DON'T ACCEPT DEFECTIVE CDs

[markgwoodruff]

Sony BMG is shipping CDs which have been improperly mastered: they are digitally clipped. Since a digitally clipped CD is inherently defective, if you have any in your collection, you should return them to Sony for a refund or a replacement that isn't defective. If enough of us do this, it will put sufficient pressure on them to fix the problem. I'm about to return $700+ of newly purchased CDs for this reason.

So far, I've found the following CDs to be defective:

Bjork, Vespertine [2 62653-2 01] (example at track 1, 1:00.3545)
Fiona Apple, Tidal [CPDP-000210] (track 1, 0:27.5415)
The Essential Chet Atkins [CTDP-113949] (1, 2:02.7123)
Erasure, Hits [3 R2 73991-2 RE-101] 1, 0:26.3976)
Blood, Sweat, and Tears, Child is Father To The Man remastered (2, 2:04.5293)
Blood, Sweat, and Tears, remastered (2, 5:05.6047)
Tracy Bonham, Down Here (1, 0:59.7975)

To see what I mean, just pull up one of the tracks as a .WAV and look at it with a wave editor (such as the freeware Audacity). Zoom in on a spot and you'll see the tops of the waveform chopped off: digital clipping.

If any of your friends have any such CDs, have them return them as well. Let's ensure that Sony doesn't put out such crappy products in the future!

[ezkcdude]

Unfortunately, it's not just Sony. It's a fairly widespread phenomenon.

[mwphoto]

Just out of curiosity is the clipping at the max level for a WAV or is it clipped at some level below MAX? Also how many samples does it tend to affect?

The reason I'm asking is that it should be fairly easy to write a tool to scan a losslessly encoded collection and identify the clipping. From that it would be fairly efficient to generate a more comprehensive database of affected CDs

I'd be interested in knowing which CDs in my collection exhibit this. . .


Malcolm

[Skunk]

Quoting markgwoodruff
> Fiona Apple, Tidal [CPDP-000210] (track 1, 0:27.5415)

I heard it the first time I listened. It amuses me how many audiophiles
around here recommend this album.

Sony must have brazen balls to jeopardize such a successful line of
hardware with childish software pranks and mishaps. Thanks, Sony, for
yet another reason to leave your product on the shelf!

[Skunk]

Quoting "Miller, Jeffrey Scott" <jscmille (AT) iupui (DOT) edu>:

> Quoting markgwoodruff
>> Fiona Apple, Tidal [CPDP-000210] (track 1, 0:27.5415)

oops, I confused this with Extraordinary Machine, which I thought I
heard clipping in as well.

[Steven Moore]

Are the tracks clipped when using a mac or linux, anyone know?

[mattg]

I wouldn't call it a defect. They are poorly mastered. Lots of CDs are. I can imagine what will happen when you go back to the store and say "These are defective because they clip". Unless you bought them at the last remaining music store that actually has knowledgeable employees, the response will be "Huh? Uh, you can't return them since you opened them already. Go grab another copy and see if it's better."

[Skunk]

Just out of curiosity is the clipping at the max level for a WAV or is it clipped at some level below MAX? Also how many samples does it tend to affect?
I'd be interested in knowing which CDs in my collection exhibit this. . .

Not knowing if my email responses are going to show up, I'll post this screenshot of Fiona Apple's please please please from extraordinary machine. They're short, but audible- IMO.

Getting your money back, as noted, is another story.

[jonheal]

Not knowing if my email responses are going to show up, I'll post this screenshot of Fiona Apple's please please please from extraordinary machine. They're short, but audible- IMO.

Getting your money back, as noted, is another story.
Haven't you thrown away that record yet!

[bludragon]

How do you know its audible if you don't have the non-clipped version to compare it with?

How do you know the artist didn't intend to create the 'clipped' sound?

kidding aside I myself was quite disappointed to notice clipping on System of a Down's last two albums including the stuff on the 'free' dvd. And in fact I suspect it is present on nearly all recent commercial cd's, especially rock and pop, as a way of boosting volume beyond normalisation and compression. Or just because a recording engineer doesn't understand digital, and having the overload indicators light up is a good thing!

[markgwoodruff]

Playing a clipped CD at realistic volumes WILL destroy your speakers just the same as having an amp that clips. Playing one at lower volume will destroy your speakers too, but it will take a little while longer.

Here's a shot from Tracy Bonham's Down Here track 1, Freed.

The only way to resolve this is to complain and return en masse. If it costs Sony more to deal with us complaining than to fix it, then they'll fix it.

We'll see what the Sony BMG music club does. I just bought over $700 worth of CDs and they're all going back if they don't fix this.

[Skunk]

having the overload indicators light up is a good thing!

Like an amp that goes to eleven?

[jonheal]

Like an amp that goes to eleven?
Hey, the SqueezeBox goes to 11!! I'm throwin' that POS out!

[jonheal]

Hey, the SqueezeBox goes to 11!! I'm throwin' that POS out!
That was a joke, by the way, although the little machine did cause me some grief yesterday.

[jonheal]

Playing a clipped CD at realistic volumes WILL destroy your speakers just the same as having an amp that clips. Playing one at lower volume will destroy your speakers too, but it will take a little while longer.

Here's a shot from Tracy Bonham's Down Here track 1, Freed.

The only way to resolve this is to complain and return en masse. If it costs Sony more to deal with us complaining than to fix it, then they'll fix it.

We'll see what the Sony BMG music club does. I just bought over $700 worth of CDs and they're all going back if they don't fix this.
No one's going to take back $700 in CDs. You'll have to strap on bombs and entrench yourself at Sony's corporate HQ.

[mwphoto]

Hi Mark,

Thanks for your research on this. Would you be able to extract a portion of clipped info to a wav file and mail it to me? I only need a fraction of second or so so there should be no real copyright issues.

If I have the fragment I'll take a detailed look and see if I can develop an analysis tool to find this effect across an entire collection.

I could start from scratch, but I'd like to start from something that I know is bad, and from an official source (ie Sony).

Cheers

Malcolm

[Pale Blue Ego]

This is just giving the music companies even more reason to try to block ripping music from CDs. Wiseacres will load the tracks in wav editors and expose the rotten recording and mastering mistakes. Life sure was easier for these guys before computers came along.

[benthos]

Playing a clipped CD at realistic volumes WILL destroy your speakers just the same as having an amp that clips. Playing one at lower volume will destroy your speakers too, but it will take a little while longer.

Why would clipping destroy your speakers?

Chris

[CardinalFang]

Why would clipping destroy your speakers?

Chris

In simple terms, if it's for long enough it's just about equivalent to placing a DC voltage across the coils and they will overheat and burn out. You can do this by clipping a low power amp as well - the output become close enough to DC to do the same thing.

[markgwoodruff]

I plan to put short samples of the defective CDs, analysis under Audacity, and a list of the defective ones on my web site (www.markgwoodruff.com). I'll let you know when it's ready (not today).

I'm also going to contact just about everyone who might be interested when I've assembled enough data. This really pisses me off and I'm willing to make as much of a stink about it as needed until they stop this crap.

Grrr....

[ezkcdude]

I plan to put short samples of the defective CDs, analysis under Audacity, and a list of the defective ones on my web site (www.markgwoodruff.com). I'll let you know when it's ready (not today).

I'm also going to contact just about everyone who might be interested when I've assembled enough data. This really pisses me off and I'm willing to make as much of a stink about it as needed until they stop this crap.

Grrr....

mark, you definitely should check out this web treatise on the subject of dynamic clipping:

http://www.mindspring.com/~mrichter/dynamics/dynamics.htm

It's quite enlightening.

[Mike Anderson]

Along the same lines"

I would pay serious money to someone willing to do a high-quality needle drop of the original master of the Stooges' classic album, "Raw Power".

The way in which that once-excellent album was remixed to clipping by Iggy Pop should be a CRIME punishable by DEATH, in my book...

[mecouc]

For the ignorant (liek myself) what does clipping sound like?

[benthos]

mark, you definitely should check out this web treatise on the subject of dynamic clipping:

http://www.mindspring.com/~mrichter/dynamics/dynamics.htm

It's quite enlightening.

From that article, it sounds like clipping of cd's is a fairly widespread phenomenon. If that's the case, I can't understand why my speakers haven't been "destroyed" yet.

Chris

[jonheal]

From that article, it sounds like clipping of cd's is a fairly widespread phenomenon. If that's the case, I can't understand why my speakers haven't been "destroyed" yet.

Chris
The level of clipping in the recordings noted is not going to destroy your speakers. But it may jar your nerves a bit. As has been noted before, clipping is essentially momentary DC current to the speakers. An extended burst of it would result in silence because your drivers would be stuck in either an extended or retracted position. Turn the volume up high enough in that scenario, and you might melt something.

