download.php?file=Sodru Dynacord DSP600 engl

download.php?file=Sodru Dynacord DSP600 engl
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 1
Dynacord DSP600 FIR Controller
With the launch of the DSP600, Dynacord is now also offering the first independent
loudspeaker controller with two inputs and six outputs to dispose over the possibility of
combined FIR and IIR filtering
Alongside the DSP244 loudspeaker controller (see test report in PRODUCTION PARTNER – Issue 10/1999), which over the last
ten or more years has proved its mettle and
technically is still to a considerable extent
in line with the state of the art, Dynacord
recently introduced two new controllers,
the DSP260 and the DSP600, each of which
boasts two inputs and six outputs, and
these are destined now in all probability to
replace the DSP244 altogether. The simpler
DSP260 is fully preset-compatible with the
DSP244 and connects to the PC via the USB
interface. A dynamic range of 111 dB (according to the data sheet), a digital AES/EBU
input, and 32-bit floating-point DSP are
among the salient features of the DSP260.
The model examined here, the DSP600, differs very little from the DSP260 on the outside, but is equipped with higher quality
AD and DA converters with a 116 dB dynamic range and employs internally two of
the latest Motorola Freescale dual core DSPs
from the Symphony DSP 567… series.
Thanks to the processing power these place
at its disposal, the DSP600 is able to offer
not only around 170 IIR bi-quad filters but
also six 512-tap FIR filters. The DSP600 can
be operated and configured via Ethernet
from a PC running the universal IRIS-Net
software. The controller can therefore be
integrated more or less at will into an
existing PC network and operated in comfort. The number of devices within such a
network is unlimited.
Outwardly, the DSP600, as is the custom
with professional devices, is somewhat
unspectacular. Design gimmicks have been
(Sub, Lo, Mid and Hi), making the channel
assignments obvious even to someone who
has never configured the device previously.
In addition, each output is equipped with a
backlit Mute button. The user interface
itself consists of 16 buttons and a three-line
As well as the general functions such as
Select, Value, Store, Recall etc., there are two
dedicated blocks of four buttons for the
processing menus and the speaker parameters. Here are to be found the groups for
parametric EQs, crossover and delays as
entirely dispensed with in the interests of
intuitive operation, which is praiseworthy.
Once the basic function of a controller has
been grasped along with the signal flow
within the device during the various
operating procedures, the operation is more
or less self-explanatory—the block diagram
printed on the top of the device being a big
help here. On the front panel of the 1U
device are LED chains for each input and
two for each output. In the case of the outputs, alongside the level meter, any gain
reduction is displayed by a second LED
chain running in the reverse direction
along with the assignment of the output
well as level and limiter in direct access for
all channels. Beneath the backlit Dynacord
emblem, there is also a USB socket, which
can be used as an alternative to the
Ethernet port.
This is to be found on the rear panel of the
device in the shape of two RJ45 sockets. Via
the integrated Ethernet switch, several
DSP600 controllers can therefore be daisychained without any external switch being
required. Above the network sockets there
are also five freely programmable control
inputs, which can be used e.g. to recall
particular setups. All the inputs of the
DSP600 (2 ¥ analogue, 1 ¥ digital AES/EBU)
H ardware
article 7-8/2010 production partner
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 2
Dynacord DSP600
are equipped with THRU outputs the signal
of which is derived from a buffer stage
taken directly from the inputs. Especially
for the digital signals, a distributor
amplifier (or the like) can therefore be
dispensed with. In the case of a device
defect, a bypass relay connects the input
signal directly to the Thru socket. The
maximum input level for the analogue
inputs is stated to be +21 dBu. Should that
prove insufficient, 6 dB of attenuation can
be introduced via software, raising the
clipping threshold to +27 dBu. The
activation of the -6 dB pad is indicated on
the rear panel by a LED placed between the
input sockets. All the outputs are designed
for a maximum output voltage of +21 dBu.
