Definition of Digital Clipping

I see a lot of threads where people are using the term "clipping" and I think it is used incorrectly a lot. I also think that I've used it incorrectly on more than one occasion.

While I was replying to a thread on cd loudness last night, I started thinking about what the term actually means and in what context it is meaningful.

I've created a list of assertions that I believe to be true. Please read them over and offer your comments. Some of these could of course be wrong and I'd like to get a dialog going so that I and other people can learn a little today.

Here goes:

1) Digital clipping happens during encoding, not during playback.
2) During playback, samples that were clipped during encoding produce compression.
3) You cannot distinguish during playback whether a sample was a) clipped or b) limited then boosted to 0 dB.
4) Maybe I should use the term "brickwall limiting" instead of clipping when talking about new CD masterings that are too loud.
5) It is brickwall limiting that is the problem in the loudness wars, not clipping.

Please comment on any/all points that you have an opinion or facts on.

Thanks.

 

Two more:

1) It is not meaningful to describe something that happens during playback as clipping.
2) The digital artifacts that we hear on newer masterings are caused by playback of the square waves that ended up being encoded, not the fact that clipping happened during encoding.

The last one I am the least sure about and would like people to give their thoughts on especially.

 

No. Compression is something totaly different. Clipping causes distortion that sounds like crackling, snapping, or tearing. Compression by itself does not cause clipping.

Not a good idea, and something that too many members here do. Compresseion is used along with limiting to get a CD louder, among other mastering tricks.

Again, "brickwall limiting" is a erroneous term that is used to describe compression that I believe started here on the SH forums.

 

Right. Clipping is not always audible. To say that CD clips just sitting in the case is hyperbole! While i'm at it, to simply say that something is loud is not the same thing as compressed. That kind of tells me that there are an awful lot of people out there, including on this forum, that do not like to touch their volume knob.

 

Does anyone have a REALLY old CD player that can't play back CDs with clipped samples?

I've read that early CD players would simply output silence instead of clipped samples, so the playback would sound as though it was skipping.

 

Hi Grant, you are one of the ones that I wanted to weigh in on this partitcularly. Thanks for the reply.

Would you please comment on #1, that clipping is something that happens during encoding?

Also, if I take a CD that produces this "crackling, snapping, or tearing" sound and reduced it to a max amplitude of -3dB wouldn't I still hear these noises? I will try with Rush "Vapor Trails" tonight when I get home from work and report back.

Isn't 32,767 just as valid as any other amplitude to a DAC? Why would a DAC "clip" if a sample or group of samples hits this level?

Thanks.

 

Not at all. I see it all over the place in audio-related literature. The assertion was made awhile back in another thread that its use for describing limiting is erroneous because it's a term commonly used in descriptions of filtering. However, in describing the concept of peak limiting that cannot go beyond a virtual "brick wall" it's entirely correct and quite common. In fact, both usages are correct as both rely on the concept of a limit beyond which the data cannot pass. I've even seen our host use it on occasion.

 

Yes. You're hearing the ugly noise because the waveform has a very unnatural shape.

Clipping isn't always that big a deal. A bit of clipping during handclaps might not even be noticeable. If vocals or guitars clip then you have more of a problem.

 

It is easy to understand, you have a pad of paper that is a fixed size and you draw something that goes off the page. This is digital clipping, that you are doing it wrong and not fitting it inside the pad of paper you are given (which is MORE than big enough as we know).

 

I've never heard one mute clipped samples.

 

"Loudness" can be introduced at any point in the recording process. For example, if an instrument or voice is recorded hot and stays hot in the mix, during mastering (adding EQ, leveling, compression, erc.) it may get worse. If it's screwed up from moment one, it's not going to get better.

It's OK to hit digital 0, but if you're hitting it all the time, that's brickwalled. Normal dynamic recordings will hit digital 0dB.

Encoding to a particular format - WAV, AIFF, MP3, FLAC, etc. - won't introduce any of the above. So encoding is also a bad term to use.

 

See my post. VT was screwed up WAY before mastering. VT would need to be remixed, and I'm not sure even that can fix it.

 

Clipping can happen either upon encoding or while processing.

If you have a CD with clipping, it's on the CD permanantly. It's f--k'ed! If you simply reduce the level, it will still be there, but your DAC may not add it's own distortion.

 

OK, I do know this, but my objection is that the two get confused. So, I think it's better to not use this term, as it applies to what we usually discuss here. Just my opinion.

But, I believe it was also you in that thread who also made the similar complaint that I did about the use of that term. I still think using "compression" is better. But, even so, as i've pointed out, compression does not cause clipping.