[Mark Lanctot]

Are the tracks clipped when using a mac or linux, anyone know?

They're recorded that way right on the disc. All ripping programs will pass it on.

[CardinalFang]

Along the same lines"

I would pay serious money to someone willing to do a high-quality needle drop of the original master of the Stooges' classic album, "Raw Power".

The way in which that once-excellent album was remixed to clipping by Iggy Pop should be a CRIME punishable by DEATH, in my book...

I shall just have to look very smug and dig out my 1970's copy then! I don't know if it's an original master though, I got it around 1975. I have a Mac with a M-Audio card and have used Audacity before to record some of my old Zappa albums, would that be good enough for you? My deck is a Pink Triangle/Mission 774/Ortofon set-up. No ultra-high end, but damn fine nonetheless. I don't remember any scratches, but I'd have to dig it out and see what state it's in. I havent played it for some years...

Paul

[Pale Blue Ego]

Just saw this article via Digg:

http://www.cdmasteringservices.com/dynamicrange.htm

What Happened To Dynamic Range?

[nicketynick]

I'd like to hear what the labels have to say about this. I'm assuming (you know what is said about that!) that they would say that recordings have been 'market-tested' and 'focus-grouped' to support the move to dynamic compression. If not, why would they do it? I sure hope the answer isn't 'because every one else is'! Because I don't think there is an adequate feedback system for us to convince them to do otherwise - they're only going to blame falling sales on mp3, etc. Perhaps the best way to get the message out is to get some high profile musicians on-board. Does anybody if U2's latest is compressed?
Have to say this has me rather depressed - it's not like we have the option of listening to a compressed vs a clean master to decide for ourselves what sounds better.

[ezkcdude]

For the ignorant (liek myself) what does clipping sound like?

The sound of settling, to quote Death Cab.

[bludragon]

OMG, this thread and those articles explain soooo much. I've always disliked pop music, and the main reason is the fact that the lack of dynamic range makes it dull when quiet, or painful when loud.

It seems to be like the record labels completely dropped the ball on high resolution audio formats - here's what they should have done:

1) 96kHz/24 bit sample rate, or equivalent (because we can more than because its needed) - minimum say around 60kHz/20 bit samples.

2) Much more important than 1 - allow the definition of multiple stereo tracks - say up to 10, each with a dsp profile set up by the recording engineer to manipulate loudness/compression etc. All players of this format can then have a control to adjust the dsp effects contained in the profile. This at minimum can have - car/outdoor (as compressed as it will go), super loud (like a modern cd), loud (for parties), normal (mildly compressed), studio (for those of us with noise isolating headphones, sound proof living rooms, or kW amplifiers). Multiple tracks are required so that more compression can be applied to e.g. vocal, than drums.

I think this would be marketable. The key would be to make the minimum dsp feature set compulsory - much like an mpeg2 decoder is compulsory in a dvd player. But not limit hi end companies from going beyond the base spec. However maybe 'beyond spec' numbers should be limited to keep things honest.

All of this could be equally applied to the next 'mp3' rather than the next 'cd'

[zooropa320]

I'm guessing that the real reason you don't like pop music isn't due to the clipping...

[CardinalFang]

From that article, it sounds like clipping of cd's is a fairly widespread phenomenon. If that's the case, I can't understand why my speakers haven't been "destroyed" yet.

Chris

It has to be sustained for a long time and at high volume to cause enough heating to damage the coils.

To be honest I'm a little surprised that so many people are shocked about this. Pop music has been compressed to Hell and back for a long, long time and mixed to sound OK on portable and car radios, not audiophile setups. It's been only a few groups like Supertramp and the Blue Nile that have tried to buck the trend with fuller range recordings. Most groups want to sell lots of records and let the final mix be squished in this way. Clipping is part of the deal, it's all part of making music sound louder.

Paul

[Mike Anderson]

I shall just have to look very smug and dig out my 1970's copy then! I don't know if it's an original master though, I got it around 1975.

Cool - Yeah, I grew up with the vinyl version too, which made the CD all that much worse to me.



I have a Mac with a M-Audio card and have used Audacity before to record some of my old Zappa albums, would that be good enough for you?

Sure! Anything would be better than the CD, frankly. I bet you could do waaaay better.

[CardinalFang]

Anything would be better than the CD, frankly. I bet you could do waaaay better.

OK, let me see what I can do over the next few days. I will have to move the Mac nearer to the deck. Getting the levels can be tricky, but at least it won't be clipped!

Paul

PS - since you have a CD of the album, I think we're Ok legally with copies of LPs - anyone know any better?

[Mark Lanctot]

For the ignorant (liek myself) what does clipping sound like?

It will sound muffled, fuzzy and distorted.

Take a look at the graphs posted in this thread. You are missing the top of the waveform. This flat line is actually inaudible since it pushes or pulls your speaker cones out momentarily and keeps them there. Since these peaks are fractions of a second long, the effect is from audibility to distortion (near clipping) to inaudibility and back. It'll sound muffled and fuzzy and your speakers may click or pop as they extend and are held there. The effect will be audible at all volumes but your speakers run the risk of being damaged at moderate to high volumes.

To hear what clipping/distortion sounds like, take a very cheap clock radio, turn it to a station and turn it up almost as high as it will go. The buzziness/muddiness you hear is caused by the very underpowered amplifier clipping and the cheap speaker reaching maximum extension.

In a digitally clipped recording you'll hear this effect slightly on very loud parts. These parts will be muffled and distorted. Your speakers may click or pop at higher volumes.

[Mark Lanctot]

Just saw this article via Digg:

http://www.cdmasteringservices.com/dynamicrange.htm

What Happened To Dynamic Range?

Facinating link. I had heard that CDs were getting louder but I had no idea it was so bad.

This does explain why stuff today not only sounds crappy because of poor artistic quality, it sounds crappy due to poor technical quality as well.

[John Stimson]

In simple terms, if it's for long enough it's just about equivalent to placing a DC voltage across the coils and they will overheat and burn out.Back to engineering class for you!

The reason that a hard-clipped signal is more dangerous to speakers than a nonclipped signal has to due with the energy spectrum of each signal. In a normal music signal, most of the energy content is in the lower frequencies. A speaker rated for 100W may have a tweeter rated for only 10W.

Clipping causes a shift in energy content towards the high frequencies. If you're blasting away at 90W and then start clipping the signal, then more energy than usual is going to be going through the tweeter and you'll toast the voice coil. Playing a high-frequency pure sine wave at high volume would produce a similar effect -- all of the power goes through the tweeter.

Clipping causes problems because it is like an infinitely fast transient -- pretty much the exact mathematical opposite to D.C. D.C. will never kill your tweeters, because it is filtered out by the crossover. And it would take an awful lot of D.C. voltage to thermally overload a woofer.

A low level clipped sample will not slowly damage your speakers. Either the signal exceeds the thermal power rating of the tweeter for a few seconds and fries it, or it doesn't. Heating copper up to 100 degrees will not cause it to eventually melt. Either you exceed the melting point or not. The same goes for the melting/ignition temperature of the insulator.

[CardinalFang]

Back to engineering class for you!.

No, I don't think they'd have me back, I finished my Applied Physics degree 25 years ago!

I was trying to couch the concept in very simple terms and obviously got too, too simple. I didn't actually mention tweeters or woofers and what I said in that very short sentence wasn't wrong, just way over-simplified I guess.

The clipped signal becomes like a square wave, so is at a fixed voltage for a short period, which is why I said it was like DC. True, it isn't a true DC signal, but a square wave, but I was trying to convey that it's the fact that the average signal is the same as the peak for that time, so there is a low "crest factor" and a lot more power to get rid of in the speaker.

Tweeters are more susceptable to power overload and can be damaged because of this low crest factor, i.e. the ratio between the peak signal power and the average (RMS) is small, so there's a lot more power, or heat, to dissipate. Normal music has a much higher crest factor, so less heat to dissipate. It all depends on volume as well, you need to have the amp cranked up too. Mathematically, square waves can be reduced to a infinite series of pure sine waves, which have 0dB crest factor, hence the high power contained in the signal.


DC can kill a speaker if there's enough power there, not all speaker have crossovers, there was a penchant for crossoverless speakers in the 80's I seem to rememeber.

Paul

[Patrick Dixon]

Back to engineering class for you!

The reason that a hard-clipped signal is more dangerous to speakers than a nonclipped signal has to due with the energy spectrum of each signal. In a normal music signal, most of the energy content is in the lower frequencies. A speaker rated for 100W may have a tweeter rated for only 10W.