A look inside the DSP600 shows compact
digital equipment on a large circuit board
that covers around 30 per cent of the area
of the base. There are additional circuit
boards for the switching power supply—
presumably bought in as a ready-made
unit—and the operating and display unit,
which is also located directly behind the
front panel and linked to the main board
via a flat ribbon cable. The core functions of
In addition to the filter possibilities and
other processing functions, in the case of a
professional loudspeaker controller the
audio quality naturally plays an important
role—above all when we are dealing with
high-quality sound reinforcement applications in theatres or concert houses, since in
such venues no compromise is tolerated
and only the best that is technically
feasible at the time will do. The last decade
has shown that as far as digital technology
is concerned, the best that is technically
feasible is subject to rapid change. In the
field of AD and DA converters in particular
great progress has been made and an excellent level today has been achieved—one
capable of satisfying even the highest
expectations. On the other, the market for
digital audio devices is being swamped at
then dips sharply just above 22 kHz. The 48
kHz sample rate here demands a relatively
steep filter, which depending upon its
characteristic can produce this little hump,
which is of no further importance.
Considerably more important is the noise
level at the outputs of the DSP600. Fig. 2
shows the corresponding spectrum with
evenly distributed white noise without
disturbing single frequency components
with a sum level of -92.5 dBu with linear
weighting and -96 dBu A-weighted. If you
relate this to the maximum output level of
+21 dBu, you arrive at an impressive dynamic range of 117 dB. If you use the digital
inputs instead of the analogue ones, a
further 1.5 dB can be extracted. If you then
avoid operating the power amplifiers that
follow with unreasonably high amounts of
gain but instead with a sensible 26 dB, it
should prove possible to retain almost all
those 117 dB with very few lost decibels.
What must be avoided at all costs is crude
mismatching of the kind, for example,
where the power amplifier is operated at its
maximum level already at +4 dBu, as doing
so results in a massive 17 dB of the dynamic
the controller are handled by the main
circuit board. These are: the input stages
with their AD converters, the digital input
with a Burr Brown 43821 sample-rate
converter, the two DSP56724 processors and
the DA converters with their following
output stages. For the critical part of the AD
conversion an AKM5388 is used, this being
equipped with four inputs with 24-bit
converters that (according to the data
sheet) deliver a S/N ratio of 120 dB. Since
only two inputs are needed, the possibility
still exists in each case of running two AD
converters in parallel, thereby gaining
another 3 dB S/N. On the DA side, Cirrus
the present time by ever cheaper products,
the scarcely believable price points of
which are achieved through the use of cutprice components that are not always of
the finest… You could argue, of course, that
even these devices deliver excellent quality,
given that they are so very cheap; they are
nonetheless a very long way from what is
technically feasible, so for the type of
applications we are considering here, it
pays always to look more closely.
Let us begin with the frequency response in
Fig. 1, which is measured from the analogue
input to the analogue output. The curve
climbs gently by 0.5 dB until 20 kHz and
range simply being squandered.
Figs. 3 to 8 show the THD measurements
related to the level and as a distortion
spectrum as well as the measurement of
the transient intermodulation distortion.
All three measurements are doubled and
measured through everything from the
analogue input to the analogue output and
only through the DAC measured from the
digital input to the analogue output.
The latter is shown by the measured values
of the DA converter. It would make sense
here to measure also the AD converter
separately, which of course would require a
digital output that is not present here. A
2 production partner 7-8/2010 article
4398 multi-bit DACs were selected, the S/N
ratios of which are also given in the data
sheet as 120 dB S/N. On the analogue side,
Dynacord relies on solid NE5532 operational
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 3
Fig. 1: Frequency response of the DSP600 without filters.
