[LAU] OT: Dynamic range compression on classical recordings

Fons Adriaensen fons at linuxaudio.org
Wed Oct 8 21:17:01 UTC 2014

On Wed, Oct 08, 2014 at 09:15:44PM +0100, Jonathan Gazeley wrote:
> However, I'm struggling a bit with dynamics processing on the
> recording. Naturally, choral music has an extremely wide dynamic
> range. If you normalise the infrequent loud parts to 0dB, the rest
> of the recording is too quiet, and people have complained that the
> CDs are "too quiet" compared to their other CDs. I know that people
> listen to these recordings on their iPod or in the car and if the
> quiet parts are too quiet, then they simply can't hear them.
> We all know about the loudness war and I certainly don't want to
> compress the crap out of these delicate and beautiful recordings.
> But I think some subtle compression would help bring up the average
> amplitude without clipping the loud parts. I've experimented a bit
> but I'm struggling to get a "natural" sound. After compression, it
> sounds fine in the quiet parts but in the louder parts it sounds
> "lumpy" and the reverb sounds unnatural.

Almost all recordings I've made during the last five years were
of contemporary or late 20th century 'classical' music, and this
can have a very wide dynamic range. They also were made for radio
broadcasting, which meant that dynamic range had to be reduced.

There are several methods, and the best approach is to make
each of them handle a few dBs, and combine them.

* If you use Ardour to record, then editing the region gain
  curve is a very convenient way to reduce the dynamics in
  function of the music itself (it's much easier than trying
  to use automation for this). The advantage compared to using
  compression is that you can anticipate level changes - as 
  you would when adjusting the level manually while reading
  the score.  

* In many cases I've used a gentle automatic compression, with
  a ratio of 1.2 or so and a low threshold, so you get a very
  gradual gain change over a wide range of levels. This was done
  using zita-dc1, which unfortunately is not released yet.

* In most cases peaks are very short, and a peak limiter will
  remove them and allow 5 dB or more extra gain without any
  perceptible artefacts. The best one for this is zita-dpl1
  which was designed exactly for this type of use. I've used
  it on all recordings I made the last years, without exception.

Combining these, it's not at all difficult to boost your average
level by 15 dB or so (and even more if needed) in a way that will
still sound completely natural. 



A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)

More information about the Linux-audio-user mailing list