Clipping causes a shift in energy content towards the high frequencies. If you're blasting away at 90W and then start clipping the signal, then more energy than usual is going to be going through the tweeter and you'll toast the voice coil. Playing a high-frequency pure sine wave at high volume would produce a similar effect -- all of the power goes through the tweeter.

Clipping causes problems because it is like an infinitely fast transient -- pretty much the exact mathematical opposite to D.C. D.C. will never kill your tweeters, because it is filtered out by the crossover. And it would take an awful lot of D.C. voltage to thermally overload a woofer.

A low level clipped sample will not slowly damage your speakers. Either the signal exceeds the thermal power rating of the tweeter for a few seconds and fries it, or it doesn't. Heating copper up to 100 degrees will not cause it to eventually melt. Either you exceed the melting point or not. The same goes for the melting/ignition temperature of the insulator.Digital clipping is a slightly different situation though, because the post DAC reconstruction filter* should bandlimit the signal to the amplifier and speakers.

Amplifier clipping sends the full bandwidth signal directly to the speaker.

* Some NOS DACs dispense with this, and rely on the bandwidth being limited elsewhere in the system.

[nicketynick]

I plan to put short samples of the defective CDs, analysis under Audacity, and a list of the defective ones on my web site (www.markgwoodruff.com). I'll let you know when it's ready (not today).

I'm also going to contact just about everyone who might be interested when I've assembled enough data. This really pisses me off and I'm willing to make as much of a stink about it as needed until they stop this crap.

Grrr....

Just wanted to keep this thread 'current', and show my support for Mark's efforts. If the dissenting voices are loud enough....

[nicketynick]

Wonder if Markgwoodruff is still out there working on his list.... I sure would like to see it!

[jonheal]

Playing a clipped CD at realistic volumes WILLThe only way to resolve this is to complain and return en masse. If it costs Sony more to deal with us complaining than to fix it, then they'll fix it.

We'll see what the Sony BMG music club does. I just bought over $700 worth of CDs and they're all going back if they don't fix this.
I agree that this seems to be problem of epic proportions that has grown worse over time. However, simply attempting to return CDs is ridiculous, because no one's going to take them back. Not buying CDs is an equally ineffective approach. I doubt the labels would attribute a quantifiable drop in sales to anything other than an increase in illegal sharing.

I think the only effective approach would be a massive, and clearly-stated petition, although I have no idea how to organize and pull such a thing off. Sure 10 or 20 of us here are sick of loud, clippy CDs with no dynamics, but does the average Joe give a rat's patooty?

[nicketynick]

I'd like to hear what the labels have to say about this. I'm assuming (you know what is said about that!) that they would say that recordings have been 'market-tested' and 'focus-grouped' to support the move to dynamic compression. If not, why would they do it? I sure hope the answer isn't 'because every one else is'! Because I don't think there is an adequate feedback system for us to convince them to do otherwise - they're only going to blame falling sales on mp3, etc. Perhaps the best way to get the message out is to get some high profile musicians on-board. Does anybody if U2's latest is compressed?
Have to say this has me rather depressed - it's not like we have the option of listening to a compressed vs a clean master to decide for ourselves what sounds better.

My previous thoughts exactly. 10 or 20 of us and the average Joey aren't going to do it, but I'm sure there are some high-profile folks in the industry who are concerned and just need to be nudged along......

[jonheal]

My previous thoughts exactly. 10 or 20 of us and the average Joey aren't going to do it, but I'm sure there are some high-profile folks in the industry who are concerned and just need to be nudged along......
I don't share your optimism. No recording engineer in his/her right mind would master their own recordings at a level that would result in a complete loss of dynamics and regular clipping.

Clearly, they are being told to do so by "high-profile folks in the industry" or company policy, or whatever because of the assertion that, all else being equal, louder sounds better.

It's a bit like the mindset that Wal-Mart has created in both consumers and the manufacturing sector that says a DVD player is supposed to cost $49. Consumers believe that's what they cost to build and manufacturers believe that is all consumers are willing to pay. Whether or not a $49 DVD is a piece of crap is irrelevant.

Now if consumers were to, en masse, inform the manufacturing sector that they believe that a DVD player actually should cost $119 and they're not going to buy another DVD player until they do cost $119, and that their quality increases commensurately, then I'm sure the manufacturing sector would respond positively, but do you think consumers are going to do that any time soon?

[nicketynick]

I refuse to believe (stubborn, huh?) that _everyone_ in the recording industry has lost their minds - however the 'louder is better' camp has been squeakier, and getting all the lubrication. Its time turn up the heat a little bit on our end of the axle. The Walmart DVD analogy doesn't quite ring true - there still is a healthy market for those who sell decent product (Slim?), and I certainly don't/won't spend my money at Walmart - you get what you pay for!

[Skunk]

there still is a healthy market for those who sell decent product (Slim?), and I certainly don't/won't spend my money at Walmart - you get what you pay for!

There is an ocean of great sounding stuff out there. Put your money where your mouth is, and buy more MFSL/DCC/XRCD/HDCD/Telarc/Chesky/etc. discs. You may not be able to find as much popular music, but that's a good thing ;-)

Here's a list of HDCD's that probably sound great on the SB:
http://www.hitbutiken.com/hdcd/index.php?p=base

[crooner]

Ok this thread got me interested in CD peak levels so I downloaded Audacity and analized a few tracks from my collection.

Jack Johnson's latest release sounded a tad compressed and loud so I analized a track from that. Peaks were close to 0 dB but no clipping. Good.

I then analized a "needle drop" recording I made of the legendary 1979 Telarc LP of the Carmen Suite conducted by Leonard Slatkin. This was recorded on the Soundstream digital recorder and mastered by none other than Stan Ricker. This is one of the most dynamic recordings ever made.

Surely enough the needle drop file clipped at the highest levels. The little red lights on the level meters are a nice touch to determine actual clipping.

I figured I must have set the levels too hot on my standalone Philips audio CD-R recorder. I need to re-record this at a lower level, I said.

Then, out of sheer curiosity I decided to analize the ripped WAV of Mercury's 1812 overture. This was mastered from the original master tapes with the help of Wilma Cozart-Fine, the widow of legendary engineer Bob Fine.

Well, lo and behold! It clipped also with the loud cannons at the end!!!

You would think a careful remaster like this would avoid clipping, even though the cannons are incredibly loud. For digital advocates it doesn't make much sense, since analog tape supposedly has narrower dynamic range. How could the master tape clip the resulting CD then?

Given these surprising results, I might not need to re-record my carmen LP after all!

[zooropa320]

Here's a list of HDCD's that probably sound great on the SB:


Does EAC read the extra bits on HDCD's? Has anyone ABX'ed an HDCD vs. non-HDCD of the same recording to determine if they sound different after being ripped?

[Mark Lanctot]

Does EAC read the extra bits on HDCD's? Has anyone ABX'ed an HDCD vs. non-HDCD of the same recording to determine if they sound different after being ripped?

Yes, if you rip to a lossless format and you have no errors the HDCD data is preserved.

I tried this today based on that list. I only had Beck - Mutations (which I don't like) in FLAC and sure enough the HDCD indicator in my receiver lit up on playback.

As to whether it sounds better, some say it does. I didn't listen long, and my receiver can only do HDCD decoding in Stereo and Source Direct (?) and I don't listen in those modes, so I'll probably never know.

[cliveb]

Ok this thread got me interested in CD peak levels so I downloaded Audacity and analized a few tracks from my collection.
I'm interested to hear what features in Audacity you used to detect this clipping. I didn't see any features to analyse levels in the version I have (Audacity 1.2.3). So I downloaded the latest 1.3.0 beta, but still didn't see any level analysis features. Also, when I played back a WAV file that I know to have some serious clipping, the playback meters never lit up their clipping indicators. So can I ask what you did to detect clipping?

[cliveb]

Sure 10 or 20 of us here are sick of loud, clippy CDs with no dynamics, but does the average Joe give a rat's patooty?
The average Joe listens on bad equipment. Indeed, a growing number of average Joe's only listen on iPods. And there is the problem: stuff that is hypercompressed can actually sound OK on an iPod (or rather, can sound not that much worse than everything else). Example: My wife bought Coldplay's X&Y and says it sounds fine on her iPod. But when we play it on the main system via the SB2, it sounds absolutely dreadful. (Doesn't bother me - I can't stand Coldplay - but my wife is pretty pissed off about it).

[Skunk]

I'm interested to hear what features in Audacity you used to detect this clipping.

You can use the zoom tool to view down to the individual sample level, then scroll through the waveform.

[cliveb]

You can use the zoom tool to view down to the individual sample level, then scroll through the waveform.
OK, that would work, but it's a tad on the labour-intensive side. I was hoping crooner would be able to point me at some facility in Audicity that I've missed.

[Skunk]

As to whether it sounds better, some say it does.