The sampling rate is fixed at 48 kHz. Correspondingly, the frequency
range ends just below 24 kHz
Fig. 2: Noise level at the output measured through everything from
the analogue input to the analogue output. The value of -92.5 dBu
(lin) or -96 dBu (A), taken with the maximum output voltage of +21
dBu, reveals a very good dynamic range: 117 dB
Fig. 3: THD at 1 kHz related to the input level measured from the
digital input to the analogue output with +3 dB internal gain. At 17
dB below full output (at + 4 dBu output voltage) an extremely good
value of -117 dB is obtained. Even directly below the clipping
threshold with +21 dBu at the output, the value is still clearly
below-100 dB
Fig. 4: THD at 1 kHz related to the input level measured from the
analogue input to the analogue output. Ch 1 (red) was measured at
0 dB internal gain and Ch 2 (blue) at -17 dB internal gain. For a more
detailed analysis, see the text
Fig. 5: Transient intermodulation distortion measured from the igital
input to the analogue output with +3 dB internal gain. The clipping
threshold with this type of test signal is in the region of -2 dBfs and
together with the 3 dB gain then in the region of -5 dBfs
Fig. 6: Transient intermodulation distortion measured from the
analogue input to the analogue output. Ch 1 (red) was measured at
0 dB internal gain and Ch 2 (blue) at – 17 dB internal gain. For a more
detailed analysis, see the text
article 7-8/2010 production partner
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 4
Dynacord DSP600
Fig. 7: Harmonic distortion spectrum at 1 kHz measured from the
digital input to the analogue output with 0 dB internal gain and
+21 dBu output voltage
Fig. 8: Harmonic distortion spectrum at 1 kHz measured from the
digital input to the analogue output with -17 dB internal gain and
+20 dBu input voltage. At this output voltage, the THD of the DAC
lies below -115 dB, so the components visible in the spectrum can
primarily be attributed to the ADC
plausible way out, however, consists in
measuring the ADC via the DAC, provided
care is taken that the DAC is driven within
its optimal working range. As Fig. 3 shows
clearly, that is precisely 17 dB below the
maximum level. Here the DAC achieves its
outstanding best distortion reading of -115
dB (see Fig. 3). Even just under the clipping
threshold, -103 dB is still achieved. If you
look at the curves in Fig. 4, the red curve
was measured with 0 dB internal
amplification from the analogue input to
the analogue output and shows the normal
operating state. The second blue curve
shows the same measurement with -17 dB
internal gain, so that with the ADCs at their
maximum level of +21 dBu, the DAC
achieves its minimum distortion of + 4 dBu.
From this it can be seen that the ADC, even
when the input voltage is at its maximum
level, only generates -108 dB of harmonic
distortion. The harmonic distortion
spectrum (Figs. 7 and 8) and the DIM distortion (Figs. 5 and 6) were measured in the
same way. The import of all these measure-
which is why we can only offer a brief summary of them here. If you wish to delve into
the subject in greater detail, I would
recommend you to visit the IRIS-Net web
site, from which after registering you can
actually download the software itself
For the test, we used IRIS-Net Version 2.3.0
together with the DSP firmware Version
1.0.0 for the DSP600. By the time this article
appears, Version 2.4.0, in which the DSP600
is already included, should already be
available, thereby obviating the need for a
user patch. The first step when creating a
new project is to enter all the devices
(DSP600, NetMax, amplifiers, interface etc.)
that are needed for the application in
question. The devices, interfaces and a
whole series of individual elements as well
as ready-made control panels, an extensive
library of bitmaps of CD players and
various large racks along with different
loudspeaker types and text fields are
selected on the left hand side of the screen
and placed on the right of the active page.
rendered highly intuitive and easily
understood through visualisation, and
tailored precisely to the needs of the device
or project in question.
Let us turn, however, to the operation of the
DSP600. Once connected to the PC network,
the device is immediately recognized by
the software without any further settings
or modifications to the network being
necessary. If no device is connected, all the
functions are available offline. If you go
online with the software, the first thing to
decide is how the device and the software
should synchronize. Thereafter direct
access to all the functions is available. The
DSP600 appears in the interface with its
front panel on which all the indicators are
legible. If need be, the keyboard of the
device can also be shown and operated as
at the device itself e.g. for presentations or
training courses. The actual software
interface is opened by pressing the DSP
button, whereupon the DSP600 is
displayed in the form of an easily
interpreted block diagram (Fig. 10). Here the
ments could be stated succinctly: in every
discipline very good results were achieved,
and these results could only be bettered by
going to considerably greater expense.
