-Noodles-
3 Initals Mixer
- Dec 20, 2007
- 1,166
- 0
- 36
Clipping analyzer?
- Thread starter ahjteam
- Start date
The software reads the waves, and if there is a clip, the wave is squared off right? The software calculates the full wave as a D/A might output it, and from there determines clips.
Generally it's standard practice to peak masters at -0.3dBFS to prevent the systems at the replicating plant from showing any overs, but with a lot of the current limiting it's very possible to make a cd that doesn't officially 'clip' but still distorts the playback AD convertor, and this analysis software can show you that...
something like that.
Generally it's standard practice to peak masters at -0.3dBFS to prevent the systems at the replicating plant from showing any overs, but with a lot of the current limiting it's very possible to make a cd that doesn't officially 'clip' but still distorts the playback AD convertor, and this analysis software can show you that...
something like that.
-Noodles-
3 Initals Mixer
- Dec 20, 2007
- 1,166
- 0
- 36
So the software is "reading" the waves.. I'll look into this more I think!
And, I was under the impression that Bob Katz has implemented the -0.3dbFS usage? I prefer his ideals of mastering, so it doesn't bother me (that, and I don't master stuff)!
Thanks.
And, I was under the impression that Bob Katz has implemented the -0.3dbFS usage? I prefer his ideals of mastering, so it doesn't bother me (that, and I don't master stuff)!
Thanks.
Clipping becomes audible when a certain number of identical consecutive samples are played. So I would guess the analyzers check for those identical consecutive samples.
-Noodles-
3 Initals Mixer
- Dec 20, 2007
- 1,166
- 0
- 36
You can master hotly, and still have several consecutive samples - which the human ear cannot detect. If my memory is working, this can be as many as 3 samples played consecutively. The identical consecutive samples however, is just the maximum number of 1s [bits] being used (in regards to amplitude) in the word length.
But that still doesn't explain how those samples which are restrained by the Limiter, and then CD medium can be reimported and read to have "clips"* [which I understand are signals going over a certain, defined level. Which the user would presumably set up, - so it could be -0.2dBFS which means that it's not actually clipping; just breaching a predefined level.]
*These clips are probably and more accurately described as digital overs.
Gomez
Member
You got there before me. Although I do pull one of these heresies from time to time when I have been working for too long and my ears are gone... nevertheless, can't trust these things!
tarnationsauce2
Member
- Apr 21, 2007
- 116
- 1
- 16
BTW guys, don't bother checking the clipping on an MP3 or a WAV converted from an MP3.
YOU may get a lot of "false positives" because of the way MP3 decompression works and the inaccuracies of the process is not as precise as CD audio or WAV where 0dBFS can be achieved without clipping but an MP3 may go over.
Just another reason MP3's suck dick.
For example:
Tool Eulogy ripped from CD 16-bit 44.1KhZ
Tool Eulogy ripped from CD 16-bit 44.1KhZ converted to 192kBps MP3
YOU may get a lot of "false positives" because of the way MP3 decompression works and the inaccuracies of the process is not as precise as CD audio or WAV where 0dBFS can be achieved without clipping but an MP3 may go over.
Just another reason MP3's suck dick.
For example:
Tool Eulogy ripped from CD 16-bit 44.1KhZ
Tool Eulogy ripped from CD 16-bit 44.1KhZ converted to 192kBps MP3
Omega_Void
Your Eventual Destroyer
In the digital domain, clipping is about the shape of the waveforms, not about the level of the samples.
Clipping represents an attempt to exceed a certain level which failed. The waveform rises as intended up to a point, then fails to rise any further - result is clipping of the peaks. This is bad because it changes the shape of the wave - which of course defines its sound.
This change in sound will persist regardless of the final level. So a limiter which produces audible clipping artifacts limiting a wave to 0dBfs will also produce audible clipping artifacts when limiting to -0.3dB or -6dB or -23dB etc...
Using a plugin to detect equal-value consecutive samples is no use. Some limiters break up the flat DC peak anyway, so output no consecutive equal-value samples even when the peak is audibly distorted. And even if the peak is flat from limiting but the maximiser then adds dithering and outputs the wave with -0.1dB ceiling the flat peak will not be flat once dither is added, and the 0dBfs maximum will never be touched.
Then there's inter-sample clips. That's down to how the hardware reconstructs the wave from the digital information, which is a different issue. Read the helpfile on SSL's free X-ism plugin for a clear explanation of how that happens.
X-ism when I tried it with a bunch of limiters consistently rang the cherries for any value higher than -0.3dB. Somebody on this forum (IIRC James Murphy?) posted a long while back that a guy from a pressing plant had recommended -0.3 dB as the maximum level for a CD, so I'd been using that anyway.
The screenshots look like Adobe Audition. It detects clipping only when samples are exactly 0dB. This method is quite common and may be ok for first time recordings, but totally useless after any signal manipulation has been applied (mastering).
So the MP3 does NOT show false positives, but the WAV statistics shows "false negatives"!
Looking at the waveform, flat parts indicate clipping, as Omega-Void explained. Better clipping detectors will examine the waveform independent of the maximum level and with some tolerance regarding the flatness.
