Dynamic Range Analysis 1970-present

Dynamic Range Analysis 1970-present

UPDATE April 2017: I’ve processed the data collected since these graphs were originally made in 2013 and updated them below. I’ve added a paragraph with any new findings/insight.

Using a bash script I developed it was possible to collect all the data that had been submitted to the Dynamic Range Database. Following some minor cleansing (to remove records with missing fields, although there is much, much more that could be done to tidy up band and album names) it was possible to do some basic analysis.

There are three main dynamic range values for each submission; album average, the maximum track dynamic range, and the minimum track dynamic range. Along with this there is the year and whether the source is ‘lossy’ or ‘lossless’. A crude attempt to identify albums from vinyl sources was made; relying of the fact that many submitters put “vinyl” in the album title field.

The collected values were averaged for each year and plotted over time. The following graph shows the averaged album, maximum and minimum track DR values from 1970 to present day for the complete sample of 48,947 submissions. The error bars indicate ±σ (standard deviation) of the album values.
DRcomplete
The graph shows that from 1970 onwards the average DRs increased to a peak in 1982, followed by slow but steady decrease up to 1990. The region marked A shows a dramatic decrease in DR; an almost straight-line increase up to 1997. After 1997 the rate of decrease tailed-off and could arguably have stopped around 2005. The area marked B shows possible signs of increasing DR in the years following to the present day.

Also of note is the standard deviation has also increased since 1990. This suggests that albums are ‘polarized’; being either crushed to death or nicely dynamic, with the resulting average somewhere in between. Indeed, there are signs the average of the minimum DR value has been increasing since 2005.

It was also possible to plot just the albums from ‘lossless’ sources such as CDs etc.
DRlossless
This shows similar trends to the overall plot, but with a more pronounced increase in album and minimum DRs since 2005.

Looking at ‘lossy’ sources, such as mp3s, gives a differing trend.
DRlossy
There is a clear trend to decreased DR values that is continuing to the present day.

It was possible to identify a small number of vinyl rips:
DRvinyl
This potentially shows modern vinyls are becoming more dynamic at best, or remaining roughly constant at worst. Although, given the potential variations in ripping techniques this could be considered dubious. In addition, there is some discussion over the suitability of DR values for vinyls.

If I had the time I’d look at trends within individual or groups of genres.

Updated graphs

The new graphs below consist of >100k data points. Processed the same as above. If anything, it could be said that the DR of vinyl has improved in the last 3 years. But the DR of lossy sources is getting worse. Lossless sources look about the same, but overall there is still a reduction in DR…