Each project can contain up to 32 pages
(layers), which serve to provide access to
the parameters and control elements of the
individual devices as well as for
supervision. Which pages are available to
which users depends upon their access
rights. By requiring users to enter a
password when launching IRIS-Net and
specifying in the Password Database dialog
which users should be allowed access to
which pages, the various user levels are
determined. Thanks to the ability to design
and store your own control panels as well
as add bitmaps to the library, the pages can
be designed in a very individual manner,
entire signal processing modules of the
inputs and outputs and the corresponding
routing is shown. In the case of each
module, you can also see immediately
which parts (e.g. which filters) are active
and the frequency response of the filter is
even displayed in the form of a small curve.
You can also see at a glance the settings of
the controller. In each input, there is a delay
(max. 1 s), a graphic 1/3-octave EQ and a
filter bank with ten parametric EQs. The
routing follows, whereby the outputs are
assigned to the inputs. In each of the
outputs there are seven further modules
divided into the groups Array Control and
T he I R IS -Ne t soft ware
For the operation via PC of all relevant
devices, Electro-Voice and Dynacord offer
their Intelligent Remote Integrated
Supervision (or IRIS)-Net software, which
offers users control over everything from
individual devices to large networked
setups. The range of functions offered by
this software is accordingly very extensive,
4 production partner 7-8/2010 article
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 5
Fig. 9: Noise level at the outputs when the digital input is used.
Compared with fully analogue use with ADC in the input, the noise
floor drops by 1.5 dB
Fig. 10: IRIS-Net user interface with a block diagram of the controller.
You can see very clearly at a glance which functions in the modules
are being used and which filter curves and limiter characteristics
have been set
Fig. 11: Control window for an EQ bank with ten fully parametric EQs.
Missing, unfortunately, are ‘Flat’ and ‘Bypass’ switches governing all
filters and the possibility of storing individual setups for the filter
Fig. 12: Control window for a graphic EQ, whereby you can choose
between three different filter types and also set values ranging from
3 to 10 for the Q factor. Here, too, one misses Flat and Bypass
functions in the shape of simple switches governing all 31 filters
Fig. 13: Operating window for the crossover functions of all channels.
Here all the usual high- and low-pass filters up to 24 dB/8ve are
available. Thanks to the beautiful graphics, you can see the filter
frequency response directly with or without displaying the other
filters. The same goes for the phase responses
Fig. 14: Control panel for the FIR filters with freely selectable highand low-pass functions and individual loudspeaker equalization.
The latter, however, is reserved for the EV and Dynacord products
included in the FIR filter library
article 7-8/2010 production partner
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 6
Dynacord DSP600
Detail showing the digital inputs and accompanying BB SRC43821 at
the top left, one of the two Motorola DSP56724 processors at the far
bottom left and the output section with CS4398 DACs and NE5532 Ops
Internal view of the DSPs with the user interface directly behind the
front panel, the main circuit board and the switching power supply at
the back right
Speaker Processing. Array Control contains
a filter bank with five parametric EQs and a
delay. To the Speaker Processing belong an
additional filter bank with six parametric
EQs, a crossover module with 1st to 4th
order high- and low-passes, a 512-tap FIR
filter as well as a further delay (max.1 s) and
the limiters. For the speaker block, entire
settings from a library can be loaded for all
Dynacord and Electro-Voice loudspeakers.
The user has the choice here for most types
of loudspeaker between IIR and FIR
filtering. For the limiter settings there are
also libraries with the power amplifiers of
both manufacturers in which the gain and
power values are stored (see Fig. 15).
If you click on a module, its operating
window opens, which as a rule shows in
the top part a graph of the filter function or
limiter characteristics and in the lower part
the filters taken
unfortunately no possibility in the filter
blocks of saving local setups. One also
misses “flat” and “bypass” switches that
would affect all the filters in a filter bank.
These, however, are the only two areas in
which one spontaneously regrets the
absence of a feature. The display of the
filter curves is beautifully realized (see also
Figs. 11-14), which according to choice can
be displayed individually or as a cumulative
function for the channel in question. With
all Dynacord and Electro-Voice loudspeakers, it is possible in addition to load their
frequency and phase responses, which
allows you to observe directly the effects of
low-frequency range of producing the desired curve. More of this in the next
paragraph but one dealing with the subject
of FIR filters.
Moving on to the conventional parametric
EQs in the DSP600, which are realized
using IIR filters, what is striking is the wide
range over which all parameters can be set.
