[sdiy] group delay (of filters) and listening

[Magnus Danielson]

From: "Czech Martin" <Martin.Czech at Micronas.com>
Subject: RE: [sdiy] group delay (of filters) and listening
Date: Tue, 5 Oct 2004 17:42:28 +0200
Message-ID: <D9D56E8FA1A73542BE9A5EC7E35D37FF01C4F1FF at EXCHANGE2.Micronas.com>

Martin,

> I used to make experiments with treble and bass speakers, that
> were mounted in certain distance. This causes nonlinear phase.
> Up to 70cm difference or so it was difficult to hear any 
> diffreence or problem.

Yes, you need to time-align your elements physically or electronically before
you start working on the filters. Also make sure that all elements have the
correct polarity (positive going voltage have cone going out, JBL have their
pole-pieces inverted!). In the end you trim all parameters to have a nice
transient and nice summing in the amplitude responce. When the ear is satisfied
with the amount of improvements one can get lay down the tools and go.

> Contrary to the headphone experiment the room reflexions come into play.
> Only few sources show the influence of an actuall room on speaker
> performance, though it is easy to measure.

The room responce and how the speaker couple to the room is quite interesting.
Either you view the speaker as a separate entity or you find ways to connect it
acoustically to the room. Put your speakers in the corner and you have a 9 dB
gain compared to it's free-space responce. In the corner it has support from
two walls and floor. For higher frequencies this support becomes less apparent
than for the base, since the speaker actually has a direction of its sound at
higher frequencies. What is important here is that you have a clear simple
spatial impulseresponce (approximating an expanding sphere - but only
approximating). Also, the directivity of a speaker shall be natural, that is
its directivity shall _smoothly_ become higher as the frequency rises. This is
what best works with rooms.

> Of course, the flat characteristics you can see in the anechoic room
> will be totally distorted, peaks of 20dB are nothinbg unusuall.
> This would not sell.

Right. You don't need anechoic rooms by the way, you can do pseudo-anechoic
measurements easilly. On and off axis naturally. Off-axis is really not only to
the side put also up down, effectively at different points around the speaker.

> There are only few web sites which show actuall room performance.

Go figure. The world is progressing very slowly.

> Considering this, the question of phase linearity of speakers gets
> even more questionable.

You *CAN* do it, but it takes a whole different approach than people are used
to work with speakers. There's really no big magic, it's just follow the trail
and see where it takes you.

In PA-speaker world, you don't have one speaker but many. The trick is to make
these work together to form one larger speaker. You definitly want to avoid the
destructive interference also known as spatial zeros. The best thing you can do
is to design the speakers so that instead of having walls to support each small
speaker the waves of the individual speakers support each other. Now you have
constructive interference. This is for instance what happends in the V-DOSC
system. In the near-field (a place where speaker-analysis have stayed off
traditionally, since the assumptions they usually make does NOT make sense
there) you then have non-spherical wave expansion and then the 6 dB per doubled
distance is also not in action. The V-DOSC system has 3 dB per doubled
distance in the near field, it's just that for higher frequencies the whole
audience is IN the near field, which is much more practical.

Now, with constructive interference you use up the emitted power much more
sensible, you don't have to play as loud in SPL sense in the front for a good
level at the back, for the benefits of everyone except some hearing specialists
that don't get to treat additional patients.

Oh, the room acoustics also create a diversity of resonances, these are the
spatial poles of a room. The reflections causes spatial zeroes since they will
cancel out (if only slightly) the signal received directly.

Acoustics is the real hell. You have three-dimensional wavefront expansion
with different waveform intensities and different wavefront impulse responces
occuring in time. Additional elements and ports creates additional sources and
all forms of reflections creates secondary emissions. Just picturing this
three-dimensional wavefront as it occurs in time is a minor hell. With modern
computers we can do it. But that doesn't help since we need to gather the
finalizing impulse responce of respectively ear and then that for all the
places we can be at in this room and you got computational disaster comming
your way.

The rules of thumb is to keep simple waveforms, one large source, smooth
transition from non-directional to directional and even spreading of energy.

Cheers,
Magnus