 

It depends on the software. For instance, some digital audio editing programs working in 32-bit floating point mode can work internally well beyond digital zero. For instance, I use Cool Edit Pro, which will preserve a peak beyond 0 until I save it in a standard format that can be read by other programs, at which time it clips any waveforms at 0 dB. In fact, each audio sample value is limited to exactly zero dB, thus producing a square wave.

No, as has already been answered, it has nothing to do with compression. Playback produces the sound of a square wave being played back. Since in most audio recordings of music, the clips are going to be very short in duration, the resulting sound is clicking, popping, rasping in nature and not like a continuous square wave played on a synthesizer.

This is really dependent on the material, the playback level, etc. The limited unclipped material will not have continues samples at 0dB so it should not sound as harsh. This is because limiting simply dictates that the peaks will touch zero dB and then the remainder of the data will be scaled using compression. Depending on the "knee" used for compression, the sound will be smooth or harsh, but rarely as bad as real clipped material.

One thing to remember is that material limited to exactly 0dB can suffer from overshoots between samples that clip on playback.

Some have both problems. However, again it's not as simple as saying either or. Brickwall limiting is definitely a big part of the problem, but what's done to the material before it's pass through the limiting phase can cause worse damage. If during the recording and mixing, heavy compression, boosting of selective frequency bands, etc., is done, the material is already compromised and then the limiting does its own final damage. As for digital clipping, no audio professional should be doing it accidentally, but there are cases, where the material is clipped on purpose to achieve a certain sound!

See my previous answer. It's not an either or thing.

 

Brick Wall limiting does not cause clipping. It's actually used to avoid clipping.

In some software limiters You can chose between Soft (Knee) or Brick Wall limiting. The first one allows some peaks to cross the threshold setting, the later one doesn't let any sample to go over the threshold (suitable for CD mastering).

A medium Brick Wall limiting (1-3db) will sound less harsh than a heavy Soft Knee limiting (5-8 db).

The term Brick Wall is often misused here at the forum, because people refer to a waveform that looks like a brick wall.

 

I define digital clipping as anytime when a digital waveform has not only hit the 0db mark, but has tried to exceed it - in effect slicing the peak of the waveform right off.

This can happen at any stage of the recording process. It can happen if the instrument, voice, or other signal is recorded too loud into the A/D converter. This is the biggest loss IMO because there is no way to recover the lost information afterwards in the mixing and/or mastering process.

It can happen during digital mixing, where the volume in the track or the output from a plugin is so loud it's clipping on the individual tracks, subgroups, or on the stereo mix buss. Here again - we add another stage where clipping can occur - on multiple levels!

And of course, it can happen during mastering. Somebody feeding an A/D converter with a really hot analog signal can introduce clipping, or somebody cranking the volume too high digitally can also introduce clipping. The engineer could even add compression and/or peak limiting (or "brickwall" limiting) and then crank that. And yes, we have all heard examples where clipping has probably been introduced not just in mastering, but in all the examples above.

Also, if I have a chance later I'll try to post some screenshots showing the difference between something that has been processed with a peak limiter and something that has just been clipped. There is a difference.

 

And that's what I was getting at!

 

Alright! You get it!

 

What software uses this terms? "Hard Limiter" is more usual.

The term is always misused here. Don't want to start holy war though... just do not understand why people want to invent wording when we have perfect one: "Clipping"?

 

Come on, it wasn't invented on this forum. It's an accepted industry expression. I saw Steve Hoffman himself using it recently. Are you saying he and everyone else is wrong and you're right?

Just a quick Google will show you it's hardly an invention of Steve Hoffman forum members:

http://www.google.com/search?q=bric...s=org.mozilla:en-GB:official&client=firefox-a

 

iZotpe's Ozone 3 for example.

 

You may be right.

What I mean by "encoding" is the analog to digital conversion itself. I don't mean converting a WAV to MP3 or anything like that.

 

I am very interested to see this.

What you have described (very well too) all happens before playback. Hence the question: Is it proper to speak of a CD "clipping" *during* playback?

What I'm trying to assert is that clipping happens *before* playback and that it is improper to say that a CD "clips". I know that people will probably still misuse the term. This is more for my education than anything.

Thanks all.

 

Correct. If something is clipping in the digital domain - it's just 1's and 0's. You can "hear" the digital clipping as result of the digital info being converted to analog.

Analog clipping is a whole other beast. In the case of playing a CD, analog clipping could potentially be added depending on what follows the D/A converters in your chain (or if something is not calibrated right at the D/A itself). But it is certainly not the fault of the digital information on the CD itself. (Do correct me if I'm wrong, experts )