13 May
2010
13 May
'10
10 a.m.
On Wed, May 12, 2010 at 10:14:41AM -1000, Joel Roth wrote:
On Wed, May 12, 2010 at 09:44:58PM +0200, Fons
Adriaensen wrote:
This is interesting. Never thought of building a DIY mastering system out of LADSPA
plugins, but I guess it couldn't hurt.
I'm not sure how to match up the settings you have here with the ASCII diagram above.
i.e. where does the spatializer and limiter fit in? And why is the parametric in front of
the three high/mid/low pass filters, and what purpose does it serve? Also, I'm
assuming that the SC4 has the same settings for each band, at least for starters.
Also, I'm not sure how to route this in stereo. I thought SC4 was mono anyway. I'd
guess that the compressors for each band should be stereo compressors, ganged togehter in
some way (RMS?), like perhaps the Calf Compressor is.
I feel limited here by ASCII art, and also by how to do the routing in for example Ardour,
which might be really complex, or maybe in something like Ingen.
The GLAME settings make sense to me: all 2-pole filters, with the lowpass rolling off at
106, the bandpass at 520 and 800, and the highpass at 1030.
How would these LADSPA filters compare to, for example, postfish? I'm really starting
to wish I'd used postfish instead of JAMIN now.
-ken
On Wed, May 12, 2010 at 11:30:11AM -0700, Ken
Restivo wrote:
Some time ago, Julien Claassen forwarded to me the
results of his exchange with Fons regarding reasonable defaults for
a three-band mastering network for Nama.
As Fons writes here, the sum of the bandsplitting filters
is unlikely to add up to the original signal, however
we did adopt what we believe to be a reasonable compromise,
with just a couple dB error.
I'm appending the default settings for the LADSPA effects
we are using. Each of the low/mid/high branches gets a compressor
and a spatialiser.
+-- low --+
mix ----> eq ---+-- mid --+-- boost ---> output
+-- high -+
You could create a similar network in Ardour or any other
sufficiently flexible software.
It might be worth experimenting with some A/B comparisons.
(Might be nice to have a program that can randomly cross
fade a couple signals with a beep indication. The listener does a
blind comparison, inputting his opinion at the keyboard.
Afterwards the program correlates the inputs with the
signals.)
eq: Parametric1 1 0 0 40 0.125 0 0 200 0.125 0 0 600 0.125 0 0 3300 0.125 0
low_pass: lowpass_iir 106 2
mid_pass: bandpass_iir 520 800 2
high_pass: highpass_iir 1030 2
compressor: sc4 0 3 16 0 1 3.25 0
spatialiser: matrixSpatialiser 0
limiter: tap_limiter 0 0
So, are the artifacts caused by the compressors
or the EQ?
The ones I referred to are caused by the FFT-based EQ.
I didn't use the EQ on JAMIN at all, just the
multiband
compressors and the final limiter.
I don't know if Jamin uses the FFT processing to implement
the bandsplitting for the multiband compressor.
If it does then the artefacts of this type of processing
will show up in the output.
If it doesn't, then Jamin needs some type of bandsplitting
fillters that add up to exactly the input if compressors
are inactive. I doubt very much if these are implemented
in Jamin.
Ciao,