I always attributed the good sound of my HDCD discs to careful mastering techniques, moreso than the proprietary decoding. They don't cost any more than standard cd's though, so I buy them when I see them- even though I don't have a decoder.

I thought there was a theory that HDCDs should sound worse on a non-HDCD DAC, compared to standard redbook. I'm not sure if that's true. I only suggested them because they seem to sound better than 77% of the standard redbook cd's I own, but the best of the redbooks DO sound better on the SB (to me) than HDCDs.

The Garcia Grisman and Rice Pizza Tapes are HDCD, and don't sound *as* good as Garcia-Grisman Self Titled. Nor does Mutations sound better than the standard cd version of Sea Change. Neil Young's Silver and Gold HDCD doesn't sound nearly as good as his newest, Prarie Wind, which is a standard disc. And while Mickey Hart's Supralingua is an incredible sounding HDCD, I have plenty of redbooks that blow it away.

I have some standard discs that sound better than MFSL cd's also, but that won't keep me from buying more gold discs. IMO it's all about the artist and engineer cooperating, and the label staying out of the way. I have a feeling coldplay just _doesn't care_ how their cd's sound. It's THEIR fault, not the label.

[Mark Lanctot]

I thought there was a theory that HDCDs should sound worse on a non-HDCD DAC, compared to standard redbook. I'm not sure if that's true. I only suggested them because they seem to sound better than 77% of the standard redbook cd's I own, but the best of the redbooks DO sound better on the SB (to me) than HDCDs.

Exactly 77%? C'mon man, be precise! 77.2% if you please! :-)

Unfortunately the point is moot with me because I don't listen in the modes my receiver does HDCD decoding in. That part is a head-scratcher for me. I'm guessing the HDCD decoding must occupy enough of the DSP that it cannot apply any other DSP mode like PLIIx simultaneously.

[crooner]

I just used the regular meters on the top center which are calibrated in the classic brickwall fashion. Anything exceeding the theoretical maximum output level (represented as 0 dB) will light a couple of red bars. This represents digital overload or clipping. Even after the event, the red lights will remain on.

A zoom in of the actual waveform will roughly show clipping as well. The two red lights confirmed it.

If you can get a hold of Mercury's recording of Tchaikovsky's 1812 Overture, try it! :-)


I'm interested to hear what features in Audacity you used to detect this clipping. I didn't see any features to analyse levels in the version I have (Audacity 1.2.3). So I downloaded the latest 1.3.0 beta, but still didn't see any level analysis features. Also, when I played back a WAV file that I know to have some serious clipping, the playback meters never lit up their clipping indicators. So can I ask what you did to detect clipping?

[crooner]

Another thought...
How accurate is this "audacity" software?
It's still hard to believe the fabled Mercury 1812 CD suffers from clipping!

[cliveb]

I just used the regular meters on the top center which are calibrated in the classic brickwall fashion. Anything exceeding the theoretical maximum output level (represented as 0 dB) will light a couple of red bars. This represents digital overload or clipping. Even after the event, the red lights will remain on.
Can I just confirm that you were playing back a WAV file (presumably ripped from a CD) when you saw the clipping indicators come on? The problem I have is that when I played back a WAV file that had plenty of clipping at 0dB, I never got the clipping indicators to come on. It seems that Audacity somehow behaves differently on our two machines.

If you think about it, a WAV file cannot contain any samples which *exceed* 0dB. Any digital clipping detection can only ever think that clipping has probably happened (eg. more than N consecutive 0dB samples); it can never know for sure that the signal was not intended that way.

[markgwoodruff]

I'm still here, and I'm still pissed off about it. I just haven't the time I've needed to address the problem.

As many of you have found out, there are no tools for easily detecting this. The closest I've found is using "sox -e stat whatever.wav" to find tracks that have 0dB peak levels on them. After that, it takes a visual inspection of them with Audacity, which is very time consuming. (I'm making a distinction between hard limiting, where the waves look flat, but the samples still have very small differences between them, and flat-out clipping. Excessive hard limiting may sound bad, but won't fry your tweeters. Hard clipping will; it represents a defective product.)

I've been tied up with other things so I haven't been able to proceed. As soon as tax time is done, though, it's near the top of my list.

My thought is to assemble the data and then start a seige against one record label in particular: Sony BMG. Although the others may be doing it also, I think picking one initial target would be more effective. My current strategy:

Document the problem. Get members of the professional community involved. Start a letter writing campaign, focusing on the popular press at first: Consumer Reports, Home and Garden, Popular Science, New York Times...essentially anybody who might pick it up as an interesting story. Move onto the professional and audiophile press. Contact various Sony investors and argue that this is why their revenue is down. Basically, contact as many people that are affected, which is just about everyone, from as many people as possible.

I don't so it being an easy battle. But given the choice is between fighting, which will take a lot of time, energy, and money, and losing any hope of having listenable music in the future, I'm dusting off my armor and stocking my war chest. You have to pick your battles in life. This is one near and dear to my heart, one that I'm willing to fight for.

[crooner]

They are PCM WAV files decoded from FLACs. It should be identical to the original WAV extracted by EAC from the original CD.

Since WAVs comprise PCM raw data in a windows compatible format, they should in theory contain the clipped waveform originally on suspect CDs.

Perhaps someone with better digital audio knowledge could enlighten us on this....




Can I just confirm that you were playing back a WAV file (presumably ripped from a CD) when you saw the clipping indicators come on? The problem I have is that when I played back a WAV file that had plenty of clipping at 0dB, I never got the clipping indicators to come on. It seems that Audacity somehow behaves differently on our two machines.

If you think about it, a WAV file cannot contain any samples which *exceed* 0dB. Any digital clipping detection can only ever think that clipping has probably happened (eg. more than N consecutive 0dB samples); it can never know for sure that the signal was not intended that way.

[crooner]

So, mark, do the overload (red) lights come on when you play the offending WAV files using Audacity?



I'm still here, and I'm still pissed off about it. I just haven't the time I've needed to address the problem.

As many of you have found out, there are no tools for easily detecting this. The closest I've found is using "sox -e stat whatever.wav" to find tracks that have 0dB peak levels on them. After that, it takes a visual inspection of them with Audacity, which is very time consuming. (I'm making a distinction between hard limiting, where the waves look flat, but the samples still have very small differences between them, and flat-out clipping. Excessive hard limiting may sound bad, but won't fry your tweeters. Hard clipping will; it represents a defective product.)

I've been tied up with other things so I haven't been able to proceed. As soon as tax time is done, though, it's near the top of my list.

My thought is to assemble the data and then start a seige against one record label in particular: Sony BMG. Although the others may be doing it also, I think picking one initial target would be more effective. My current strategy:

Document the problem. Get members of the professional community involved. Start a letter writing campaign, focusing on the popular press at first: Consumer Reports, Home and Garden, Popular Science, New York Times...essentially anybody who might pick it up as an interesting story. Move onto the professional and audiophile press. Contact various Sony investors and argue that this is why their revenue is down. Basically, contact as many people that are affected, which is just about everyone, from as many people as possible.

I don't so it being an easy battle. But given the choice is between fighting, which will take a lot of time, energy, and money, and losing any hope of having listenable music in the future, I'm dusting off my armor and stocking my war chest. You have to pick your battles in life. This is one near and dear to my heart, one that I'm willing to fight for.

[Skunk]

The problem I have is that when I played back a WAV file that had plenty of clipping at 0dB, I never got the clipping indicators to come on.

Audacity 1.2.4 behaves the same for me. I was getting 25 samples in a row at 0db, with no red indicators. If you up the gain 3db the indicators go crazy, of course.

[cliveb]

As many of you have found out, there are no tools for easily detecting this. The closest I've found is using "sox -e stat whatever.wav" to find tracks that have 0dB peak levels on them. After that, it takes a visual inspection of them with Audacity, which is very time consuming.
I didn't want to mention this before in case it sounded like a bit of advertising, but my shareware Wave Repair (www.delback.co.uk/wavrep) might help here. It has a facility to search for clipping (Markers | Find Clipping in Selection). The default behaviour is that 4 consecutive samples at 0dB is considered clipping (while 3 or fewer isn't). The user can change this default and set the amplitude and number of consecutive samples to be considered as clipping (on the Operations tab of the Options dialog).


Document the problem. Get members of the professional community involved.
The problem is that the professional engineers are already well aware of the problem. It's the record companies that are insisting on what's being done, and those mastering engineers do have to eat. He who pays the piper calls the tune.

[Skunk]

He who pays the piper calls the tune.

We pay the piper! :-)

Surely they would just ignore a petition or any other civilized way of trying to communicate. Maybe the EFF would file a class action lawsuit if convinced they can bite at Sony/BMG's heels again.

