dev builds
themes manual
device status forums
mailing lists
IRC bugs
dev guide

Rockbox mail archive

Subject: Re: output level equalisation
From: Chris Holt (
Date: 2003-11-29

On Sat, 29 Nov 2003 11:59:12 +1300, ten Velde family wrote:
> I've seen a number of posts here describing settings for a sound that's
> close to the sound produced by a CD player.
> I'm trying to use the Jukebox through a PA and found that it's rather
> thin when all settings are on 0 or off, so decided to create a series
> of test tones of various frequencies and put them on the Jukebox as mp3
> files, so I can adjust the settings to give me a reasonable level
> frequency response using the level meter on the PA mixer.
> I created a series of files of sine waves of one frequency each in wav
> format, then bulk converted them using CDex without normalisation.
> To check whether these tones were indeed all of the same level, I used
> the analising function of MP3Gain. I thought that would be the best
> way since I run all my files through this program to ensure a
> reasonably "evenness" of levels.
> The results of this are rather alarming: at 50Hz the level detected is
> 72.9 (below that it's even less, but I'm not too worried about those),
> the levels then rise to 98.8 at 3840Hz and drop back down again to 62.0
> at 15kHz. The mp3s are encoded at 192kbps to ensure a reasonably flat
> response at the higher frequencies.
> When I load these three files into Goldwave, the levels are: 50Hz
> maximum 0.2500 RMS 0.1767 3840Hz maximum 0.2510 RMS 0.1767 15kHz
> maximum 0.2722 RMS 0.1767
> which would indicate no significant variation in levels.
> The question is: should I use the files as they are and assume that
> MP3Gain would distort the real levels, or should I normalise using
> MP3Gain to get a true evenness of levels? Also, since I have processed
> all the music files through MP3Gain, and if that has "adjusted" the
> frequency response of these files, would I be better off to process my
> test tones the same way?

MP3Gain does not adjust the frequency response in any way, (it simply
adjusts the global gain variable on each frame) but it's analysis is based
on perceived loudness, so it will not treat all frequencies the same (when
it comes to analysis, not processing)

> Only one can be right, surely - and a level variation of some 40dB over
> the audio spectrum is rather too large to ignore.

Either could be considered "right" unless you discount weighted dB
measuments as a "real" measurement. The human ear is not a linear

> John ten Velde


Page was last modified "Jan 10 2012" The Rockbox Crew