On Wed, Jul 23, 2014 at 07:52:52AM +0200, Gabriel Nordeborn wrote:
> So, my question is: How on earth do I do this?! Are there FLOSS tools
It's not basic and not easy.
A room doesn't have 'a' frequency response, it has a zillion FRs.
You'll get a different one for each position and orientation of the
speaker and for each position of the mic. And if the mic isn't an
omni (yours isn't), also for each orientation of the mic.
And each of these, unless doctored to look presentable, will look
quite chaotic and completely different from all the others. In
other words it's near impossible to interpret a raw FR of a room
unless it has been processed to remove irrelevant detail and keep
only the features that matter. And then you need to understand
the nature of that processing and its consequences. And even then
just a FR isn't very interesting. How a room sounds depends very
much on how it behaves in the time domain. A raw impulse response
will tell you more than raw frequency response. But the relevant
information is a combination of time and frequency domain data.
It's quite easy to 'flatten' a room FR in the mid in high frequency
regions, a combination of damping and scattering (diffusing) will
do it. But unless your acoustic treatment includes the low frequency
end as well you could end up with a room that is very boomy and still
has a lot of resonances (room modes) in the lower frequencies. And
removing those is not easy and not cheap. So in the end it's better
to not overdo the mid and high frequency treatment unless you have
the means to treat the low end as well.
This is why near-field monitoring is popular in budget limited
conditions, by placing yourself closer to the speakers you reduce
the influence of the room on what you hear - it eanbles your brain
to ignore the room to some extent, provided it's not too small.
If you want to correct your system after the room is finished, the
DRC software is the way to go.
Make sure to read the excellent documentation, all of it, so you
have and idea of what you're doing. It will explain some of the
hairy things I hinted at earlier. DRC will produce IR filters
which you can use with jconvolver (or with any other convolution
processor). The result will be much better than when using e.g.
a graphic EQ, because DRC does things that a simple equaliser
If you want to get a quick idea of the overall FR you can use Japa.
Send pink noise (from Japa) to a speaker (one at a time), connect
the mic to one of Japa's inputs, select 'Prop' response (using the
control below the display) and 'Slow' or 'Noise' speed. This is
the type of measusement that would be used to adjust a 1/3 octave
graphic equaliser. Note that the target is normally not a flat
response but one that drops off gently above 2 kHz.
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)
Linux-audio-user mailing list