[ezkcdude]

An easy way to look for clipping is to use MP3Gain to do a batch analysis of your library. It will tell you which songs are clipped.

[poyntzj]

this is what I use which does recognise clipping. Noticable on most new albums, more so the rock / metal ones.
I have no problem with the above and make all tracks 89db so that there are no sudden volume increases between albums as there are otherwise.

I did post a couple of days ago on here as someone else has used MP3Gain and has a problem

[Mark Lanctot]

I'm going to analyze Red Hot Chili Peppers - Californication. I never realized how poorly-sounding this album was. (It's a shame, because the music is good!)

But jeez, nearly every song there's audible clipping (buzzing on loud vocals). ReplayGain seems to recognize that the entire album is recorded way too high - it's giving track gain values between -10 and -12 dB, by far the highest in my collection.

The recording engineer should be ashamed of himself.

[markgwoodruff]

The BMG record club refunded my money for the CDs that I complained about! No explanation why; they just credited the amount to my credit card.

I'm not sure if someone there actually acknowledges the problem (but may be powerless to do anything about it) or if they just got sick of my bitching. (In one case, I returned the same CD four times because the mother had been incorrectly made, causing every single pressing to have the same, visible defects).

Anyway, I'm thinking of a softer approach now. Perhaps there are some sympathetic ears we could reach at Sony who might do something about it. After all, the idea is to get them to master the music tolerably, not just raise a ruckus.

In the mean time, I'll check out mp3gain and Wave Repair and see how they do at detecting these.

[Mark Lanctot]

The BMG record club refunded my money for the CDs that I complained about! No explanation why; they just credited the amount to my credit card.

I'm not sure if someone there actually acknowledges the problem (but may be powerless to do anything about it) or if they just got sick of my bitching. (In one case, I returned the same CD four times because the mother had been incorrectly made, causing every single pressing to have the same, visible defects).

Anyway, I'm thinking of a softer approach now. Perhaps there are some sympathetic ears we could reach at Sony who might do something about it. After all, the idea is to get them to master the music tolerably, not just raise a ruckus.

Wow, knock me over with a feather, congratulations Mark!

Unfortunately it doesn't address the issue, but there seems some hope.

[opaqueice]

It's clear to me, and probably to everyone else reading this thread, that this sort of digital clipping is a bad idea, unless you purposely want the music to sound disorted. It's simply a stupid thing to do - it reduces the dynamic range and makes music sound worse. However I'm very skeptical of the idea that it will "fry your speakers".

Post number 8 on the first page of this thread shows an example, and it's pretty clear from looking at that there isn't going to be any problem with the speakers. The transitions due to the clipping are less sharp than some others already present in that sample (for example look at the little peak close to the middle), so there can't be a problem with high-frequency harmonics blowing the tweeter. What was earlier described as "DC" - the flat clipped parts of the waveform - are around .001 seconds long, much farther from DC than, say, a 50 Hz sine wave, so again, no problem. Of course it might be that some more significantly clipped samples could be so bad as to cause a problem, but I think this is pretty unlikely. Does anyone have any evidence of speaker damage from this actually occuring?

I'm pointing this out because, if people really want to stop this, talking about speaker damage isn't the way to go - it will probably just be dismissed by audio engineers as a cranky rant and ignored. Rather, it seems to me a more powerful argument is to simply point out that this a silly misuse of the digital format, and sounds bad!

[Mark Lanctot]

I'm going to analyze Red Hot Chili Peppers - Californication. I never realized how poorly-sounding this album was. (It's a shame, because the music is good!)

But jeez, nearly every song there's audible clipping (buzzing on loud vocals). ReplayGain seems to recognize that the entire album is recorded way too high - it's giving track gain values between -10 and -12 dB, by far the highest in my collection.

The recording engineer should be ashamed of himself.

Here's a shot of the whole track "Parellel Universe", where the chorus is particularly clipped. Kind of a clicky/staticky/buzzy sound over the vocals and loud guitar chords. Very unpleasant to listen to.

No wonder why ReplayGain went nuts on this file and assigned it a -14.17 dB track gain! The highest adjustment in my 2400-track collection.

There's also a zoom in of 5-sample then right beside it, 7-sample clipping. 4- and 8-sample clipping in the opposite channel. Nasty.

[Mark Lanctot]

I'm pointing this out because, if people really want to stop this, talking about speaker damage isn't the way to go - it will probably just be dismissed by audio engineers as a cranky rant and ignored. Rather, it seems to me a more powerful argument is to simply point out that this a silly misuse of the digital format, and sounds bad!

Yes, you've got a point there. But yeah, it does sound bad, and what's worse is there's nothing that anyone can do about it. It's right there on the source - we're literally missing the top of the waveform, and attenuating it won't bring the lost data back.

[crooner]

So...did you get the red clipping indicators to light?
It seems I'm the only one who's seen this...


Here's a shot of the whole track "Parellel Universe", where the chorus is particularly clipped. Kind of a clicky/staticky/buzzy sound over the vocals and loud guitar chords. Very unpleasant to listen to.

No wonder why ReplayGain went nuts on this file and assigned it a -14.17 dB track gain! The highest adjustment in my 2400-track collection.

There's also a zoom in of 5-sample then right beside it, 7-sample clipping. 4- and 8-sample clipping in the opposite channel. Nasty.

[Mark Lanctot]

So...did you get the red clipping indicators to light?
It seems I'm the only one who's seen this...

Yes, you're right, my clipping indicators don't light.

For that matter, where are the clipping indicators?

[crooner]

They are the last couple of indicator lights on the level display. They represent overload after the digital maximum level of "0 dB". With Audacity and certain CDs such as the Mercury 1812 mentioned in an earlier post, the meters will light red when they detect clipping and will remain lighted for the remainder of the file (track).

Below, I have marked in red the two little bars that should light if clipping is detected.

I'll try to take a snapshot of my screen reproducing the mercury disc extracted file later

[Mark Lanctot]

They are the last couple of indicator lights on the level display. They represent overload after the digital maximum level of "0 dB". With Audacity and certain CDs such as the Mercury 1812 mentioned in an earlier post, the meters will light red when they detect clipping and will remain lighted for the remainder of the file (track).

Huh. I wonder if that setting is adjustable, i.e. after "x" number of samples at 0 dB FS it'll indicate clipping. Mine definitely didn't indicate clipping, yet it was audible. Whenever I had a passage like that indicated, it had that buzzy/crackly sound.

BTW if anyone else has this CD (or "Greatest Hits") and you want to find out what I mean it's during the chorus with the words "California King".

[crooner]

Ok here it is.

This is a WAV file extracted using EAC of Tchaikovsky's 1812 Overture, Mercury Living Presence version.

You can clearly see the two red bars light up during the cannon shot passages.

[Mark Lanctot]

Ok here it is.

This is a WAV file extracted using EAC of Tchaikovsky's 1812 Overture, Mercury Living Presence version.

You can clearly see the two red bars light up during the cannon shot passages.

Oh! I thought you meant the two entire bars lit up red. Those little bits are so small I might've missed it.

I'll check it again.

[Mark Lanctot]

Well I get da-aa-ang close, but no clipping.

The attached screenshot is about the closest I get.

I disagree with Audacity's assessment, it's clearly audibly clipped.

Two other things:

1. I also see several spots of "negative clipping" where the signal peaks at -1.0. The Audacity documentation indicates this is equivalent to "positive clipping" and produces distortion as well.

2. As mentioned earlier in this thread, CD mastering volumes have been creeping up over the years. A brief sample indicates my older CDs have track ReplayGain values of -1 dB or even +0.5 dB. But recent CDs have track RG values of -8 to -12 dB.

[bsquare]

The average Joe listens on bad equipment. Indeed, a growing number of average Joe's only listen on iPods. And there is the problem: stuff that is hypercompressed can actually sound OK on an iPod (or rather, can sound not that much worse than everything else). Example: My wife bought Coldplay's X&Y and says it sounds fine on her iPod. But when we play it on the main system via the SB2, it sounds absolutely dreadful. (Doesn't bother me - I can't stand Coldplay - but my wife is pretty pissed off about it).

there is actually a really simple way to make coldplay sound better on any system: buy a radiohead album instead. works every time!


bb

[cliveb]

Ok here it is.

This is a WAV file extracted using EAC of Tchaikovsky's 1812 Overture, Mercury Living Presence version.

You can clearly see the two red bars light up during the cannon shot passages.
I have been trying to replicate your ability to get the clipping indiciators to come on. I prepared a WAV file with known clipping. When played in Audacity, the indicators did not come on. So I then applied some amplification (1.0dB added using Effect | Amplify, with the "Don't allow clipping" option turned off). Playing the result caused the clipping indicators to come on, as we might expect.

