And that ^^^, mike_321, was the reason for my "radio competitors" comment. Too subtle, maybe.

Other Mike mentions Refugee. I'm sure I'm not the only one who can't resist cranking that one up when it comes on the radio! Pull up a few of your other favorite radio songs and you'll see that none of them are volume-maximized. Try "Radar Love", or "More than a Feeling". Crank 'em up and enjoy.

Just a quick FYI, the mp3 scoping is never correct....all mp's I've ever scoped out had "possibly clipped samples" as well as DC's. Something in the conversion causes this. Try it with one of your own songs and you'll see what I mean. You may not get  the possibly clipped samples at -1dB or the DC offsets, but if you were to master out at -0.1 dB (anything -0.3 or louder will do it) then convert to mp3, you'd see the DC's as well as the PCS's.

Another thing to keep in mind, if you're using Adobe Audition to scan in, it has a sensitive level of threshold for scoping that stuff out, which I like..but sometimes it is literally giving you false alarms. For example, sometimes it will show PCS's yet your LED meter will never flash in the red.

My point is...don't put too much stock into scoping mp3's. If you had the same song in wav format, the numbers would change pretty drastically. I'll give you an example in case anyone thinks I'm out of my tree. :)

Philip posted a song we did on the song forum called "So Long". Here are the stats for both the wave and the mp3. The mp3 was converted directly from this wave file using Wave Lab. Notice how the statistics have changed? The wave is 16/44. The mp3 I've chose is 320 kb, constant bit rate, slow encoding.

Wave File:

View in browser: https://dl.dropbox.com/u/4909348/SoLongWaveStats.jpg

Mp3 File:

View in browser: https://dl.dropbox.com/u/4909348/SoLongMp3Stats.JPG

See? The results are quite different. Scanning both the wave and the mp3 in Wavelab gives me even different results. So never put too much stock into mp3's. They will always give you a false representation/misrepresentation every time. :)

-Danny

I'd also like to add to my post above....a lower bit rate mp3 is even going to give you more false representation. You see how mine changed using a 320...imagine how it changes if you are reviewing something of lesser quality. Also, variable bit rate and constant bit rate can play a role here too. You just can't trust mp3 encoding to give you the right numbers because as you can see, it's a bit off.

-Danny

It looks like it ended up with 14 something-like-intersample peaks and the noise floor came up a bit.


best regards,
mike

Thank you batsbrew and Danny-- interesting stuff!!

Here's an old thread on the topic of headroom for MP3s, with some pictures: http://forum.cakewalk.com/tm.aspx?m=1689770

I simply use a different method.... I know about the levels issue and db's and such..... but my goal is not to make redbook speced waves.... I just want them to sound good as mp3 and waves.

I keep everything int he tracks and busses and master... for the most part...out of the red. An occasional hit into the red is not a cause to freak.

  I don't overly compress and when I export I normally will normalize the file to less than 100% (I think the example below was taken to 100%) ..... this gives me waves that look like this:  



This song.... if you wish to hear the results.... is on page 2 on my sound click site.

I'm happy with the results but... again I'm not planning to burn them to a CD.... I'm plugging songs to publishers and such things where the song is very likely to be recorded again by someone else..... or possibly used in a "less than demanding" listening environment. 

Just my way of working..... and my 2 cents on this topic.

The final level of the master is only as important as the amount of compression used to achieve it. A -0.1db level which has been compressed lightly will still sound quieter than achieving that level through a 10:1 compression ratio and low threshold. Whether it sounds better or worse is another matter.
 
If your song has all the sonic quality it needs, and you want to retain air and dynamics, then push it up to around -1db with light compression and the quality will shine through

BOTTOM LINE IS...
 
most people listen to mp3's these days...
and the quality of your encoder will have a say-so on the final rip quality..
 
if you master too high, you'll get clipping on your mp3's.
just turn it down.
 
me, i master to -0.2, and depending on the quality of the rip (i use wavelab typically, but sometimes CDEX) it can still clip.
 
old CD players are the same, that cannot handle peaks near -0.1 even.....

There are a few loudness "standards" which most people tend to ignore.   -20dB RMS for classical, -14 to -12 dB for dynamic pop and jazz.  In reality, for popular music, the audio usually gets squashed to -6 to -4 dB RMS.  This is the dynamic range of the music, and keeping in mind that 1dB is considered the smallest noticeable volume change, most popular music today just sounds like a loud wall of sound.  
 
For peaks, when I started out the standard was -0.3dB.  This was only to prevent inter-sample peaks during the DA stage, it has nothing to do with the loudness level of the recording and it is irrellevant if the peak is set to -.2, -.3, or -1,  they all will have the same perceived average loudness.  With modern peak limiters that have oversampling, this is not an issue and the music can be mastered to 0.0dB without a problem, but to avoid error messages for Red Book mastering you need to drop the output to -0.2dB.