CD Waveforms Illustrated

[MDS]

There have been quite a few questions/discussions on CDs and the 'loudness wars' so I thought I would post a few examples from my own collection. A picture is worth a thousand words (or so they say) so seeing the trend visually may be worthwhile. Hopefully this will also help with understanding terms like 'average level', 'peak level', and other terms related to digital audio in general.

First: The Way It Used to Be. This waveform is Kansas - Carry On Wayward Son from their first album Leftoverture. The album was released in 1976 but the CD was probably released in the early '80s. It appears to be a straight copy from the LP mastered to CD with few changes.

Statistics from Sound Forge:

                                        	Left Channel        	Right Channel       
Cursor position (Time)                      00:02:24.118        00:02:24.118        
Sample value at cursor (dB)              -47.325                -31.071             
Minimum sample position (Time)         00:02:47.410         00:05:07.630        
Minimum sample value (dB)               -3.351                  -3.977
Maximum sample position (Time)        00:00:17.127         00:02:37.215
Maximum sample value (dB)              -3.628                  -4.210
RMS level (dB)                               -19.183                 -20.266
Average value (dB)                         -46.729                 -49.671
Zero crossings (Hz)                         1,942.52                2,004.69

As you can see, the average level is ~-20 dB and the peak level is -3.35 dB - nowhere near the max of 0 dB. The meters bounce up and down wildly and I captured the image when they were reading about -16 dB. This waveform has a large dynamic range and the 'crest factor' (difference between average and peak is rather large as well). This song sounds great and the louder you turn it up, the better it sounds.

 

[MDS]

Oasis - Hello

This is a particularly egregious example of over compression. The song is Oasis - Hello from the CD (What's The Story) Morning Glory? released in 1995.

Statistics from Sound Forge:

                                                   Left Channel        	Right Channel       
Cursor position (Time)                  	00:00:58.607       	00:00:58.607        
Sample value at cursor (dB)             	-32.247             	-23.343             
Minimum sample position (Time)         	00:01:07.231        00:00:40.908        
Minimum sample value (dB)               	-2.599              	-3.078              
Maximum sample position (Time)       	00:00:40.838        00:00:34.501        
Maximum sample value (dB)              	-2.701              	-2.967   
RMS level (dB)                          	-12.188             	-12.184             
Average value (dB)                      	-Inf.               	-Inf.               
Zero crossings (Hz)                     	1,686.66            	1,495.81

The average level is ~-12 dB which is par for the course for '90s and later CDs, but the peak level is - 3dB and the waveform spends it's entire time at roughly that level. No ups and downs, no contrast - just always the same level at all times. The louder you turn it up, the worse it sounds.

 

[MDS]

Ozzy Osbourne - Crazy Train

Ozzy Osbourne - Crazy Train from the CD The Ozzman Cometh, released in 1997.

Statistics from Sound Forge:

                                        	Left Channel        	Right Channel       
Cursor position (Time)                  	00:01:59.461        00:01:59.461        
Sample value at cursor (dB)             	-11.116             	-25.170             
Minimum sample position (Time)         	00:04:14.217        00:00:24.587        
Minimum sample value (dB)               	-0.009              	-0.009              
Maximum sample position (Time)       	00:00:19.541        00:01:00.367        
Maximum sample value (dB)              	-0.009              	-0.009              
RMS level (dB)                          	-12.474             	-12.284             
Average value (dB)                      	-85.923             	-Inf.               
Zero crossings (Hz)                     	1,681.52            	1,571.56

The average level is ~-12 dB but the peak level is 0 dB as is the case on most newer CDs. However, this song actually has some reasonable crest factor and dynamic range. The waveform isn't as compressed as the Oasis example above, but wait and see what a few years can do to things...

 

[MDS]

Ozzy Osbourne - Crazy Train

Ozzy Osbourne - Crazy Train but this time from The Essential Ozzy Osbourne, released in 2003.

Statistics from Sound Forge:

                                        	Left Channel        	Right Channel       
Cursor position (Time)                  	00:02:06.409        00:02:06.409        
Sample value at cursor (dB)             	-3.779              	-7.873              
Minimum sample position (Time)         	00:04:36.239        00:04:35.491        
Minimum sample value (dB)               	-0.067              	-0.010              
Maximum sample position (Time)        	00:04:34.723        00:00:16.597        
Maximum sample value (dB)              	-0.010              	-0.010              
RMS level (dB)                          	-10.261             	-10.280             
Average value (dB)                      	-60.016             	-60.908             
Zero crossings (Hz)                     	1,531.21            	1,501.76

What a difference a few short years makes. This version of the same song is louder than the previous with an average level of ~-10 dB but also is much more compressed. The level meters in Sound Forge rarely dip below -3 dB and you can see from the screen capture that the majority of the time they are right around the peak of 0 dB.