Now here comes the really spooky bit. I saved the amplified file back as a brand new WAV file. For sure, it has more clipping than before, but it's still just a 16 bit WAV file with some 0dB samples in it. But when it was loaded back into Audacity (a fresh session, to make sure it didn't have anything cached in memory) and played, the clipping indicators came on! I checked with a hex editor that it didn't have any additional RIFF chunks with extra Audacity-specific data in it. How does Audacity know that this particular 16 bit WAV file has clipping, when it didn't detect it in the other file? I then prepared a third file (using a different audio editor) with even MORE clipping than the one which caused the clipping indicators to come on, and when this was played in Audacity, the indicators did NOT come on. This really doesn't make sense.

The only rational explanation I have for this behaviour is that Audacity must keep some kind of database with information about files it has processed.

[opaqueice]

The only rational explanation I have for this behaviour is that Audacity must keep some kind of database with information about files it has processed.

Couldn't it be that it has a standard for what it decides to call clipping which isn't the same as yours? For example, a single 0dB sample, or many seperated from each other, shouldn't automatically be considered clipping (might be just full use of the dynamic range). What about two 0dB samples in a row? Maybe it has some threshold which has to be exceeded before it warns you, and your modified file passed it while the others didn't.

[Mark Lanctot]

The only rational explanation I have for this behaviour is that Audacity must keep some kind of database with information about files it has processed.

...and unfortunately, since you first processed a file with known clipping and it did not indicate clipping, it's inaccurate.

I notice the documentation is inaccurate too: it indicates that clipping is samples "beyond" 1.0 or -1.0. How is this possible with a CD? If you have a sample at 0 dB FS, it can't get any louder than that, as shown by the examples of clipping in this thread - it goes to 0 dB FS and stays pegged there for a few samples. Am I missing something?

[cliveb]

Couldn't it be that it has a standard for what it decides to call clipping which isn't the same as yours? For example, a single 0dB sample, or many seperated from each other, shouldn't automatically be considered clipping (might be just full use of the dynamic range). What about two 0dB samples in a row? Maybe it has some threshold which has to be exceeded before it warns you, and your modified file passed it while the others didn't.
It's not that straightforward. I've attached two screen shots of highly zoomed waveforms. TEST2.JPG shows a fragment of waveform that causes Audacity's clipping indicators to come on. TEST3.JPG shows a fragment of waveform which does NOT cause the indicators to come on. And yet, the number of consecutive 0dB samples is GREATER in test3 than test2. This makes no sense to me.

[edit - replaced PNG files with JPGs for better quality]

[crooner]

Yep, that's what I suspected. Audacity is not accurate enough to detect any digital clipping present in the original recordings. The clipping on the Mercury 1812 Overture track is very odd. One would think Wilma Cozart-Fine and the mastering engineer would keep an eye on this!

[opaqueice]

It's not that straightforward. I've attached two screen shots of highly zoomed waveforms. TEST2.JPG shows a fragment of waveform that causes Audacity's clipping indicators to come on. TEST3.JPG shows a fragment of waveform which does NOT cause the indicators to come on. And yet, the number of consecutive 0dB samples is GREATER in test3 than test2. This makes no sense to me.

[edit - replaced PNG files with JPGs for better quality]

Yeah, certainly doesn't look good that test3 doesn't light up the clipping indicators. My guess is programming error of some sort - the test-for-clipping algorithm is probably just broken.

[bsquare]

Yep, that's what I suspected. Audacity is not accurate enough to detect any digital clipping present in the original recordings. The clipping on the Mercury 1812 Overture track is very odd. One would think Wilma Cozart-Fine and the mastering engineer would keep an eye on this!

have you considered that they _did_ keep an eye on things and determined that, given the limited dynamic range of their recording technology, they would get a better recording by letting the cannons clip so that the rest of the music wouldn't be lost in the noise floor?


bb

[crooner]

That's a possibility. But consider this was digitally remixed from the original 3 channel analog tapes. Supposedly the new digital mix (spars ADD) would yield a reduction in noise and increased dynamic range.


have you considered that they _did_ keep an eye on things and determined that, given the limited dynamic range of their recording technology, they would get a better recording by letting the cannons clip so that the rest of the music wouldn't be lost in the noise floor?


bb

[cliveb]

Yeah, certainly doesn't look good that test3 doesn't light up the clipping indicators. My guess is programming error of some sort - the test-for-clipping algorithm is probably just broken.
OK, I've got to the bottom of this. Bear in mind that these experiments are with 16 bit WAV files (ie. the sort you get when you rip a CD):

Audacity only switches on the clipping indicators if it sees samples with an absolute numeric value of 32768. This is unfortunate, because the positive swing of the waveform can only ever reach a maximum value of 32767. In other words, a waveform that has massive clipping in the positive direction will never switch on the indicators. Only clipping in the negative direction (where sample values can reach -32768) is detected. In the TEST3 file that I prepared, the clipping in positive and negative directions was at +32767 and -32767, which explains why Audacity's indicators don't come on. I used Ultraedit (a hex editor) to change some sample values to -32768 and then the indicators came on.

After a little more experimentation, it looks like Audacity needs to see 4 consecutive samples at -32768 before it considers that clipping has occured: 3 consecutive -32768 samples do NOT switch on the indicator. It's interesting that they chose 4 as the number. When I wrote Wave Repair, I also used 4 consecutive full scale samples as the default clipping threshold. I chose the number because that's what my Tascam DAT recorder used (and I figured what was good enough for Tascam would do for me). There must be some kind of informal industry standard for this (not that I've ever seen it written down). Or maybe the Audacity authors also happen to use Tascam DAT recorders :-)

[crooner]

Thanks for your research cliveB, very useful information!

[bsquare]

That's a possibility. But consider this was digitally remixed from the original 3 channel analog tapes. Supposedly the new digital mix (spars ADD) would yield a reduction in noise and increased dynamic range.

i think you missed my point: i'm suggesting the original tapes have clipped audio on them because either the cannon exceeded the range of the mics or they just didn't have a good way to make a decent recording of the rest of the music without clipping on the cannon.

all the increased range of the new technology would do is reproduce the clipped original recording.


bb

[bsquare]

[b]
So far, I've found the following CDs to be defective:

Bjork, Vespertine [2 62653-2 01] (example at track 1, 1:00.3545)


for a long time i only listened to this on my ipod and i thought the mighty power of bjork was exceeding its meager capabilities. later, i tried it on a real system and made a frowny face. as much energy as she pours into production, you'd think she could manage to resist going LOUD.

i actually have several versions of vespertine: CD, hybrid SACD, and audio DVD (not DVD-A). the DVD version is _not_ the same mix (especially noticeable in unison where there is all sorts of new, wonderful, atmospheric sound clicking and chirping in the background), so i wonder if it has similar clipping problems. time to rearrange the gear so i can play discs again. spoiled by squeezebox.


bb

[crooner]

What's impressive is that an older analog recording could tax the digital system to the max, as the waveform displayed on Audacity clearly shows.

When remixing, the mastering engineer could have recorded the original 3 track signal (clipped or not) at a lower level. This is for the final two track digital master used to manufacture the CD.

Any artifacts, such as clipping, from the original tapes could still be heard, but the average level of the digital master would be in the normal range...


i think you missed my point: i'm suggesting the original tapes have clipped audio on them because either the cannon exceeded the range of the mics or they just didn't have a good way to make a decent recording of the rest of the music without clipping on the cannon.

all the increased range of the new technology would do is reproduce the clipped original recording.


bb

[bsquare]

What's impressive is that an older analog recording could tax the digital system to the max, as the waveform displayed on Audacity clearly shows.

When remixing, the mastering engineer could have recorded the original 3 track signal (clipped or not) at a lower level. This is for the final two track digital master used to manufacture the CD.

Any artifacts, such as clipping, from the original tapes could still be heard, but the average level of the digital master would be in the normal range...

it's possible it's an error or an intentional bad move, but i have so few classical discs with clipping issues (one particularly egregious example on an arvo part disc springs to mind) that it just seems more likely it was done on purpose, for a good reason.


bb

[cliveb]

What's impressive is that an older analog recording could tax the digital system to the max, as the waveform displayed on Audacity clearly shows.
Actually, all the clipping reported by Audacity tells us is that the person who transferred the recording to digital allowed digital clipping to occur. No analogue tape (with the possible exception of a Dolby SR recording) can come close to exceeding the dynamic range available on CD, so in principle there is no excuse for getting this wrong. But as someone pointed out elsewhere in this thread, maybe keeping the canons within headroom would have rendered the rest of the disc so quiet it wasn't an option. There again, some kind of graceful limiting could have been used to avoid hard digital clipping. Unless there is some subtle reason known only to professional mastering engineers for allowing this clipping, I have to say it smacks of negligence.