Because of bandwith limitation and other lossy operations, flat parts in WAV will be less flat in MP3. So, a better clipping detector will count LESS clipped samples in MP3 than in WAV.
Another method to detect clipping is to look at the histogram. For unclipped sound, there should be a gaussian distribution of sample values (the louder the less), especially at sample values near the maximum. When the occurance of samples increases near the peak level of a track, it indicates clipping. (You need a histogram of better resolution than Adobe Audition can give to see this!)
See ClippingAnalyzer to get some nice pictures of a clipped wave and histograms.
And here is the result for Tool Eulogy in ClippingAnalyzer:
There are some clippings indeed (ca. 500 per channel in WAV, 300 in MP3).
The higher number in MP3 by Adobe Audition is because they count every sample having 0dB, even if it is the only one at the top of a wave.
There are some clippings indeed (ca. 500 per channel in WAV, 300 in MP3).
The higher number in MP3 by Adobe Audition is because they count every sample having 0dB, even if it is the only one at the top of a wave.
tarnationsauce2
Member
- Apr 21, 2007
- 116
- 1
- 16
It all depends on what you're calling "clipping". If it's hitting 0dBFS then using most clipping statistic such as the one in AA will give false positives when using an MP3. People should be aware of it if this is the way they are going about it and how useless it is UNLESS you are mastering 16-bit song. That can lead to digital nasty overs on people's CD players.
That "Clipping Analyzer" appears to concentrate on flattened waveforms even if it doesn't hit 0dBFS. I imagine it takes the trajectory of a waveform and if it flattens to DC for a given amount of time within a given amount of time it then calls it a "clip". That's just speculation. Square waves are used in music all the time, if that's what the analyzer measures then it too will give false positives.
With that said. ALL of this over-analysis is pretty stupid. It's enabling the people who cannot hear for themselves to think they know if something is a good recording or mastering job. The bottom line and golden rule of audio engineering is if it sounds good, that's what matters. Hell a lot of people regularly use "clippers" instead of hard limiting for that extra bite of energy and harmonics. In the right application clipping sounds.... great.
I guess in a way improper / over use of the analyzer hampers the little audio engineers if they try and reach unrealistic goals of "no clipping" as the Clipping Analyzer defines it. Though it can be a good tool in the right hands, used in the right application.
That "Clipping Analyzer" appears to concentrate on flattened waveforms even if it doesn't hit 0dBFS. I imagine it takes the trajectory of a waveform and if it flattens to DC for a given amount of time within a given amount of time it then calls it a "clip". That's just speculation. Square waves are used in music all the time, if that's what the analyzer measures then it too will give false positives.
With that said. ALL of this over-analysis is pretty stupid. It's enabling the people who cannot hear for themselves to think they know if something is a good recording or mastering job. The bottom line and golden rule of audio engineering is if it sounds good, that's what matters. Hell a lot of people regularly use "clippers" instead of hard limiting for that extra bite of energy and harmonics. In the right application clipping sounds.... great.
I guess in a way improper / over use of the analyzer hampers the little audio engineers if they try and reach unrealistic goals of "no clipping" as the Clipping Analyzer defines it. Though it can be a good tool in the right hands, used in the right application.
In my opinion, intersample overs are relatively rare and generate much less audible distortion than many clippings.
Of course, its impossible to define the overall sound quality of a recording simply by examining clipping, and not every clipping can be heard.
Many people try to get more "energy" by pushing up the digital volume over the limit, so the signal gets clipped many times. At first glance, it may sound hotter. But this is pure imagination, because the only reason is the higher loudness. When played at the same loudness as the original recording, the clipped one will sound rather unnatural, distorted, mashed up. (Heavy soft limiting isn't necesseraly better.)
Therefore, I think its better when those "little audio engineers" - and especially the big ones! - are aware of clipping and its negative effects.
A tool like ClippingAnalyzer primarly helps finding critical tracks and the critical parts therein. It's just faster and easier to check a CD or a audio collection this way than to listen to it completely. And its handy to communicate the results.
When recording and mastering, there should always be enough headroom to avoid unwanted clipping, which is absolutely no problem with contemporary equipment. Therefore good volume meters, good software (which doesn't clip internally) and careful engineers are needed rather than retrospective clipping analysers.
Of course, its impossible to define the overall sound quality of a recording simply by examining clipping, and not every clipping can be heard.
Many people try to get more "energy" by pushing up the digital volume over the limit, so the signal gets clipped many times. At first glance, it may sound hotter. But this is pure imagination, because the only reason is the higher loudness. When played at the same loudness as the original recording, the clipped one will sound rather unnatural, distorted, mashed up. (Heavy soft limiting isn't necesseraly better.)
Therefore, I think its better when those "little audio engineers" - and especially the big ones! - are aware of clipping and its negative effects.
A tool like ClippingAnalyzer primarly helps finding critical tracks and the critical parts therein. It's just faster and easier to check a CD or a audio collection this way than to listen to it completely. And its handy to communicate the results.
When recording and mastering, there should always be enough headroom to avoid unwanted clipping, which is absolutely no problem with contemporary equipment. Therefore good volume meters, good software (which doesn't clip internally) and careful engineers are needed rather than retrospective clipping analysers.