 

[MACCA350]

Great illustration MDS.
The Essential Ozzy Osbourne, released in 2003. This is just ridiculous completely butchered, I personally don't listen to Ozzy's stuff but its painful to see this happening to so many releases It makes me wonder if this loss of dynamic range is partially why many people say some of their LP's sound better than the CD release(snap, crackle and pop aside )

cheers

 

[JKL1960]

Great illustration MDS.
The Essential Ozzy Osbourne, released in 2003. This is just ridiculous completely butchered, I personally don't listen to Ozzy's stuff but its painful to see this happening to so many releases It makes me wonder if this loss of dynamic range is partially why many people say some of their LP's sound better than the CD release(snap, crackle and pop aside )

cheers

I'm sure there is some truth to that. It is so bizzarre. I remember that getting extra dynamic range out of vinyl was something everyone was after. Now with a medium that actually has more than enough dynamic range they are compressing to almost less than vinyl had.

It makes me ill.

Part of the problem is that you need better gear to expose it. On a boom box or computer speakers they sound fine. On better stuff still not everyone notices but point it out, explain it and play some samples and I haven't seen anyone yet not notice. With no frame of reference they don't know.

I've been called crazy for suggesting that MP3s aren't good enough for serious listening. I don't think that serious listening is as popular as some of us wish. (I mean sit in your sweet spot and turn it up to real volumes, serious listening.)

 

[no. 5]

it does'nt even have to be that good of equpment to notice the defrence, but when will it end?
when every sound on a CD is within 3dB of 0dB?
you'r right JKL, people tend to be shocked with MP3's, that you can't compress music and have it sound the same.

great thread MDS.

 

[mtrycrafts]

It makes me wonder if this loss of dynamic range is partially why many people say some of their LP's sound better than the CD release(snap, crackle and pop aside )

cheers

I don't see any other explanation that it could be. With this much compression and practically every note has the same level, I just don't know what anyone can enjoy from a recording like this?

 

[mtrycrafts]

This is a particularly egregious example of over compression. The song is Oasis - Hello from the CD (What's The Story) Morning Glory? released in 1995.

Statistics from Sound Forge:

                                                   Left Channel        	Right Channel       
Cursor position (Time)                  	00:00:58.607       	00:00:58.607        
Sample value at cursor (dB)             	-32.247             	-23.343             
Minimum sample position (Time)         	00:01:07.231        00:00:40.908        
Minimum sample value (dB)               	-2.599              	-3.078              
Maximum sample position (Time)       	00:00:40.838        00:00:34.501        
Maximum sample value (dB)              	-2.701              	-2.967   
RMS level (dB)                          	-12.188             	-12.184             
Average value (dB)                      	-Inf.               	-Inf.               
Zero crossings (Hz)                     	1,686.66            	1,495.81

The average level is ~-12 dB which is par for the course for '90s and later CDs, but the peak level is - 3dB and the waveform spends it's entire time at roughly that level. No ups and downs, no contrast - just always the same level at all times. The louder you turn it up, the worse it sounds.

Great pictures, yes.
This one seems like it has the least dynamic range in it. Hardly any white dips; even less than that 2003 release.

 

[mtrycrafts]

it does'nt even have to be that good of equpment to notice the defrence, but when will it end?
when every sound on a CD is within 3dB of 0dB?
you'r right JKL, people tend to be shocked with MP3's, that you can't compress music and have it sound the same.

great thread MDS.

No. I think it will end when all the notes are at 0dB with no dynamics in it
Just look at that 2nd chart. It is hard to find any dynamics there even when it is that far from 0dB. Is that music? Or just a shouting match?

 

[mtrycrafts]

Again, great pictures. Do you have any classical recordings to illustrate as well?

 

[MDS]

Sorry, no classical music in my collection. It's mostly '70s and '80s rock/pop/disco/electronica with a small amount of '90s pop and rock. I just picked a few random songs that I thought would help illustrate the point.

You'd really be shocked if I posted waveforms of things like Air Supply. Super low average level but huge dynamic range and sounds really good if you like that kind of light pop music.