Any artifacts, such as clipping, from the original tapes could still be heard, but the average level of the digital master would be in the normal range...
If this was indeed originally recorded onto analogue tape, then "clipping" isn't something that would have happened. When the signal level gets too high for analogue tape, it gets progressively "squashed": in effect analogue tape behaves as a pretty good (and fairly graceful) peak limiter, albeit with lots of distortion.

[earthbased]

This is an excellent discussion as I have learned a lot. I have an Eagles tune that I was listening to a few weeks ago with my brother and we noticed distortion on the lead guitar at a certain point at moderate volume. Now I am wondering if this is clipping as I assumed something went wrong in the ripping to FLAC. Will have to dig out that CD and do a comparison.

[earthbased]

This is an excellent discussion as I have learned a lot. I have an Eagles tune that I was listening to a few weeks ago with my brother and we noticed distortion on the lead guitar at a certain point at moderate volume. Now I am wondering if this is clipping as I assumed something went wrong in the ripping to FLAC. Will have to dig out that CD and do a comparison.

Well I just got to test "Peaceful Easy Feeling" by The Eagles using SB2 and CD on two different stereos. When the guitar solo comes in wait for the second part where it intensifies and you will hear distortion which must be due to clipping. This http://www.cduniverse.com/productinfo.asp?pid=6299958&BAB=E is the 2-CD set and the tune is on disc 1. Will next try to load the FLAC tune into Audacity (once I load that and figure out how to use it) and see if I can find a flat spot at around that time point.

[dagordon]

Please count me in on any crusade to rid the world of digital clipping.

Following the purchase of my SB3 I upgraded my audio gear a bit, from a Harman Kardon mini system to a Charlize amp and a pair of Klipsch bookshelf speakers.

A couple of days after setting up the Charlize I was playing disc 1 of Bruce Hornsby's "Spirit Trail" and noticed some absolutely hideous distortion on track 3, Preacher in the Ring Part I. I began to panic, thinking that it was something wrong with the new amp or speakers, as it wasn't happening with the Harman Kardon system, or a pair of Etymotic ER-ri headphones plugged directly into the SB3 or a portable CD player or my computer.

On inspection of the actual track, though, excerpt here:

http://www.princeton.edu/~dagordon/test.wav

there's digital clipping right where the distortion is occuring (for example, the 's' in 'say' in the excerpt).

What's amazing is, as others have said, that it seems that the higher-end the system (as in, the more revealing), the worse the clipping sounds. On my system it sounds almost like static. Again, on my other equipment I don't hear anything odd at all! This is infuriating.

[SteveEast]

I just ripped The Seekers Ultimate Collection from 2003 and was wondering why it sounded so poor. When I checked the replay gain values they varied from -10 to -12 dB. On the Seekers?! What it does to Judith Durham's voice is criminal.

Steve.

[Mark Lanctot]

I just ripped The Seekers Ultimate Collection from 2003 and was wondering why it sounded so poor. When I checked the replay gain values they varied from -10 to -12 dB. On the Seekers?! What it does to Judith Durham's voice is criminal.

Steve.

Some record exec must have wanted her to sound as loud as Kid Rock or Ozzy Osbourne...

[Craig]

As a frequent contributor to these loudness debates I just want to mention that Wilco's new album Sky Blue Sky is resfreshingly dynamic. Analysing it in Wavelab shows not one single instance of clipping even though peaks do touch 0db. It's easy to play loud and there's even cymbal hits that'll make you jump :-)

Craig

[ezkcdude]

There's a lot of well recorded (and good) music out there. You just don't hear it on the radio.

[Timothy Stockman]

The big problem that causes the "Loudness Wars" is that 9 out of 10 listeners equate loudness with quality. 20 years ago or so, when the loudness wars were taking hold in the broadcast industry, I was the engineer at a classic rock station. I spent about a year enduring comments like "I don't get your station very well, it's not as loud as the others" and finally the general manager told me I had to make our station the loudest in town. Well, I cranked up the processing, but I turned in my resignation about a week later.

Now we've got the same problem with compact discs, 9 out of 10 listeners think louder is better. Marketing types follow what the buying public wants; if 9 out of 10 want louder CDs, the record companies will sell them louder CDs, even if they are distorted. There is a reason that most people think that louder is better. It is because, in comparison, louder does sound better. This is why it is so critical to precisely balance the levels in a ABX test. But most people don't know that, they just "know" that a louder CD sounds better than a "wimpy" disc. Its almost ridiculous these days to see a CD crammed up against the top few bits while the bottom half of the dynamic range is filled with random noise. At least, I assume that's what's down there, because FLAC does a rather poor job of compressing these overly loud discs.

[peejay]

American Doll Posse is actually louder than any other Tori Amos album in my collection. I noticed firstly from the humble VU meters on the SB3. No evidence of clipping, however. It seems we are likely to never escape from this loudness competitive nonsense.

[Phil Leigh]

The big problem that causes the "Loudness Wars" is that 9 out of 10 listeners equate loudness with quality. 20 years ago or so, when the loudness wars were taking hold in the broadcast industry, I was the engineer at a classic rock station. I spent about a year enduring comments like "I don't get your station very well, it's not as loud as the others" and finally the general manager told me I had to make our station the loudest in town. Well, I cranked up the processing, but I turned in my resignation about a week later.

Now we've got the same problem with compact discs, 9 out of 10 listeners think louder is better. Marketing types follow what the buying public wants; if 9 out of 10 want louder CDs, the record companies will sell them louder CDs, even if they are distorted. There is a reason that most people think that louder is better. It is because, in comparison, louder does sound better. This is why it is so critical to precisely balance the levels in a ABX test. But most people don't know that, they just "know" that a louder CD sounds better than a "wimpy" disc. Its almost ridiculous these days to see a CD crammed up against the top few bits while the bottom half of the dynamic range is filled with random noise. At least, I assume that's what's down there, because FLAC does a rather poor job of compressing these overly loud discs.

FLAC does a perfectly fine job of losslessly compressing those bits, whatever their values may be. It is 100% accurate.

[crooner]

Add Paul McCartney's new release (Memory Half Full) to the list of compressed to death and distorted CDs.

This is quite possibly the worst sounding modern recording I have ever heard. Horrible distortion and digital clipping everywhere. The amount of compression is insane. The antithesis of high fidelity, IMO.

[JimC]

A couple of us were discussing this internally, and we have a suspect as to the motivation for what we all consider excessive compression... the listening environment and the record labels reaction to it.

Up to about 15 years ago, CDs were mostly used at home and required substantial investment in both gear and in the CDs themselves. The ambient noise level in an average home is something on the order of 50+dB, and average speech ranges in the 60-70dB. In this environment, a wide dynamic range is totally acceptable, since you can hear the "quiet" parts of the music easily, without the volume turned way up (which cranks the loudness the top end of the range).

Then came the CD Walkman and the advent of car stereos with CD capabilities. Wham! Suddenly, the noise threshold for the listening environment went through the roof (the ambient noice level in a car is around 85dB, for example). To hear the quiet parts of a song, you have to raise the volume for that section to something above 85dB. Since volume is applied across the dynamic range, the loud portions are boosted to very high levels.

The alternative to increasing overall loudness is to apply compression to the audio, lifting the quiet passages above the noise threshold without pushing the loud sections higher. Sure, you lose dynamic range, but you can at least hear the entire piece.

The proper place to apply this compression is non-destructively in the rendering hardware, rather than in the source material. Unfortunately, very few devices can perform compression (either on-the-fly, or from pre-calculated data about the song), so compression ends up being applied to the CD source material. This is destructive to the audio, since you remove dynamic range from the source; performing the compression at the time the audio is rendered allows it to be applied non-destructively and with respect to the actual noise threshold of the environment.

The record labels are misapplying a perfectly valid technology to solve a problem that should be handled in the rendering hardware.

[Timothy Stockman]

FLAC does a perfectly fine job of losslessly compressing those bits, whatever their values may be. It is 100% accurate.
FLAC is lossless (100% accurate), but random bits cannot be compressed. FLAC can only remove "redundancy". Therefore, it cannot get as much compression when the random noise floor is increased. This happens either when the overall audio level is raised (rasing the noise floor as a side effect) or when the bit depth is increased to 24-bits. On 24-bit files, much of the added resolution is usually taken up by the noise floor, since no recording has the insanely low noise floor which 24-bits could potentially provide.

[Phil Leigh]

FLAC is lossless (100% accurate), but random bits cannot be compressed. FLAC can only remove "redundancy". Therefore, it cannot get as much compression when the random noise floor is increased. This happens either when the overall audio level is raised (rasing the noise floor as a side effect) or when the bit depth is increased to 24-bits. On 24-bit files, much of the added resolution is usually taken up by the noise floor, since no recording has the insanely low noise floor which 24-bits could potentially provide.

