I've recently been re-encoding some of my CDs to FLAC, and I noticed a certain pattern in resulting bitrates. My only two classical discs (Star Wars 1 OST/Laputa OST) compressed down to 618 & 736kbps, whilst my usual fare of electronica was coming out at 900-1000kbps. As a test, I took a track which I've now got on three different discs and compressed it (using compression level 5):

1988 disc:
Peak Amplitude: -2.42 dB -2.1 dB
Minimum RMS Power: -91.78 dB -92.5 dB
Maximum RMS Power: -13.77 dB -12.81 dB
Average RMS Power: -26.06 dB -25.29 dB
Total RMS Power: -24.6 dB -23.79 dB
Compressed bitrate: 853kbps

1993 disc:
Peak Amplitude: 0 dB 0 dB
Minimum RMS Power: -81.43 dB -81.51 dB
Maximum RMS Power: -8.4 dB -7.73 dB
Average RMS Power: -19.74 dB -19.39 dB
Total RMS Power: -18.51 dB -18.12 dB
Compressed bitrate: 933 kbps

2003 disc:
Peak Amplitude: -.06 dB -.05 dB
Minimum RMS Power: -72.92 dB -68.22 dB
Maximum RMS Power: -4.17 dB -4.25 dB
Average RMS Power: -14.57 dB -14.72 dB
Total RMS Power: -13.43 dB -13.54 dB
Compressed bitrate: 1001 kbps

This is the same track, just mastered at different levels (the last one being painful to listen to). Will highly compressed (limited?) music always result in higher bitrates, or have I just picked a bad example?

If nothing else, I guess this shows the trend towards louder CDs

Jim

Simplifying down to my level quiet music doesn't take as many bits to encode, pretty much whatever codec you use (though I suspect for different reasons). If the original is compressed then there's not as much quiet, so FLAC doesn't get to use a low bitrate on those sections, hence a higher average.

Cheers, Paul

Just look at the RMS... the louder the music, the higher the bitrate, because there's almost nothing to throw away...

I always thought lossless encoders like FLAC don't throw away anything ...

We can talk about "throwing away" one bit when the signal is belows -6 db, for example. An uncompressed format uses 16 bits per sample to encode digital silence, while a lossless format can "throw away" all this unuseful accuracy.

Until a certain level of compression the bitrate increases, above this limit it falls.
We don't reached this break even yet, which I expect in 6 or 7 year when the loudness
increases in the same way as in the last 10 years.

The example is not bad, it is common for most albums you have older and newer releases.

Frank, that was tongue-in-cheek, right? As in, when everything is compressed to digital full scale, we will effectively have one bit of information per sample?

well, basically music consists of some valueable information and noise (that has very little information). lossy encoder can discard noise and replace it with simmilar, but different noise. lossless encoder has to compress noise losslessly, and the more loud noise is the more information it (pretends to) have and the more bits it takes to store.

also, from the point of lossless compression, this is not the same track, since these files are not bit-identical

-Eugene

The information per sample is less than 1 bit/sample, because there are remaining dependencies
between the samples, i.e. they are not independent.

...
-32768
-32768
-32768
+32767
+32767
+32767
+32767
-32768
-32768
-32768
-32768
-32768
-32768
+32767
+32767
-32768
+32767
+32767
+32767
+32767
+32767
+32767
...

Well, I guess what I'm asking for is that the record companies stop mastering their CDs at way high levels. I can dream can't I?

ask *them* then For that matter, has anyone organized some sort of campaign to ask all the majors to put the brakes on and reverse this?

Then your thoughts were wrong... Lossless encoders "throw away garbage" in a very similar way as RAR or ZIP does...

Just educate the people. Most people don't want to be stupid consumers, but for some reason loudness sells. Once enough people become aware that they are getting CD's with blown-out sound it's possible this compression-trend will fail.

At least there is more awareness than before. It was funny that the other day this guy at work was telling me about this very same issue as if it was a new issue, and I'd been aware of it for years. 1990 is about the latest date I have on a CD where it still sounds like analog music instead of a 45-minute TV commercial.