I went through some of my CD's and analyzed the levels so that you could see how they have changed since CD's were commercially produced.

Although a few were made in 1983, production was not on a large scale until 1984.  

If you have followed us here, you know that we want to maintain a True Peak level of -1 dB in order to prevent any problems at all with any playback hardware and software, and to assure that any conversion to a lossy format will not result in any clipped samples.  

The LKFS number is an average Loudness number that helps us to quickly see where the level of a track is.  We can look at great recorded music through the years and see what is good in terms of dynamic range and loudness levels.  The LKFS number would to a certain extent be controlled by the dynamic range of the song.  Orchestral music with very wide dynamic range would be closer to -20  dB, and compressed pop music could be about -12 to -16dB.  I like to use -16 dB as a standard because virtually all music will fit into that range and it is the highest level that iTunes and most broadcasters will allow before applying their own compressors.  So a level of -16 will assure that your track will be heard exactly as intended.

You can see from the chart that we passed the safe level in the mid '90's.  From 1997 until now, the levels have remained squashed and clipped.  So for 20 years, we have talked about the evils of the "Loudness Wars", but have done nothing to change our behavior.  The biggest names in the mastering world are the worst offenders.  

Thousands of examples could be listed, but I think the chart makes it clear enough.  This should help to dispel the notion that only experts can make good music.  Experts can ruin good music.  And high resolution recordings at 192 khz or 384khz and at 64 bit is not the issue.  Good music is the result of following a few rules and learning from the past.

​

Let's take a look at some real life examples of some CD's from the early days.  The first one up is a CD from one of the sweetest sounding guitarists of the 20th century, Tony Mottola.  This was basically a direct level transfer from the vinyl master and you will see that the track we are looking at (Close To You) has an LKFS of -23.8  dB, a true peak of -9.4  dB, and a dynamic range (LU) of 5.1  dB.  If we want to raise that level on a CD master, we can raise it 8.4  dB without doing any damage to the original recording.  That would bring the new LKFS number to -15.4  dB, the true peak level to -1  dB, and the dynamic range of course is unchanged at 5.1  dB.  The pictures of the overall waveform appear below.

​

I have mentioned that at the dawn of CD production, the albums were transferred from the master tapes to 1630 Umatic tapes at the same level as they were for vinyl and then cassettes.  There were slight variations in different countries, but in this country, the standard was from -18 dBFS to -20 dBFS.

The goal of the standard was to prevent the highest peaks of the particular recording from going beyond 0 dB.  The 1630 Sony machines which were used as the transfer medium between Mastering  and CD production had a readout which told you if the peak level had been exceeded during the recording or playback of the tape.

 If 3 consecutive 0 dB levels were detected, the recording was rejected because of the certainty of a clipped or distorted waveform.  I had 3 of these machines in my room at the CD plant and spent years doing this until the advent of CDR masters, and then internet supplied DDP masters made the very large and complex 1630 Umatics obsolete.

So the most important order of business in the production of CD's, as it was with the production of vinyl records, was to prevent the highest peaks from being distorted in the transfer to the final consumer product.  (Cassettes were exactly the same)

If you have followed this blog and watched the videos, you have a good idea of why we have to keep the actual samples to -1 dB or less to prevent the resultant waveform from going above 0 dB.  There of course are waveforms that could go above 1 dB over the sample level on either side of them, but the vast majority will fit comfortably in this parameter.  If a recording has waveforms that exceed that level, the waveforms can be individually lowered, run through a limiter, or the entire recording lowered in level to compensate for that peak.

I have done all of the above at some point and each method has pros and cons.  With modern software, it is simple to do one or all of the above fixes.

What we have with modern pop music is a situation where the music has such a small dynamic range that there is rarely just one peak at  the highest level.   The song is compressed and limited to the point that every beat is at the maximum level.  So instead of dealing with one or two peaks, you need to use software to deal with thousands of peaks, or lower the level of the entire recording away from the 0 dB mark.

The previous blog showed how 100% of the masters from the Grammy's 2016 CD were incorrect in the simple parameters that we have described here.

​Obviously these things are not obvious.  We will examine some more soon.