parts sound loud.

The Death of High Fidelity

[Ford Freak]

I would be ****** off if all labels started doing that. I just got a couple CDs in the mail yesterday. Somebody at work asked me why I didn't download them. I said "because it will sound like crap unless I find an exact CD image."

[k0NG0]

Rips of CDs at 256kbps or higher are virtually of CD quality. Especially on non-professional audio equipment, and pretty much there too (you can do spectral analysis and almost not tell PCM apart from an MP3 at 320kbps). Rippers very rarely do anything but data-rate compression, which only cuts certain frequencies out of the mix, it doesn't compress the sound itself.

The dynamic range compression goes on CDs as well these days. Most of them are compressed to hell and back during mastering. One of the reasons I like music from "olden days" is that very reason. It has nothing to do with the fact that it was recorded on analog, it's that it was produced and mastered to a higher standard. If they had digital back in the day, they'd have used it.

Sentences like "analog sound to me is more emotionally affecting" are tainted, because the subject knows it's sound from an analog recording source in the speakers ("sound" is always analog) and adjusts the mental state accordingly, based on those preconceptions. I fail to see how anyone would really tell an analog recording from, say, a DVD-Audio recording of the same thing (properly mastered of course; DVD-A has excellent dynamic range at 24-bit) in a double blind-test, provided the analog recording is superb and they're played through the same set of amps/speakers.

I'm also fairly certain that if you made the DVD-Audio recording sound more analog-y with carefully crafted noise added to the sound, you could get "experts" to say it was warmer and more rich, or whatever they usually say.

There's a lot of snake oil in this business. A lot. Never mind that a 44.1kHz sample rate means you can recreate frequencies up to 22kHz (see: Nyquist) usually above audible range unless you're a Human/Golden Retriever blend.

[XJ.]

Originally Posted by k0NG0

Never mind that a 44.1kHz sample rate means you can recreate frequencies up to 22kHz (see: Nyquist) usually above audible range unless you're a Human/Golden Retriever blend.

I used to be able to hear well above 20khz (sonic welders drove me nuts), sadly, as was proven at a Boston Museum of Science demo, I'm down to ~15khz

I remember doing the derivation for digital sampling back in my sophomore year, though I can't remember why we chose a sample rate of 2.2 samples/hz over the theoretical value of 2 samples/hz.

[ctal]

I agree with much of the estimable kONGO's learned comments, particularly about the snake oil that pervades much of the audiophile business, and the fact that most recordings are highly compressed and poorly engineered. But I'd like to make the following points:

-- A real-world low pass anti-aliasing filter, which is necessary to keep spectral components greater than 22.05kHz out of the sampler (where they would produce aliasing artifacts), will have frequency response ripple and phase anomalies within the passband (i.e., at lower frequencies). I believe that these effects can be audible on high quality recordings played back on high quality systems, at least with filters that are designed with typical cost compromises. The higher effective sample rates that can be provided by SACD and DVD-A can reduce these effects.

-- Spectral comparisons of mp3 vs. cd music I would expect to be of little use with respect to these issues. First, the human ear has extraordinary dynamic range, which I suspect is better than most or all spectrum analyzers. Also, spectral analysis ignores the time domain, to which the ear is very sensitive (I'm referring here to things like the time alignment of the spectral components relative to each other, and also between channels, phase distortions, transient intermodulation distortion, the ear's tendency to "latch on" to the leading edge of a transient waveform, etc.).

-- Most recordings have been highly compressed for the last several decades; this is not a new phenomenon although it very well may be getting worse. In the old days, a major reason for compression was to catch listeners' attention on FM radio, so they wouldn't tune past the station while driving.

-- I have not done any critical listening to mp3's, but I would strongly suspect that a well engineered, essentially uncompressed recording of acoustic instruments in a natural space (e.g., a classical chamber group or symphony orchestra, or a well done jazz recording), played back on a high quality system, would easily demonstrate serious shortcomings of even very well done mp3 encoding.

-- Many years ago, as my stereo system evolved from low-end to mid-fi to what I would call "common sense high-end," I expected that as the system got better the majority of recordings (which I think we all agree are poorly engineered) would sound worse, not better, due to the system's more revealing character. But I was surprised to find that virtually everything sounded better. The reason was that just about every recording manages to get something right (say, the mid-range of some instruments), and my attention would be automatically drawn by the greater realism to what was right about the recording, not what was wrong.

-- Given that most recordings are poorly engineered, and that trend is getting worse, it is still not difficult (especially in the classical area, but also in pop) to find a lot of good sounding material. I for one hope that continues to be the case, despite the poor state of the cd business, and the trend towards mp3's and computer listening.

