How to determine a CD's loudness and clipping before buying?

March 11, 2007 12:35 PM

How can you determine if a CD is a casualty of the loudness wars - lots of clipping, mushy sound, and so on - before buying it? Or after buying it, if it comes to that?

Listening to a CD that grated on my ears, I thought about all the good info on the Loudness War I'd read a while ago. I threw some tracks into Audacity to discover, indeed, my ears were right: it looked almost sausage-like. No lows, and I could zoom in to find a ridiculous number of highs clipped.

Are there any resources to look up this kind of thing pre-purchase?

Reviewers seem happy to talk about this CD's place in the post-rock canon but don't mention that it sounds horrible. I'd love to find some place that says "CD x has 2.5 db of headroom, no clipping, and isn't mushy and compressed" and "CD y runs up to 100% loudness, there's horrible clipping throughout, all the instruments sound the same, and if you get through one listen you'll be shocked."

Failing that, are there better ways for me to diagnose this besides loading them into an audio program and just eyeballing it?

posted by dmz to media & arts (19 comments total)

Failing that, are there better ways for me to diagnose this besides loading them into an audio program and just eyeballing it?

Compare it to a reference known-good album at the record store, if there's a listening station available. Listen to the album mixed to your preferences; then put on the disc you're checking out. Too loud? Too compressed? Voila.

Implicit in that is the notion that what matters is how it sounds; if you can't recognize an overly-loud mix just at a listen, what is the goal in IDing such albums?
posted by cortex at 12:48 PM on March 11, 2007

Piggy back question: is there a way for a non-audio genius to somehow mitigate this on a CD they bought?
posted by popechunk at 12:49 PM on March 11, 2007

Professionally mastered CDs don't clip. If you're hearing clipping, then something with your playback setup is messed up. Also, you can't just look at a waveform and determine if the "lows" or "highs" are clipping, only whether or not it's clipping at all (and even then, it would be very difficult to eyeball this on a final mix.)

Are there any resources to look up this kind of thing pre-purchase?

Download it and listen before opening the shrink wrap on the CD.

are there better ways for me to diagnose this besides loading them into an audio program and just eyeballing it? / is there a way for a non-audio genius to somehow mitigate this on a CD they bought?

Many music players can scan and calculate the RMS (average loudness, not peak value) of a song. iTunes can do this, though not very well (turn on "Sound Check" in the options). One of the best implementations is ReplayGain, which is in the popular foobar2000. It allows you to scan albums (or songs individually, though you'll always want to do an album as a whole for albums) to find the average loudness, and then adjusts the volume to average it with the rest of your library. The volume is almost always adjusted down, so that there's clipping, and the adjustment data is stored as part of the song's tag and the adjustment is made as the song is played, so that your data isn't altered.

iTunes' implementation is similar, but it's incompatible with ReplayGain (which was around long before Sound Check) and it screws up albums – each song on the album is calculated individually, which means a quiet song on an album now might be louder compared to the others than it was supposed to be.
posted by tumult at 1:14 PM on March 11, 2007

make that, "so that there's no clipping"

oops.
posted by tumult at 1:15 PM on March 11, 2007

I should have been clearer about what I meant by "clipping" there - I meant that I could look at the waveform and see many peaks with square tops, like the classic sliced-off-mountaintop that shows up in articles on this.
posted by dmz at 1:20 PM on March 11, 2007

I can't find the site again, but there was a website (I believe linked here on MeFi) where this guy listed specific albums and appraised how 'damaged' they were by overloudness, even assigning different categories of infractions.

He did dock albums for things like analog tape saturation, though, which I feel is a perfectly legitimate production technique.
posted by neckro23 at 1:27 PM on March 11, 2007

It's called brickwall limiting, or often just limiting. Clipping means something else entirely, and avoiding it is the whole point of limiting :) (that, and making your song loud as hell)
posted by tumult at 1:28 PM on March 11, 2007

I'm not sure about other places, but every HMV I've ever been to will crack open any CD you give them and let you try before you buy at their listening stations.
posted by furtive at 2:12 PM on March 11, 2007

Tumult, as I understand it, digital clipping is defined as losing musical information by boosting volume past the limits of the medium to resolve. dmz's observation of flat wave tops is a classic symptom.

There's 'amplifier clipping', which is when the amp simply can't produce enough power for a desired volume (result: flattish wave tops), and then there's digital clipping. The official definition may be 'signal at 100%, losing original information", but they cap it at 99.9% instead. It's still losing info, and it still sounds like shit. The flat wave tops are at 99.9% instead of 100%... big freakin' deal.

Arguing that professional CDs 'don't clip' is really just semantics. The actual effect when it's delivered to the listener's ears is the exact same thing.

