Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Warbird777]

Can "Peak Level" be a good indicator of clipping potential?

I have noticed some of my compressed (both lossy and loseless) audio files contain harsh audible clipping.
And I am wondering how good an indicator of clipping potential are the "Peak Level" values found by MC's "Analyze Audio" function?

Thinking about the Digital to Analog process, Nyquist theory, and this excellent article...

I've thought that many (most?) recent CDs are way too hot, most show very unrealistic levels.
Here's a little paper on the subject. Lends a new perspective for me.
http://www.cadenzarecording.com/papers/Digitaldistortion.pdf

...I have the following Observations, Conclusions and Questions:

Observations:
- all of the audio files that contain harsh audible clipping, do have a Peak Level of 100%
- not all of the audio files with Peak Level of 100%, contain harsh audible clipping
- most of the audio files that contain harsh audible clipping, are in Albums where most of the tracks have a Peak Level of 100%
- some albums where most of the files have a Peak Level of 100%, do not contain harsh audible clipping

Conclusions:
- it appears files with a Peak Level well under 100% are safe from clipping
- a Peak Level of 100% is not a sufficient determination of clipping potential by itself
- the best one can do is to review the Peak Levels and ultimately the subjective sound quality of every track after each rip

Questions:
- Is "Peak Level" measuring the maximum sample value, or the maximum value of the reconstructed waveform?
- what other tools could be used to scan libraries for audio files that were recorded too hot (greater than +6dB)?

This is a bit disheartening. One would almost rather have the known distortions of rumble, groove noise, and 3rd order tube clipping; than wondering where and when the digital noise demon will appear.

Michael

Addendum:

Volume Leveling (Reply Gain, Clip Protection) does not remove the harsh audible clipping.

Correct me if I am wrong, but Replay Gain only reduces level (except in rare +dB cases with low Peak Level values).

And as AlexB mentions here, my understanding is that Clip Protection...

... should automatically prevent tracks from clipping if the chosen DSP Studio settings would otherwise cause that.

If clipping exists in the source, such that the reconstructed waveform goes beyond +6dB, clipping occurs (unless the playback hardware is specifically designed to handle it.
I wonder which players/clients handle clipping better than others?

[Vincent Kars]

Peak level is a single number.
It won’t tell you how many times this level has been reached nor how long the peak lasted.
As 100 is the max, it can’t tell you how much the original signal was in excess of this level either (digital clipping)
If all settings are ‘straight’ this is simply a property of the source, produced ways to loud

[Alex B]

The audio data on Audio CDs is never technically "clipped". The data format can store only unclipped audio data. I.e. the maximum peaks can be 100% aka 0 dBFS.

The contents may be more or less compressed and level limited before they are pressed on the CD. The replay gain peak value is not a good indicator of this. A very compressed album (aka a typical victim of the so called loudness war) may have been set to 98% maximum peak value instead of 100% and that does not change the nature of the audio data.

A better indicator is the Replay Gain value.

Regarding the Clip Protection feature, it only prevents clipping that would be caused by the combined effect of MC's DSP features (which may cause the volume level to be amplified over 0 dBFS).

[hit_ny]

So could DSP effects applied on these files cause clipping ?

Easy to tell, just play any 'known' clipped file with all effects off and note if it clips.

Otherwise if the files have been modified in a wav editor is it possible to make them clip ?
 --say some parts were a tad too soft and were gained a little and in the bargain pushed the limits with louder areas.

[Warbird777]

Using Media Editor (built into MC) I have loaded one of the files which exhibits the harsh audible clipping I am addressing here.
There are several sections which display at 100% (0 dBFS) for twenty consecutive samples.
Clearly when this waveform is reconstructed by the DAC, a "clipped" waveform will result.

What is needed, is an analyzer that can report stats that would be a indicator of potential clipping in the reconstructed waveform.
Some of those stats may include:
A. when more than 10 consecutive 0 dBFS samples occur
B. when condition A. occurs more than 10 times in a single file
etc.

Without this type of tool, the recourse is to load suspect files into the editor for closer scrutiny, one by one.
Suspect files have a Peak Value of 100%, and are part of an Album where most of the tracks do too.
But that is far from sufficient...hundreds of files meet that criteria in my library.

Correct me if I am wrong but, here is my understanding of the following terms in MC:

Peak Level:
- My guess is Peak Level measures the peak sample value of the source file, not the peak reconstructed waveform value? As the latter could/should have a +dB value associated with it.
- It is the reconstructed value that is relevent here. As the article above notes, anything over +6dB will overload the DAC.

Replay Gain:
- Only reduces the average level, so that albums/playlists have the same loudness.
- Is irrelevant to clipping, unless it is positive (as in some previously mentioned rare classical music cases)

Clip Protection:
- An algorithm in MC that reduces the likehood that gains introduced by MC's DSP features will not overload the DAC.
- It cannot correct "clipping" in the source (as defined above).

PS. Does this forum support images? I would like to show the clipped waveform.

[hit_ny]

The audio data on Audio CDs is never technically "clipped". The data format can store only unclipped audio data. I.e. the maximum peaks can be 100% aka 0 dBFS.

Right so question is how is it these files are clipping, one potential suspect is a bad rip as indicated here.

[Warbird777]

...as indicated here.

This waveform on the Accurate rip site is the clipping I have multiple examples of in my Library.
Who wants to write the tool that will find these? I'll help and we can provide the code to vendors as needed.