Regards,
-- Al

[Acid8000]

Originally Posted by ctal

-- Most recordings have been highly compressed for the last several decades; this is not a new phenomenon although it very well may be getting worse. In the old days, a major reason for compression was to catch listeners' attention on FM radio, so they wouldn't tune past the station while driving.

I think the clipping distortion tended to be worse for CDs released between 2000-2004 than those today. However the dynamic compression is still just as bad. It's a pity, but in more recent times I've had a tendency to buy music from earlier periods that are mastered better than those released now (plenty of it is rubbish now anyway, and I'm a younger person saying that). Of course it can get a bit annoying trying to find an older pressing if it recently has been remastered to be as dynamically compressed as anything from today.

[XJ.]

Well, for those whose ears are still 100% and well trained, this sort of thing matters. I think that most of the gen X&y crowd probably don't care (not trained) and some of us a bit older just don't have the ears for it.

I can tell the difference between 128kbps and 192kbps, but I can't tell the difference between 320kbps and a CD (on my computer). I rip my CDs at 128b AAC for our IPods an I'm satisfied with that since I only listen to music when exercising or driving, where the ambient noise would cancel out the benefit of higher bit rates (plus I can fit my CD collection on a 4GB IPod). I'd probably like 256kbps if I had my choice, but I'd wind up with mixed results due to iTunes purchases. I have friends who'd need >100GB to store their collections at 128kbps - and they probably wouldn't be happy with that! Some people really love their music!

[Acid8000]

For data compression I use VBR ranging from usually 160-240 kbps and it sounds transparent on my setup. Of course it may be possible to hear artifacts and such in a studio or other high quality setup, but when I'm at home, on a computer, I don't need any more quality. This article seems more about dynamic compression than lossy audio compression though.

[ctal]

Originally Posted by XJ.

Well, for those whose ears are still 100% and well trained, this sort of thing matters. I think that most of the gen X&y crowd probably don't care (not trained) and some of us a bit older just don't have the ears for it.

Well put, and I agree 100%. It is sad, though, that most of those who don't care will never be exposed to what is possible in high quality music reproduction, and will never have a chance to decide on an informed basis if it matters to them or not.

Regards,

-- Al

[Acid8000]

I only really took interest in higher quality recordings a few years ago. It gets irritating when I try to discuss it with friends my own age and they wonder what the heck I'm talking about and that older recordings are not loud enough.

[ctal]

Originally Posted by Acid8000

I only really took interest in higher quality recordings a few years ago. It gets irritating when I try to discuss it with friends my own age and they wonder what the heck I'm talking about and that older recordings are not loud enough.

It's exactly the same talking with many older folks as well! Besides "not loud enough," invariably a listener used to a typical mass market system, when exposed to a well done recording on a powerful but accurate system, that has a good low end, will ask "where's the bass?" (The answer being, of course, that it is there when it is present in the music, and not there when it is not present, as opposed to inexpensive less accurate systems that have a phony mid-bass emphasis all the time).

BTW Acid8000 -- good clarification on dynamic compression vs. data (bit rate) compression. I should have made clearer that my comment about compression to catch people's attention on FM pertained to dynamic compression, not bit rate compression.

Regards,

-- Al

[Baud]

To make the things short, I do totally agree with k0NG0 and ctal.

[Quarkmachine]

Fascinating read!

[TQ]

Well, all I can add is that I can certainly hear the difference between a 256kbps Mp3-encoded song and its original .wav source.
I can also hear the differences in playing back a CD on my studio's Denon DN-C630 CD player then playing back the same song through my PC. Having both routed through my Mackie mixer makes A/B comparisons quite easy.
Science be damned, my ears can discern the difference and if I performed that same A/B for anyone they'd hear the difference, as well.
A 320kbps CBR Mp3 is acceptable, though not the optimal, way for me to listen to music in my truck while I'm on the road. I'll sacrifice some quality for quantity, but 320kbps as far as I prefer to go.


The other day I purchased and downloaded the Moody Blue's song "Your Wildest Dreams" from Amazon.com to see if I could stomach the 256kbps encode and I found that I couldn't.
The Mp3 sounded harsh and brittle to my ears and the detail I know to be in the song was truncated to an unlistenable state. Yargh.
I plan to go over to Best Buy soon and pick up the Moodies greatest hits CD so I can have a ripped .wav of it for an Mp3 CD I'm making for road-trip use.
I'll rip the tune at a CBR 320kbps and compare it with the original. I'm sure it will sound quite a bit better than the 256 I bought from the Amazon store.
It's my un-scientific contention that within those frequencies that are removed during the .wav to Mp3 conversion much musical information and energy lives.
I know.
I can hear it missing.

[k0NG0]

TQ, everyone knows you're a robot. No need hiding it.