Replaygain is generally fine. It's very rare that digitally lowering the volume loses significant music information; on these modern mixes, there's virtually never anything down there to care about. It's either slam-to-the-max loud or not there at all. And, even if you do clip off some of the very quietest bits, you're losing stuff that's probably quieter than your heartbeat in your ears. It's not at all obvious or distressing, unlike clipping at the TOP of the volume scale, which is glaringly apparent.

It's hard to find CDs these days that don't do this. All mainstream music does. The whole point of the CD format is dynamic range and low noise floor, and to get an extra scoche of apparent loudness on the radio, they sacrifice that completely.

I'd buy more music if they mixed it decently.

But, yeah, listening in the store or downloading are the only ways to tell without buying it. Assuming it's overmixed, however, is a fine start.
posted by Malor at 2:33 PM on March 11, 2007

(or, of course, eyeballing it with an audio program; that also works. :) )
posted by Malor at 2:35 PM on March 11, 2007

Tumult, as I understand it, digital clipping is defined as losing musical information by boosting volume past the limits of the medium to resolve. dmz's observation of flat wave tops is a classic symptom.

Things that have been brickwall limited only look flat unless you zoom in enough. What limiting does is move the mean level closer to 0db while keeping the peak level under 0db, so more of the peaks are just short of 0db after limiting. Most actual clipping would be on transients (i.e. attacks) and will be too sporadic to see unless you are looking for it. Digital clipping is a harsh crackling noise that sounds very bad when you hear it; it happens when sound actually goes over 0db. If a song was continuously clipping it would be completely unlistenable. Limiting/compression keeps the sound under 0db and sounds completely different, and while it is bad for the sound, it affects the sound in a completely different way (destroying the dynamics, as well as introducing compression artifacts).

There are some free analysis tools, though you need a vst/au host for them. They will actually tell you how much clipping there is, and how much headroom. inspector free is what I use, but that's not because it's the best, but because it is the only UB free one.
posted by advil at 2:54 PM on March 11, 2007

while it is bad for the sound

I should say: "while it can be bad for the sound".
posted by advil at 2:55 PM on March 11, 2007

Commercial albums can and do clip, where clipping is defined as multiple continuous samples at 0 dB, although sometimes the ceiling is actually below that. It does sound bad, but because it's so few samples, relatively, it still sounds like music.

To the original poster: listen before you buy or ask an audiophile friend who already has the album. I don't know of any review sites that provide such statistics. And besides, a good audio engineer would probably tell you that you should be listening with your ears and not getting too hung up on how the waveform looks.
posted by ludwig_van at 3:26 PM on March 11, 2007

You could always download a tester track from one of the various p2p services out there. Most albums are available online in advance of their release date these days, and while it's arguably illegal to do so, downloading one track per album for the purpose of checking the sound quality before purchasing a legal copy seems relatively harmless. Of course, an mp3 copy will be slightly different from an uncompressed original, but for this purpose, it should be sufficient. Oh, and you may open yourself up to being sued by the RIAA. For a less risky approach, try googling for a preview track. Many record labels make one or two tracks available before the release date, and there are hundreds of mp3 blogs out there posting promo tracks as well.
posted by cathodeheart at 8:52 PM on March 11, 2007

If you really wanna go o-taku on this, Post-Ut can start training you to identify dynamic range by ear. Then you could go to the record store, listen to the preview, and just know how much range it has and how much headroom that leaves. How cool is that?
posted by eritain at 10:08 PM on March 11, 2007

Of course, an mp3 copy will be slightly different from an uncompressed original, but for this purpose, it should be sufficient.

I'm inclined to disagree.
posted by ludwig_van at 6:09 PM on March 12, 2007

Why, ludwig_van? Whatever artifacts transcoding to mp3 may introduce, a radical change in the overall dynamics and levels aren't among them. Unless, granted, the transcoder actually purposefully releveled or companded/expanded the source audio, but that's not exactly standard cricket.
posted by cortex at 6:19 PM on March 12, 2007

Yeah, massive changes in dynamics are not typically a hallmark of your standard mp3 encoders. You can certainly set them to alter the dynamics in some cases, but if you just want to find out if the music has already been crushed, I see no problem in 99 out of 100 cases.
posted by cathodeheart at 7:35 PM on March 12, 2007

It probably isn't going to radically change the overall dynamics, but there are too many variables in the interaction between source material and mp3 encoder for me to think it's wise to do any close-up waveform judging of an mp3. mp3s often end up hotter than the source material after conversion, and even non-overcompressed popular music isn't very dynamic to begin with, so it seems very hit or miss. Of course if it's a high quality mp3 and it was encoded well it's going to be a better representation.
posted by ludwig_van at 6:19 PM on March 13, 2007