[LAD] So what do you think sucks about Linux audio ?

Fons Adriaensen fons at linuxaudio.org
Sat Feb 9 12:08:35 UTC 2013

On Sat, Feb 09, 2013 at 11:16:48AM +0000, John Rigg wrote:

> By "dumbed down" I mean restricted in a way which may result in
> inexperienced users making fewer mistakes, but will also inconvenience
> more advanced users. An example of this would be a high mid EQ that won't
> sweep above 8kHz. What if I need to EQ 12kHz? There's some excuse for this
> kind of thing on analogue hardware, as component cost has to be kept down,
> but in a plugin it's totally unnecessary. 

I've seen some EQ plugins that actually become unstable near FS/2. And even
some where the author didn't bother to limit the F range in that case - move
the slider too far with some others at the right value and watch the smoke
from your tweeters. 

Lots of plugins are written by people who apparently never have used real
audio engineering gear ever, and don't bother to test their code or even
understand what exactly it will do.

Some ten years ago someone wrote an 'RMS' routine as part of a compressor
plugin. It does a sliding window calculation over 256 samples. You can
find verbatim copies of it in at least 5 other dynamics plugins and also
in 'VU' meters (which are not even supposed to use RMS). While it's not
entirely wrong, the result is that the compressor (or whatever) will be
completely insensitive to level variations at some frequencies. For example,
at 48 or 44.1 kHz, any modulation by around 90 Hz (or integer multiples)
will be completely ignored. Five millisecs attack time ? Forget it - you get
some attack time that depends on input content in a completely haphazard way.
And of course the same lenght is used at whatever sample rate. The result is
then presented as an 'RMS' compressor... 

> Another pet peeve is lack of a text entry field on controls, as it makes
> it difficult to set a parameter to an exact amount. Even worse are detented
> controls. What if I need an intermediate setting? The reason for detents on
> analogue hardware is for repeatability of settings, but it's totally
> redundant in software, unless the developer has neglected to provide
> text entry!

I generally dislike text input on audio engineering tools. The controls
should have sufficient resolution. More often than not the author just
accepts the display resolution (e.g. standard toolkit sliders), and if
these don't have fixed sizes you don't even have repeatable settings.
For EQ frequencies I usually have steps 1/3 or 1/6 octave depending on 
filter type, with finer steps availiable using a modifier key.


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-dev mailing list