Different average RMS value between recent version of Adobe Audition and Adobe Audition 3.0

[hyunsoo Jeong]

Why is the calculation method for the average RMS value different between the most recent version of Adobe Audition that I am currently using and the previous Adobe Audition 3.0? The results calculated from the same file in each program turn out differently. ​Please explain how the calculation methods differ in each version and which one provides the accurate value.

[michelle_4392]

Hello!

The difference in average RMS value calculations between Adobe Audition versions is due to changes in the algorithms used. The newer version likely uses a different method, such as time-windowing, which can result in slightly different values. Unfortunately, Adobe doesn't disclose the exact formulas, so it's hard to say which one is more accurate.

[hyunsoo Jeong]

Of course, fully understand, but the results is not slightly different. It's huge. for example, the value measured in 3.0 showed -26dBFS, but same file can be measured in recent Audition as -51dBFS. So it's too big. Confused. which one is correct? Even if the method of the algorithm cannot be opened, I think there should be at least a minimum explanation of how it has changed, or which is more accurate, or that a certain part has changed.

[SteveG_AudioMasters_]

I'm afraid that Michelle is essentially incorrect in her assertions. The algorithm for calculating RMS values hasn't changed at all - it's defined as taking the root of the mean value squared over a given windowing period. It's done this way - universally - to give always positive absolute results, which are displayed as a level below 0dB, hence the minus sign.

I checked the same file out on Audition 3 and the current version. The results were identical except for the average RMS value, which seemed to be rather far astray, by about 19dB. Which got me wondering. I looked at the file I was using, and it has several seconds of silence at the end of it. Removing that silence and rescanning gave a much more reasonable total RMS figure. I tried exactly the same thing on Audition 3 and got a much smaller jump in value.

What does this mean? (pun intended). What it implies is that Audition 3 only gives an accurate result when there's no large silences in a track. The silences - quite correctly - alter the average signal value across the file quite significantly, and Audition 3 doesn't pick this up correctly. This has been corrected at some point, and it means that you get more accurate values from any of the more recent releases.

[hyunsoo Jeong]

Hello Guys.

when I checked this issue by myself, I made one script file by Matlab and I can find something different between different versions. As someone mentioned windowing, I applied that method and second, I just applied just full data of file iteslf without any windowing. If I applied the 10ms of window to calculate the Average RMS, I can find same results of Adobe Audition 3.0, but if I applied the square and mean on the data, I can get very similar value of the latest version of Audtion. So, It seems that they have modified the algorithm to always work with a single file and scan specific portions of that file to calculate the average RMS.​ While using a 10ms windowing method is very useful in real-time applications, Audition does not provide the feature for real-time average RMS calculation. Instead, it only allows for reading the entire data set or scanning specific sections to calculate the average RMS for those sections. I believe this is the reason for their algorithm change. This is just my opinion, and I’m not sure if it is correct or not...