The centre frequency is variable between 20 Hz and 20 kHz, the gain between -18 and
+12 dB, and the Q factor between 0.4 and 40
(!). The bandwidth (speaking figuratively)
of the possible settings is illustrated in
Figures 16 and 17. Naturally there is also a
corresponding compensation for the filter
6 production partner 7-8/2010 article
Fi lte rs
In the case DSP600’s filter functions, a
distinction has to be drawn between the
FIR and the IIR filters. The IIR filters are set
in the familiar manner using parameters
like Frequency, Gain and Q Factor and
biquads (a feedback filter structure with
four coefficients). The FIR filters can currently only be configured by users as high-,
low- and band-passes. Using an IFFT, the filter coefficients for a linear-phase filter with
the desired behaviour are then calculated
from the filter curve. The critical factor here
is the number of coefficients, which determines on the one hand the latency of the
filter and also its chances above all in the
curves, which evens out the otherwise
unavoidable distortion of the curve as you
approach half the sampling rate, so that the
curves correspond exactly to the desired
ideal until just before the boundary
frequency of 24 kHz.
There was no sign of the distortions that
are the bugbear of IIR filters, when high Q
factors coincide with low frequencies, and
that are caused by rounding errors with the
coefficients and in the computing
algorithm. Given the fixed-point DSP56724
processors with two 250 MHz 56300 cores
from Motorola Freescale that are being
used here, this indicates that the filters are
working in double-precision mode in 48-bit
resolution--a solution even better than that
afforded by the usual floating-point DSPs
as far as minimizing distortion is
For the crossover functions, the DSP600
offers the choice between ‘normal’ IIR
filters of the first to fourth orders and
linear-phase FIR filters 512 taps long. With
the IIR filters, the cutoff frequency of the
high-and low-passes is adjustable between
20 Hz and 20 kHz with 6 dB/8ve and as Butterworth or Bessel filters with 12, 18 and 24
dB/8ve of roll-off. The Linkwitz-Riley filters
come with 12 and 24 dB/8ve of roll-off (see
Figs. 21 and 22). Higher order filters cannot
be selected, the assumption being here that
the use of minimal-phase filters with
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 7
Fig. 15: Control panel for the limiters with power amplifier peak
limiters and loudspeaker thermo limiters. The thermo limiters are
constituents of the loudspeaker data sets and therefore only
available for EV and Dynacord products
Fig. 16: Filter curves of the parametric EQs at 12 dB gain and a Q factor of 4 for frequencies from 20 Hz to 20 kHz. Through
compensation, the filters emulate the desired curve of an analogue
filter until just below 24 kHz (half the sampling rate). The 20 kHz
filter is necessarily asymmetric however. For the purposes of
comparison, the curves of the equivalent analogue filters are
shown below
Fig. 17: (Above) Adjustable range of the gain from -18 dB to +12 dB
taking the example of a PEQ at 1 kHz with a Q factor of 1. (Below)
Adjustable range of the Q factor from 0.4 to 40 taking the example
of a PEQ at 1 kHz with a gain of -18 dB
Fig. 18: Examples of filter curves of the graphic EQ at 12 dB gain and
a Q factor of 4 for frequencies from 20 Hz to 20 kHz. For the purposes
of comparison, the curves of the equivalent analogue filters are
shown below
Fig. 19: (Above) Filter functions of the graphic EQ for fader settings
from 500 Hz to 2 kHz on +6 dB for various Q factors between 3 and
10. (Below) Filter functions of the graphic EQ for fader settings from
100 Hz and 4 kHz on ±12dB for various Q factors between 3 and 10
Fig. 20: Effect of the filter type with the graphic EQ. At 100 Hz
constant Q, at 1 kHz balanced Q, and at 10 kHz proportional Q
article 7-8/2010 production partner
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 8
Dynacord DSP600
slopes in excess of 24 dB/8ve makes little
sense due to the pronounced phase shifts
that would result, and that it would be
better in such cases to use FIR filters directly, these being capable—at least, at
medium and high frequencies--of delivering that amount of roll-off effortlessly and
without any phase shifts. Fig. 23 shows FIR
crossover filters at 100 Hz, 1 kHz and 10 kHz
in each case 512 taps long. What can be seen
nicely here is that at 100 Hz only around 10
dB/8ve of roll-off is achieved, which at 1 kHz
however has already risen to around 250
dB/8ve, and by 10 kHz we are in the realm
of real ‘brick-wall’ filters. What this means
in concrete terms is that the FIR filters of
the DSP600 can be used for the crossover
functions from around 250 Hz upwards, as
their slopes are by then sufficiently steep.