So, it doesn't do "a poor job" - it does the best job it can under the circumstances.

I'm no fan of hyper-compression BTW.

[Timothy Stockman]

So, it doesn't do "a poor job" - it does the best job it can under the circumstances.
A few quiet discs I have got ratios almost as low as .300; these are recorded at low levels, so there is a lot of redundancy in the "zero" bits at the top. Discs recorded loudly have no corresponding redundancy at the "bottom", because the increased noise floor tends to be random bits, rather than "zero" bits. Therefore, FLAC does a comparatively poor job of compressing them, yielding ratios between .600 and .700. In other words, FLAC achieves about twice the compression on the average "queit" disc as on the average "loud" disc. My experience with FLAC has taught me that a ratio of near .300 is very good, while a ratio of near .700 is rather poor. Of course, even FLAC's poor performance at .700 is still better than an uncompressed copy of the file: 1.000. Snd regardless of the compression ratio, even if it is poor, the file is still decoded losslessly. That is, even the random noise of the noise floor is decoded losslessly, even when the compression ratio is poor.

Think about it, why couldn't a second FLAC pass further compress an already compressed file, yielding an even better compression ratio? The answer: because the first pass already removed the redundancy. With random noise, there's very little redundancy to remove, in the first place, so the higher the noise floor, the less compression is possible.

[Mark Lanctot]

A few quiet discs I have got ratios almost as low as .300; these are recorded at low levels, so there is a lot of redundancy in the "zero" bits at the top. Discs recorded loudly have no corresponding redundancy at the "bottom", because the increased noise floor tends to be random bits, rather than "zero" bits. Therefore, FLAC does a comparatively poor job of compressing them, yielding ratios between .600 and .700. In other words, FLAC achieves about twice the compression on the average "queit" disc as on the average "loud" disc. My experience with FLAC has taught me that a ratio of near .300 is very good, while a ratio of near .700 is rather poor. Of course, even FLAC's poor performance at .700 is still better than an uncompressed copy of the file: 1.000. Snd regardless of the compression ratio, even if it is poor, the file is still decoded losslessly. That is, even the random noise of the noise floor is decoded losslessly, even when the compression ratio is poor.

Think about it, why couldn't a second FLAC pass further compress an already compressed file, yielding an even better compression ratio? The answer: because the first pass already removed the redundancy. With random noise, there's very little redundancy to remove, in the first place, so the higher the noise floor, the less compression is possible.

I suppose all lossless codecs would perform similarly as they try to do the same thing.

I notice when there's one of those dreaded "hidden" tracks at the end of a CD, the FLAC encoder trips up on the silence. It will slow down tremendously. Sometimes these silent portions are 10+ minutes long (Nirvana - Nevermind) which makes for very long encodes.

As you say, there isn't much you can do - the noise is random, there's no redundancy to remove.

[Timothy Stockman]

I notice when there's one of those dreaded "hidden" tracks at the end of a CD, the FLAC encoder trips up on the silence.
Interesting... I never noticed that! I always try to unhide the "hidden" track first by extracting to WAV and splitting it to 2 files in Cool Edit before feeding it to FLAC. I have little use for "hidden" tracks on the hard drive.

[Mark Lanctot]

Interesting... I never noticed that! I always try to unhide the "hidden" track first by extracting to WAV and splitting it to 2 files in Cool Edit before feeding it to FLAC. I have little use for "hidden" tracks on the hard drive.

Yes, I do the same (although I use Audacity) - I have no use for 10 minutes of silence, even if the artist intended it or not. Frankly, I think more than 2 seconds of silence is hard to take.

[Phil Leigh]

A few quiet discs I have got ratios almost as low as .300; these are recorded at low levels, so there is a lot of redundancy in the "zero" bits at the top. Discs recorded loudly have no corresponding redundancy at the "bottom", because the increased noise floor tends to be random bits, rather than "zero" bits. Therefore, FLAC does a comparatively poor job of compressing them, yielding ratios between .600 and .700. In other words, FLAC achieves about twice the compression on the average "queit" disc as on the average "loud" disc. My experience with FLAC has taught me that a ratio of near .300 is very good, while a ratio of near .700 is rather poor. Of course, even FLAC's poor performance at .700 is still better than an uncompressed copy of the file: 1.000. Snd regardless of the compression ratio, even if it is poor, the file is still decoded losslessly. That is, even the random noise of the noise floor is decoded losslessly, even when the compression ratio is poor.

Think about it, why couldn't a second FLAC pass further compress an already compressed file, yielding an even better compression ratio? The answer: because the first pass already removed the redundancy. With random noise, there's very little redundancy to remove, in the first place, so the higher the noise floor, the less compression is possible.


At the risk of splitting semantic hairs, I re-iterate. FLAC does not do a poor job - it simply does the best that is possible. This is true of all RLE compressors. Random in = less compression out!.

The issue is not with FLAC, it is with the source file.

[joshk]

A couple of us were discussing this internally, and we have a suspect as to the motivation for what we all consider excessive compression... the listening environment and the record labels reaction to it.

Up to about 15 years ago, CDs were mostly used at home and required substantial investment in both gear and in the CDs themselves. The ambient noise level in an average home is something on the order of 50+dB, and average speech ranges in the 60-70dB. In this environment, a wide dynamic range is totally acceptable, since you can hear the "quiet" parts of the music easily, without the volume turned way up (which cranks the loudness the top end of the range).

Then came the CD Walkman and the advent of car stereos with CD capabilities. Wham! Suddenly, the noise threshold for the listening environment went through the roof (the ambient noice level in a car is around 85dB, for example). To hear the quiet parts of a song, you have to raise the volume for that section to something above 85dB. Since volume is applied across the dynamic range, the loud portions are boosted to very high levels.

The alternative to increasing overall loudness is to apply compression to the audio, lifting the quiet passages above the noise threshold without pushing the loud sections higher. Sure, you lose dynamic range, but you can at least hear the entire piece.

The proper place to apply this compression is non-destructively in the rendering hardware, rather than in the source material. Unfortunately, very few devices can perform compression (either on-the-fly, or from pre-calculated data about the song), so compression ends up being applied to the CD source material. This is destructive to the audio, since you remove dynamic range from the source; performing the compression at the time the audio is rendered allows it to be applied non-destructively and with respect to the actual noise threshold of the environment.

The record labels are misapplying a perfectly valid technology to solve a problem that should be handled in the rendering hardware.

Absolutely agree! Or, the recording companies can use the new space available with modern formats to provide two versions, one for the road (compressed), one for home (uncompressed).

Even if the recording industry provided two versions of every CD, using simple redbook, one for the road, one for home, then some of the consumers could buy the one for the home. Just a thought.

[Loneseer.....]

Hello all. I was searching for forums on defective CD's and came across this forum.

There is nothing more irritating than a badly mastered CD. That is why I didn't order a CD from CD Baby recently, even though I really liked the music. The woman's voice hurt my ears, because it was so severely clipped. I have a number of CD's from well known companies with the same problem as well and won't listen to them though I really like the songs.

I just got through producing my own CD. And I was wanting to make sure the waveform wasn't severely clipped. I had some problems with this and had to redo the songs being very careful of this digital clipping. My CD sounds quite good now and I hope the music distrubitor doesn't do anything to cause this clipping since they make modifications on people's CDs. Don't think much of their work. Sent them an email as to what I did for this reason.

Digital clipping is such a wide spread problem.

[cgull]

As I absorb this most enlightening thread it occurs to me that what the industry is doing is a microcosm of modern corporation driven USA. It fits perfectly with the "Super Size Me" culture; big cars, big food, loud music etc. Don't get me wrong, I am no tree hugger but am concerned about what is happening in our culture and how it manifests in normal every day things.

I am fairly certain that millions are spent by the record companies on the psychology of music purchasing by the average consumer. If deeper market penetration means altering the music from its natural state then it will be done. That is unfortunate for us purist but lets face it, we are a sliver on the revenue pie chart.

The same thing is happening with video. Walk into any consumer grade a/v store and look at the artificial color bump on the tv's. Better red than dead is the saying for flesh tones. Me and a buddy were talking about how 1080i NFL games appear to have artificially inflated color/saturation levels. You can correct this to some degree but something is surely lost.

This is the age of the iX where X=Pod, Phone, Car, House... well Life. The record industry is transitioning away from the old static mediums into the digital age. Preserving the natural "intended" form of media is not as important as preserving the revenue stream. Kinda makes you wonder where we will be in say 10 years.

Just my .02


----------------
cgull