calculate entire RMS of an audiofile ,are audio statistics accurate?

[JohnVo]

hi

about are the audtio statistics of an entire audio file accurate ?

reading a forum about it , i have read

it's impractical to calculate the true RMS of the entire file, so most algorithms do some time-windowing and take an average-of-RMS values (or maybe an RMS-of-RMS values if that makes sense mathematically). So... two different applications may report two slightly different RMS values.

&

and, the average may  also be an average-of-averages over the length of the file.

is it true?

thanks

[SteveG_AudioMasters_ · Accepted solution]

Giovannivolontè  wrote

is it true?

Yes.

Now what?

[JohnVo]

it's impractical to calculate the true RMS of the entire file

&

Yes.

Now what?

hi

so audtio statistics is almost unless

i use to find out the peak amplitude & the minimum rsm ampliture

these value are misleading aren't they?

for example i working on this audio , i watch the value with the red arrows

thanks

[SteveG_AudioMasters_]

No they aren't. The errors are so small as to be audibly imperceptible. Just because they are there, it doesn't in the slightest mean that they are significant. We are talking about small fractions of a dB here, and your ears can't realistically distinguish a 1dB level shift, never mind anything smaller.

On top of that, this only becomes significant if you scan a large section of a file - and information gleaned from this about averages is pretty useless anyway, never mind how accurate it is or isn't. The only things you need accuracy about, Audition calculates accurately anyway. So if you want to normalize waveform so that the highest peak hits a specific value, then this is very accurately determined, and it works absolutely fine.

What do you need this information for, anyway?

[JohnVo]

hi

i use to fix manually clipping , noise (speech records ) and son

i use to add markers and work on them

may i know Steve if you use and when?

thanks SteveG

[SteveG_AudioMasters_]

Giovannivolontè  wrote

i use to fix manually clipping , noise (speech records ) and son

i use to add markers and work on them

may i know Steve if you use and when?

The RMS statistics won't help you fix anything to do with clipping. If you want to fix that, you use Diagnostics and the de-clipper setting to determine and mark them.

Personally I don't have to do this for myself, because I don't ever record anything that hot - quite deliberately. About -12dB is about the highest level I ever encounter, so clipping isn't an issue. The only times I ever encounter clipped audio is if somebody sends it to me to see if I can fix it.

[JohnVo]

Diagnostics and the de-clipper setting to determine and mark them

hi

never worked for me

look

here you can see there is not clipped audio

thanks

or i can't understand Diagnostics or they don't work or it's buggy on my system

[SteveG_AudioMasters_]

Giovannivolontè  wrote

or i can't understand Diagnostics or they don't work or it's buggy on my system

Try resetting the tolerance to zero...

[JohnVo]

hi

it works! i set tolerance to zero !

the default is 1% but set to 0 work perfectly!

wondering about how should i set the declicker

thanks StegeG!!!

[SteveG_AudioMasters_]

Giovannivolontè  wrote

hi

it works! i set tolerance to zero !

the default is 1% but set to 0 work perfectly!

Why are you so surprised? All default settings in Audition are only starting points - you have to adapt them to suit your particular circumstances.

wondering about how should i set the declicker

Which declicker? There are three, although two of them are very similar. The general principle though is to use your ears. The click/pop eliminator is the most effective, but you have to set it up with care. Let it loop around a section after you've determined the threshold levels, and only advance the controls as far as you need to make them inaudible.

[JohnVo]

Why are you so surprised? All default settings in Audition are only starting points - you have to adapt them to suit your particular circumstances.

from adobe

Tolerance

Specifies the amplitude variation in clipped regions. A value of 0% detects clipping only in perfectly horizontal lines at maximum amplitude; 1% detects clipping beginning at 1% below maximum amplitude, and so on. (A value of 1% detects most clipping.)

if at tollerance 1% , for audition clipping is -3.97db L -6,12dB i really i don't understand it

Which declicker?

in the Diagnostics , there is delete silence, declipper,mark audio and declicker

i'm talking about it

There are three, although two of them are very similar.

1)automatic click remover

2) click/pop elimination

3) is the declicker in the Diagnostics ??

thanks SteveG

[SteveG_AudioMasters_]

Giovannivolontè  wrote

if at tollerance 1% , for audition clipping is -3.97db L -6,12dB i really i don't understand it

Technically that's a bug, as in your example, the duration is zero and you can't have a clip lasting zero time! I suspect this has a lot to do with the rapid rate of change at the start of your selection. You will also note that in Amplitude Statistics it says Possibly Clipped Samples - strictly you need to consider the waveform either side of any given sample in order to estimate whether it's part of a clipped sequence. If you make a selection within a file, the rules say that you can only consider the selected part, but that you can't exclude the possibility that in a rapidly changing selection at the start, that the first sample could be a part of a clip in the section immediately prior to your selection - so basically it's playing safe (although I don't think that really, it should do this). But as it's not the easiest thing to detect, there is allowance for errors, and they aren't all obvious. For instance, it's quite possible to create sequences of samples that will overload the output of your sound device without technically clipping at all - by anything up to about 8.5dB. And you need, especially if you're going to normalize a file, to take account of this. Now, these sample sequences are actually legal - even though the clip detector would say they were clipped. So, this is not an exact science, and you have to use a bit of common sense with it.

No, I am not expecting you to take my word for the overload situation. Years ago I created a document (using Audition 1) showing how this can happen, and since this is all based on the Laws of Physics, nothing has changed. You may have a look at the document and its waveforms here: Dropbox - Audition_1.0_overshoot_doc.pdf

[JohnVo]

hi

SteveG

i have loaded other audio clips and check for clipping tollerance %1

i know there are not clipping sample ,but audition found many and many clipping samples

would like to know what the develop team can say about it

thanks

[ryclark]

All you have to do to make sure that there are no clipped samples is to Normalise the audio files to a slightly lower level. -1dB should be Ok but you can go down to -3dB if you want to be absolutely sure. After all there is no real need normally to push your audio within a fraction of a dB of clipping. But at the same time you should be able to see most audio peaks that get anywhere near clipping without having to measure anything I would have thought.