At first sight, the curve of the high- and
low-passes at 100 Hz looks somewhat
unusual. Due to the frequency-linear
behaviour of the filters, the curves in the
logarithmic frequency display get ever
steeper as the frequency increases,
whereupon an asymmetry arises. The two
curves nonetheless add up to a constant
0 dB.
FI R /I I R fi lte rs
FIR filters for loudspeakers are nowadays
standard in the current top level controller
generation. This type of filter with a nonfeedback structure (hence the name FIR,
which stands for ‘finite impulse response’)
makes it possible, in contrast to the IIR
filters (IIR stands for ‘infinite impulse
response’) otherwise used, to determine
the amplitude and phase response
independently. If you want to create e.g. a
crossover filter with 24 dB/8ve of roll-off,
then if you are using an IIR filter, just as
with an analogue filter, this will inevitably
be accompanied by a phase shift of a 360°
in all. This so-called minimum-phase
coherence is unavoidable with this type of
filter. A FIR filter makes it possible on the
other hand to achieve the desired
amplitude response without any phase
shift, making it a linear-phase filter. The
advantage in the case of filters with only 24
dB/8ve of roll-off is limited, but with
steeper slopes of 48 or even 96 dB/8ve it is
of decisive importance, as the adverse
8 production partner 7-8/2010 article
effects of the phase shifts produced by
minimal-phase IIR filters are sufficiently
audible to preclude their use. On the other
hand, steepness in this context is what is
desired, in order to ensure that the area on
either side of the crossover frequency in
which overlapping and therefore interference occurs is as small as possible.
Even with FIR filters, of course, the news
isn’t all good. For one thing, they demand
considerably more processing power and
therefore DSP capacity than IIR filters—not
that this is all that tragic, given the
performance of modern DSPs—but they
also, depending upon their length, produce
a latency that is almost always undesirable.
This is determined by the filter length and
the sampling rate and is proportional to the
resolution of the filter in the frequency
range. That means, the more a FIR filter is
expected to deliver e.g. in terms of roll-off,
the longer the filter needs to be and therefore the greater the amount of latency.
Since the resolution is linear over the
frequency, the demands at low frequencies
are considerably greater than at high ones.
The more the coefficients of the FIR filters
therefore, the greater the overall latency of
the system, which in the case of the
DSP600, regardless of the number of signal
processing functions, would otherwise be
around 1 ms. Given the sampling rate of 48
kHz, introducing a FIR filter with 512
coefficients to a signal path within the
DSP600 adds a further 5.3 ms of latency.
The figure is obtained by dividing the
number of coefficients by twice the sampling rate: in this case, then, 512 divided by
2 ¥ 48.000 1/s. This basic calculation only
applies, however, when the FIR filter is
established as a linear-phase filter. If you
select other, non-linear phase-functions,
the latency can also be reduced to zero. In
the DSP600, however, the FIR filter
functions are always linear-phase, since for
all the other functions the IIR filters are
FIR filters offer another important advantage especially when deployed as crossover
filters for loudspeakers, since as soon as the
complex transfer functions (amplitude and
phase response or impulse response) of the
individual signal paths are known, these
can be equalized at the same time by the
FIR filters in the individual signal paths.
That means the filter not only produces a
linear-phase high- or low-pass, but also
performs the equalization of the
loudspeakers in such a way that the overall
function of the crossover filter and the
loudspeaker taken together becomes a
linear-phase filter. Eager use is made in the
DSP600 as in other devices from Dynacord
of this possibility where the loudspeakers
of its sister brand Electro-Voice are
concerned. For loudspeakers manufactured
by other companies this possibility is
unfortunately not yet available. There is
talk in-house already of opening up the
procedure to third-party users. For the time
being, though, one must be content with
the setting of conventional filter
parameters, which are then implemented
as linear-phase FIR filters.
