[LAU] Mackie Onyx like channel strips?

David Kastrup dak at gnu.org
Sun Nov 19 15:30:58 UTC 2017


Robert Edge <thumbknucklerocks-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org>
writes:

>> The canonical filter realisations as found in textbooks (e.g.
>> biquad) are usually the worst way to actually implement them
>> if things like numerical stability and parameter smoothing
>> are taken into account.
>>
> People say this a lot.  Also warping at the extreme end of the
> frequency response is supposed to be a real problem.

Here is a graph:

-------------- next part --------------
A non-text attachment was scrubbed...
Name: tp.pdf
Type: application/pdf
Size: 11003 bytes
Desc: not available
URL: <http://lists.linuxaudio.org/pipermail/linux-audio-user/attachments/20171119/444545e3/attachment.pdf>
-------------- next part --------------

Is this a problem?  It may be if you are trying to estimate filter
parameters for a given signal filtered in the analog domain.

> I dunno.  I incorporate the regular ass Robert Bristow Johnson
> cookbook filters in to stuff all the time, and have been doing so for
> years.

It's also a no-go if you use Bessel filters for maximizing phase
linearity in the passband.  Then it's really a waste of resources to use
the bilinear transform and get a comparatively large IIR filter that
does _not_ asymptotically have linear phase in the passband.

> I can't imagine why a sane human being in 2017 would go out of their
> way to emulate a Mackie.

Because the circuit diagrams are openly available and that particular EQ
design matches a particular purpose?

I play the accordion.  The usual approach to standard bass registrations
(of which there typically are three to seven) is to pick the register
sounding best.

Now my favorite accordion has 168 separate bass registrations.  That
means that you don't just try out every one for a given passage and pick
the best.  It means that you make a plan that reduces your options to a
few ones and _then_ pick out the best among those.

This kind of mechanism has not caught on.  People want a few good enough
and diverse choices and then work with them.

On the treble side, a big accordion typically has 11 registrations.  I
have, strictly speaking, 7.  Only 3 or 4 make real musical sense.  Who
cares as long as you don't tire of hearing them as fast as all the
ordinary 11 registers?  A violin has 1 register.

Stock settings matching given analog designs have the advantage that

a) you may already know what to expect
b) someone has invested a lot of time for making a coherent result
   appealing to experts and listeners
c) you can reasonably easily match a particular sound you have been able
   to achieve with the analog console, perhaps even autogenerate
   automation from the analog mix

> I mean, they make stuff that is fine if that's what happens to be
> there, but nothing I'd go looking for.

The main idea was that you don't need to "go looking for it" since they
make their circuit diagrams available, and the modular and opamp-based
design means that converting them to formulas and then to digital is
easy.

Basically low-hanging fruit of good quality.  If it doesn't appeal to
people not in possession of the respective mixer, I'm actually glad.  It
will make it easier to get Mackie to be ok with it if it only means
added value if you actually own one of their devices.

-- 
David Kastrup


More information about the Linux-audio-user mailing list