crystal clear : was : RE: [sdiy] Simple discrete Unity-Gain Follower ?

[Magnus Danielson]

From: "Czech Martin" <Martin.Czech at Micronas.com>
Subject: crystal clear : was : RE: [sdiy] Simple discrete Unity-Gain Follower ?
Date: Tue, 6 May 2003 15:53:13 +0200

Hi Martin,

> >Cheers,
> >Magnus - thinks crystal clear for large audiences is much more of a challenge
> 
> yes, that's what I think so often: why do PA systems sound so bad, muddy?

Where should I start? There are so many ways they can go wrong and this is why
it is a challenge.

> Is it because they want as much SPL as thermally possible (no burned coils,
> but high excursion, partial oscillations, air pressure distortion, etc. etc.)
> instead of as much SPL as acoustically possible?

Try as much bang for the buck as possible, just that there ins't much buck to
start with and it is really more BANG then good sound which impresses in the
end for those who usually calls the shots. Sigh.

I've already mentioned alot of things that could go wrong, but here is a
shortlist:

You don't want passive crossover-filters. It is a total waste of money really.
They both become big and bulky, there arn't as easy to design well as you would
be inclined to think from the glossy books. They will also have considerable
powerloss so you need large amps, not only since the amp needs to feed two
speakers, but also since it needs to waste it on the filtersection. Additional
negative effects is that you get slower cone-tracking since you effectively is
driving the elements with a higher impedance due to the lossy filter. It is
just a messy legacy which is best forgotten. Passive crossovers is just where
you loose out before even turning the rig on. I don't like it in any speakers
actually. Active cross-overs and separate amplifiers is the natural choice here
to get any form of control.

Cross-over adaptation. You really must see the speakers, amplifiers and active
crossover filter as a system. You need to make sure that the cross-over between
the bands becomes as smooth as possible. There are *many* ways to go wrong
here. First of all, you want the speakers phase center to be aligned, which
can be different due to mechanical differances and also due to mode of membrane
in the crossover section. You want to have the same polarity (i.e. the
membranes gives a push outwards for positive going edge) and naturally you
want to engineer a cross-over filter such that the energy is smoothly moved
over from one register to another, such that you get a flat responce with not
too much upset in phase and impulse. You may need to apply per-register
equalization to even bumps out. You probably want a 24 db/Oct crossover, so
that you remove energy output from an element sufficiently quickly as you go
outside of the register, or else you get summing problems. You don't want very
resonant cross-over filters as a rule of thumb, but rather very smooth and
well-damped which doesn't bring supprises in the amplitude, phase-responce or
group-delay responce. Butterworth, Bessel-Thomson and Gauss filters responces
springs to mind.

Smoothness in off-axis characteristics. As a rule of thumb you want as smooth
changes in responce as you go off-axix as possible. Big lobes in different
directions make it harder (impossible?) to tame. Ported bass-cabinets have the
drawback of an additional radiator(s), and this muddles up the 3D impulse-
responce in time. Also, for PAs you are bound to not use a single cabinett on
each side, but rather alot of them to sum up. This also brings the question how
these will operate as a system when viewed on and off-axis. The interference
(essentially spatially dependent zeros) becomes much more complex in a large
scale system as a rule. For bas cabinetts you really have to spend some time
to think about the acoustical impedance. You can gain alot by mounting baffles
on the speakers. If you can corner a bas-speaker you gain alot (up to 9 dB for
tossing the same power onto it!, this is explained by radiating into pi/2
steradians instead of 4pi steradians). However, best result is acheived when
multiple speakers operate in constructive interference instead of destructive
interference such that they support each other and essentially starts to behave
as a common larger element instead of multiple elements splashing into each
others space. The effect is naturally that the near-field of the speaker
becomes bigger and when taken to a limit you put the whole audience in the
nearfield of the speaker, with the benefit of only 3 dB/doubled distance
instead of the traditional 6 dB, which is what to expect in the fair field.
For flat speakers you can reach 0 dB, but that is of lesser use in PA systems.
This effect naturally allows you to use a lower sound pressure level at 1 meter
from the speakers to make the same sound pressure level at 100 meters, which is
great for those in the audience just standing 5-10 meters away. 

Other things to think of is amplifiers. First rule of thumb is that your amp
should be able to push out more power before it clips than the speaker element
is able to handle thermically. When the amp clips it pushes very quickly much
more energy in clip-distorsion than you would expect and this naturally reaches
the ears of the listeners as well as make sure that the speaker element is
pushed to point of extintion. What many people tend to forget is that quite
alot of the power actually is converted into thermical energy in the coil. When
in the beginning of the concert the coil and speaker is cold, but the longer
the concert is going on, the hotter the speaker assembly gets and the less
efficient will the cooling of the coil becomes, since it radiates its excess
energy into its surrounding, meaning the magnet assembly, and the radiation is
relative to the temperature difference, just like voltage differance. You want
to make sure your elements doesn't heat up too quickly, something which DJs do
very efficiently by their supercompressed music (which is pure speaker torture
compared to a live rock act or symphonic music - so DJs should have a much
LOWER peak volume than a live act on the same system). Also, let's not even go
into the non-linear effects that creep up with a stressed system. So called
catastroph limiters should be applied per-band so that the amplifiers never
push more effect to the elements than they can handle. Propper catastroph
limiter design is a blessing. I've seen *really* stupid designs misstakes in
catastroph limiters - let's just say half-wave rectifiers is not only on the
chart of stupidity, but they also contribute to excess distorsion since they
rarely push alot in the end of the peaks.

As for linear filter properties and equalizations you can really fluke it. You
need to start with good material, the elements. You want to stay of the
elements whos charts would make good rolercoaster material. The box-design is
also very important. You don't want uncontrolled resonances and zeros. Reflexes
forms zeros and also tends to be spatially dependent. Ported boxes both has a
more unnaturall 12 dB/Oct roll-off than the smoother 6 dB/Oct roll-off as well
as you easilly get a resonance peak which you want to avoid or at least have
well damped and the port gives a secondary emission surface (i.e. delay and
phase-shifted signal emissioned from a different surface) which causes
spatially zeros. Spatial zeros is a drag, since no electric EQ in the world is
able to handle this 3D problem in the 1D world of the electrical signal you
feed the box with. What you want is to keep the whole area around the jw-axis
clear of poles and zeros as much as you can. If you drop a ton of equalization
onto the system they usually end up here and help to confuse the hell out of
the signal and let's not keep the ears perception out of it. What you want to
approximate is an as clear impulse response as possible. Free of resonances
near the musical range and free of zeros to swamp signal. Many people experince
systems to complex in their response that they over-compensate everything in
order to try to make it work properly. Often people experience things like
being too piercy or swallows the bas-impulses or whatever, those are resonances
and zeros at work, which may cause greater greif than a flat frequency responce
might initially explain.

Naturally you need to hunt noise and hum. There are cures for these, but they
seem all to seldom used. Balanced signals and transformators isn't there for
fun you know. The system which peaked at 139 dB(SPL) on a meter we hunted away
noise and hum so well that you had to (don't try this at home!!!) lean your
head against the grill of the speaker to hear a very faint hiss in the speaker
with housemixer, crossoverfilter and amps at full volume. If you estimate the
hiss to about 10 dB(SPL) we are talking about 130 dB of dynamic. Clean of
major distorsion factors and pretty cleaned up impulse responce and therefore
amplitude/phase responses. That is a clean system and it went very well.

The way you rigg and fly your system matters. There are loads of ways to do
things wrong. Interference, reflections, time-relations with other elements
etc. etc. is all there, toss in accoustical issues and you see the problem.
For a good system, this can be your main concern, since you already took care
of all the other problems back home.

You have to excuse me for not being an expert in all the ways you can do to
make your system worse, but giving you hints on the reasoning on how you get
your system well should give an indication what to avoid not doing so to speak.

Cheers,
Magnus