Despite this, one is not inevitably obliged
to dispense with the possibilities of a
linear-phase loudspeaker as overall system.
As long as the possibility exists of
measuring complex frequency responses
under adequate free-field conditions, the
following procedure is to be recommended:
For the filter design, first, with the help of
various IIR filters and the level setting in
each path, the frequency responses of the
signal paths involved are equalized until
linearity is achieved at the same level for at
least one octave above the crossover
frequency and beyond. Next comes the
time alignment using the delays in the
output paths, so that the phase responses
of the signal paths concerned generally
match in the region of the crossover
frequencies. System theory now states that
into this ready-equalized system any
combination of equivalent high- and lowpass filters can be inserted provided their
sums are constant. If now you make use of
linear-phase FIR filters, the overall result
will be a linear-phase system over a wide
frequency range. The phase shift caused by
the acoustic high-pass behaviour of the
woofer, however, remains unaffected.
Even short FIR filters are capable of linearphase separation in the medium and high
frequency ranges that are markedly
superior to anything achieved using
conventional IIR crossover filters. If in
addition you want an extremely steep filter
slope e.g. in order to make the interference
band between the two paths as small as
Sodru_Dynacord_DSP600_engl_Layout 1 04.08.10 16:25 Seite 9
Fig. 21: High-pass filter functions with 6 dB per octave (dark blue)
and with 12 dB per octave with Q factor variable from 0.5 to 2.
The roll-off frequency is adjustable from 20 Hz to 20 kHz.
The corresponding low-pass filters perform satisfactorily
Detail showing the analogue inputs with the AKM 5388 ADC below in
the centre
possible, a very great deal can be achieved with somewhat longer
FIR filters, the only drawback being the somewhat longer latencies
that result.
Summ ar y
Fig. 22: Crossover functions of the 4th order with Bessel (blue),
Butterworth (red) and Linkwitz-Riley (green) characteristics as
well as their summed functions
Fig. 23: Crossover functions with FIR filters of a fixed length of
512 taps at 100 Hz, 1 kHz and 10 kHz. Due to the frequency-linear
behaviour of the filters, the curves in the logarithmic frequency
display get steadily steeper as the frequency increases. The
asymmetry of the curves is a further result of this. Due to the fixed
length of the filters of 10.66 ms, the latency with a linear-phase
setting is fixed at 5.33 ms
The DSP600 from Dynacord is a classic 2-into-6 controller offering
all the possibilities with which we were already familiar from the
Dynacord Netmax and the RCM modules in the power amplifiers.
Remote control from the PC is performed via Ethernet using the
familiar, intuitive and user-friendly IRIS-Net software. For the AD
and DA converters, Dynacord has splashed out on first-class chips
from AKM and Cirrus, as they have also for the DSPs, treating the
DSP600 to two dual-core DSP56724 processors from Freescale,
which make possible in addition to a wealth of IIR filters also 512tap FIR filters in each of the six outputs. From the metrological
evaluation, the DSP600 emerged with flying colours, achieving
such excellent results as a signal-to-noise ratio of 117 dB and THD
of -110 dB.
The DSP600 is a real joy to operate, with it’s clearly laid out and
sensibly designed interface that is sufficiently intuitive to
encourage spontaneity, and offers a wealth of possibilities for the
handling of all types of loudspeaker. One of its particular charms
lies in the possibility, through an artful combination of IIR and FIR
filters, to achieve a largely linear-phase overall system whilst
keeping latency, at only 5.3 ms, within acceptable limits. A special
advantage for users of Dynacord or EV loudspeakers are the readyprepared filter setups that are available for the DSP600 offering
inter alia linear-phase equalization of the loudspeakers based on
their measured values.
It only remains to consider the price, which at 1,499 euros is
extremely good news. All in all, the DSP600 offers possibilities and
qualities that were hitherto only to be found in considerably more
expensive devices.
◊Text and measurements: Anselm Goertz
Photos: Dieter Stork and Anselm Goertz
article 7-8/2010 production partner
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF