There is a lot of discussion in the hi-fi-audiophile community about the dynamic range and volume level of different recordings. I could also say this is a neverending debate. There are countless writings on the web on the issue, but it is also a significant issue in the broadcasting and recording industries. The theme begins with the simple question: „why are commercials louder than film”? In more complex form it is the infamous loudness war, which has been an important issue almost since the dawn of sound recordings, especially in the pop music industry.
Subjectively, large differences can also be discovered when listening to different recordings. The loudness level of recordings is now also addressed by industry organizations. Such is the specific recommendations and regulations of the European Broadcasting Union (EBU). (For example, recordings mastered above a specified volume level will not be broadcast.)
But back to the High-Fidelity! I have a relatively large collection of classical music CDs (original albums, not copies). I analyzed the loudness (LUFS) and loudness range (LU) parameters on a lot’s of my CDs. This had no particular purpose, simply when I have experienced what drastic differences there are between the dynamic range and loudness levels of different recordings, I became interested in the question. After collecting such data, I decided to make the collected data publicly available as well. The spreadsheets below are not complete, I am uploading new data day by day.
Methodology
A word about measurements. I first ripped the disc with the EAC program (in the case of SACDs, of course, only the CD layer and without normalization). After ripping I analyzed the full content with Goldwave 6.5 sound editing program. I ran the analysis on the whole disc content, which results in such a summary:
Spreadsheets below contain the Loudness (LUFS) and Loudness range (LU) values and the Possibly clipped values. (A good video about the LUFS value.)
The results therefore include the two stereo channels in aggregate (clipped samples are summed). The methodology has one big problem: in many cases, a disc can contain several different programs, or even several programs recorded with completely different techniques and dynamic ranges. As the table always contains data for the entire disc, for discs containing such multiple materials, it is therefore not specific for each different program. However, in case of classical music, it is very common for a record to contain only one piece of composition or possibly a few pieces of compositions made with the same technique. In the case of opera, it is even less critical problem because an act is often on two disks. In any case, I also indicated the recording dates in the table (if this is given on the album). For each album, I also included the matrix number of the disc or discs for easier identification. Regarding singers/musicians, however, I only include the most repersentative in the table. In this case it is enough for identification of the recording.
If CD made with pre-emphasis, I corrected the ripped wav manually with Goldwave’s spectrum filter before calculate LU and LUFS values. Charts contain a „pre-emphasis” column too. I write pre-emphasis information (no or yes) into this column (including pre-emphasis flag location in case of „yes”; TOC or subcode).
I am constantly uploading data to the discs, so the list will continue to expand in the future. I currently have around 1,600 CDs in my collection.
LUFS, LU, and clipped columns have colour codes. At LUFS and LU column, darker green cells indicate recordings with ever smaller dynamic ranges and increasingly louder.
You can find digital clipping on many recordings. Goldwave counts how many samples reach the 0 dB level, which can potentially cause clipping during playback. I also included this number in the table, there are some CDs that contain a lot of clipped samples. This indicates mastering/sound engineering fault. However, there are some cases where Goldwave cannot determine the number of these clipped samples because the recording is mastered so that the samples do not reach the 0 dB level. An examination of the waveform shows that there is overflow in the recording in this case also. Disks with this type of problem marked with a „yes” option because the number of samples cannot be determined here. I could determine this problem with visual examination only. And only on recordings where I suspect this problem occurs. Of course, there may be times where this has not been revealed, as it is impossible to go through all the recordings so closely.
Overloaded samples on a recording. In this case Goldwave can count these.
Clearly overloaded recording, but mastered well below 0 dB
Even if there are some overloaded patterns on a recording, that usually doesn’t cause an audible problem. One second of sound requires 44,100 samples. Even 1-2 thousands of overloaded pattern is negligible on a 74 minute long CD (unless, if they are grouped at a very short part). A well-made DAC can handle these.
Classical music list
Others list
I also uploaded analysis some of the pop/rock albums in my collection. These are shown in the table below.
Róbert Sipos (2021)