One thing I think is really funny is I have a CD of Barenaked Ladies and most of the songs on the disc have an average level of -8 dB. That's louder than my Who CDs!

 

[no. 5]

No. I think it will end when all the notes are at 0dB with no dynamics in it

lol .
I'll bet you'r right though.

 

[no. 5]

Sorry, no classical music in my collection.

you could pick up a whole wack of lightly used classical CD's from Amazon.com for around $50, just so you could post what those waveforms look like.

 

[MDS]

Air Supply - All Out Of Love

These waveforms are Air Supply - All Out Of Love from the Greatest Hits CD released in 1984.

The first is the original untouched version:
Average Level: Left = -19.404, Right = -19.896
Min Sample Value: Left = -.758, Right = -2.271
Max Sample Value: Left = -.206, Right = -2.039

It peaks right around 0 dB (actually -.2 dB) but there is only ONE such peak and the next highest peak is -2.039. The majority of the time the meters stay below -6 dB. This is conservative mastering no doubt.

To make it louder you could normalize it to 0 dB but there is only .2 dB between the current highest peak and 0 dB, so normalizing will just raise the volume of the entire file by .2 dB and the waveform looks practically identical to the original.

To really make it loud requires compression and/or limiting and that comes next...

 

[MDS]

Air Supply - compression

The compressor takes a few parameters:
Threshold - the level at which the compressor kicks in to do its thing
Attack - How long the signal must rise above the threshold before the compressor will start compressing it.
Release - How long after the compressor kicks in the signal must remain above the threshold before it stops compressing the signal.
Ratio - The ratio of input to output.

This waveform I think would be considered 'mildly compressed'. The parameters are threshold = -15 dB, attack = 1 ms, release = 500 ms, and ratio = 3:1. That basically means that anytime the signal rises above -15 dB and stays there for at least 1 ms, the compressor will increase every 1 dB change above -15 dB by 3 dB and will continue to do so as long as the input signal remains above -15 dB for 500 ms.

So if in 1 ms the input rises to -14 dB, the compressor will change it to -11 dB. Got it? Confusing I know...

After compression the new statistics are:
Average Level: Left = -14.043, Right = -14.412 (~-20 dB before)
Min Sample Value: Left = -.428, Right = -.326
Max Sample Value: Left = -0.00, Right = -0.00

Now the music is on average +6 dB louder and the peaks are now at 0 dB but still there are only a few peaks at 0 dB (unlike some of the prior waveforms).

 

[MDS]

Air Supply - limited

At the opposite end of mild compression is extreme limiting. The Sound Forge preset for this case is called 'maximize volume'. It is a -12 dB limiter with auto-gain compensation (attempts to keep the input and output levels constant). Without auto-gain compensation it would look like the Oasis - Hello waveform where the entire waveform falls between -12 dB on both the positive and negative portions of the wave.

Average Level: Left = -11.442, Right = -11.783
[right in line with many new CDs]
Min and Max sample values now -0.00 in both channels.

It is now twice as loud as the original and clips repeatedly.

 

[MACCA350]

Again, great pictures. Do you have any classical recordings to illustrate as well?

Done

P.I. Tchaikovsky, Symphony No. 4, OVERTURE "1812"
By Point Classics Digital Recording, CD released 1994

.......................................................Left Channel.............Right Channel
Cursor position (Time)..........................00:00:00.000............00:00:00.000
Sample value at cursor (dB)...................-Inf........................-Inf.
Minimum sample position (Time)..............00:12:28.581............00:12:29.014
Minimum sample value (dB)....................-0.452.....................-0.804
Maximum sample position (Time).............00:03:13.515............00:14:35.533
Maximum sample value (dB)...................-0.114......................-0.310
RMS level (dB)....................................-23.654....................-24.415
Average value (dB)..............................-56.283....................-64.668
Zero crossings (Hz)..............................1,006.38...................810.14

cheers

 

[MDS]

Thanks for the classical waveform. It's actually what I would have expected - low average level but peaking near the max. The difference in philosophy for mastering classical vs pop/rock is interesting. I wish the pop/rock would go back to looking more like that classical waveform.

 

[mtrycrafts]

Done
P.I. Tchaikovsky, Symphony No. 4, OVERTURE "1812"
By Point Classics Digital Recording, CD released 1994
cheers

Yes, great shots, thanks. Just what I was hoping for