Lectrosonics DMTH4, Справочное руководство

Страница: 7 / 70

Digital Telephone Hybrid
Rio Rancho, NM
7
In a simple sound system arrangement, the local micro-
phone can be muted when nobody is talking in the local
room. A simple gated mixer can provide this function.
With no open microphones locally, there is obviously
no return echo signal. This requires that a threshold
level be set high enough to keep the microphone from
being opened by background noise, but low enough
to allow it to open when someone speaks. When the
local microphone is open, a return echo path is created,
which is when a DSP echo canceller is needed. Given
the wide variety of human voices and the dynamics of
noise in a meeting room, a gated mixer is often not the
best choice.
Using a dedicated DSP echo canceller on each input of
the local mixer (referred to as “distributed echo cancel-
lation”) is an expensive but effective approach to reduc-
ing the return echo. The process requires the algorithm
to “converge,” which is to identify the far-end audio
and subtract it from the signal sent to the far-end. This
requires at least a brief moment when there is very little
local sound or noise, with significant far-end audio pres-
ent in the room. If nobody moves and there are no gain
changes made to local microphones and loudspeakers,
it is possible (in theory) to effectively remove return
echo, but this is not a very realistic situation.
The theory behind distributed echo cancelling is that
once the DSP has converged, it can continue to sub-
tract far-end audio even when the local microphone is
open and far-end audio is present at the same time. If
there are any changes in gain, noise or acoustics in the
local space and equipment, the DSP must re-converge,
which requires another brief moment with little or no
local noise or sound, and significant far-end audio pres-
ent.
A gated automatic mixer does not change the gain
when the microphone is open, it just turns the channel
off and on abruptly. This helps with distributed echo
cancelling since the microphone is completely muted
when not in use, but it is very “choppy” sounding in the
local sound reinforcement system.
A gain proportional automatic mixer applies the most
gain to the most active microphone with smooth,
continuous changes. This makes it extremely effec-
tive for local sound reinforcement, but the continuous
gain changes make it difficult for the echo canceller to
remain converged and effectively reduce the echoes at
the far end.
The DMTH4 in conjuction with a DM Series processor
offers a unique approach to the problems with simulta-
neous teleconferencing and sound reinforcement. The
patented adaptive gain proportional mixing algorithm
works in conjunction with a centralized echo canceller
to address a variety of issues. The automatic mixer pro-
vides seamless allocation of gain to local microphones
through a mix-minus matrix to reduce background noise
and decouple loudspeaker and microphones, while a
very fast converging DSP echo canceller operates on
the composite transmitted signal being sent to the far
end. This combination of processes is possible only
with the latest DSP technology.
The auto mixing algorithm adapts to changes in back-
ground noise continuously, and unlike a gated mixer
there are no threshold levels to adjust. A sum of all
channels is the reference signal, each channel level
is compared to this reference and the individual chan-
nel gain is adjusted to apply NOM attenuation. Gain
is adjusted continuously to eliminate audible artifacts
that gating and abrupt level changes can cause. As the
common mode noise in the room changes, all channels
are affected equally. The end result is seamless, adap-
tive auto mixing that requires no calibration or threshold
adjustments.
Each individual output of the matrix operates as a sepa-
rate NOM bus, so a particular input can be assigned to
multiple outputs with mix parameters adjusted differently
for each output. In other words, gain and mix mode are
configured independently for each matrix crosspoint, re-
sulting in great flexibility. Four mix modes are supported:
Auto, Direct, Override and Background.
The echo canceller converges continuously when the
level of the far side signal exceeds a minimum level,
and the ratio of the far side signal to local room sound
exceeds a minimum ratio. This dynamic control pre-
vents divergence during periods of silence from the
far side room or in “doubletalk” situations. The con-
vergence takes place very quickly to keep up with the
changes made by the automatic mixing algorithm and
other changes that occur in the room. Setup is greatly
simplified and any adjustments, such as level changes
made with a remote control system, are accommodated
automatically.
The convergence speed is adjustable in the control
panel GUI to fine tune it to a particular situation. Faster
convergence times can track changes in the room al-
most instantaneously, but the depth of echo cancellation
will be reduced. Slower convergence times take a bit
longer to fully converge, but produce greater echo can-
cellation. The ERLE value achieved by the echo cancel-
ler is displayed on the GUI and the effects of altering the
convergence rate will be immediately visible and audible.
An important final note on the DMTH4 is the fact that the
echo canceller will never “diverge” (lose convergence).
This unique algorithm will also converge on a continuous
sine wave, which is especially important when DTMF
tones are present in the room. Since the echo canceller
will never diverge, there is no need for a “panic button”
(as is used in other designs) to generate a noise burst to
help the echo canceller re-converge.