Loudspeaker system
NL
Latest change 2023-03-07
Extra: In the sept / oct 2019 edition of
Elektor-Labs an article written by me about the Bass unit of
this system was published. The Elektor editors have asked me to
write about the whole system. You wil find it here:
http://www.breem.nl/lsp/LoudspeakerSystem.htm
(English only)
In
brief:
An unconventional loudspeaker system
with as most important properties:
- 3-way, Bass, Mid, Tweeter. Each loudspeaker has its
own power amplifier.
- Active filtering with compensation of loudspeaker /
box properties.
- Reproduces frequencies well below 20 Hz.
- Concentric around a vertical axis, radiating over 360
degrees in the horizontal plane.
- Optimized time / impulse response.
- Subtractive crossover of second order for bass and
tweeter, first order for mid range. Vector sum = 1.
- Bass and mid speakers are electro dynamic, the tweeter
is a monopole electrostatic speaker.
- The relative volumes of the loudspeakers can be
remotely adjusted by software.
- Class D power amplifiers located in a free space on
the back of the bass unit.
- The control amplifier from which the speakers are
driven and the cabling supports symmetrical transmission of
the analogue audio signal, symmetric RS422 digital
communication and a 5 Volt line to switch the power on / off.
An important source of inspiration was Hans van Maanen, have a
look at the "diamant" system on his website: https://www.temporalcoherence.nl
The setup
All parts except the cone and the tweeter box have a
pentagonal ground plane.
The lower part has the bass speaker radiating upwards, above
that the somewhat diamond shaped mid range compartment,
also radiating upwards, and on top the tweeter radiating
downwards. Between mid and tweeter sits a double cone
approximately on ear-height when the listener is sitting.
The photo on the right gives the status of November 2017,
still in ground color.
Important
considerations:
Why 3-way?
Not any single loudspeaker is able to reproduce the full
audio spectrum of over 3 decades.
One could think of a full-range electrostatic speaker,
ESL's are known for their low coloring in the mid range
and their very transparent high. However they also have
some drawbacks. For the lower frequencies a large surface
is required, and for real bass power a large excursion of
the membrane. That is difficult. Additionally, they are
almost exclusively constructed as dipoles, so reproducing
very low frequencies is nearly impossible because of the
shorting effect. For the highest frequencies the
directivity is often a problem. In my concentric setup
that is no problem at all.
With a two-way system the bass speaker must
produce well into the mid range, and / or the tweeter
must do well in the higher mid range. Such a combination
is difficult, because the bass speaker must be quite large
for enough power at extreme low frequencies, and for the
tweeter I had already decided for an ESL because of the
transparency in the higher frequencies.
So almost automatically I ended up with a 3-way system.
Why has each loudspeaker its own
power amplifier?
This approach has so many advantages that it is
astonishing that it is not done much more often, even in
the very high priced segment where people easily pay
thousands of euros/dollars/pounds for a speaker set or
an amplifier.
- Every loudspeaker is damped well over the full
frequency range by the low output impedance of the
amplifier. In the conventional approach with passive
filters the damping is impeded because the filter often
represents a high-Q resonator and other reactive
properties.
- With active filters one can realize much nicer filters
with cheap components.
- Passive filters are often designed with the assumption
that the loudspeaker impedance is nicely 8 or 4 Ohms.
Far besides the truth; a 4 Ohm loudspeaker can
easily reach 40 Ohm at resonance, and then your filter
does something different.
- Equalizing the sensitivities of the loudspeakers
can easily be done with a potmeter or so. In the passive
situation one needs power resistors which have an effect
on the filter characteristics, or certain combinations
of loudspeakers do not fit because the sensitivities
differ to much or the impedances do not match. 4-Ohm and
8-Ohm speakers often do not mix very well.
- The collection of power amplifiers can easily be placed nearby or inside the loudspeaker cabinet,
resulting in very short cables.
- Distortion produced by one amp only appears in
one speaker. E.g. high frequency distortion in the bass
channel does not end up in the tweeter. Sure in the bass
speaker, but that hopefully does not reproduce it very
well.
- Per channel less power is required in comparison
to a full-range amplifier.
- Experimenting with the filter properties is more
attractive, because the components cost a few cents,
while the passive components may cost tens of
euros/dollars/pounds.
- In particular it is possible to make filters
such that the summed output is perfectly flat in
amplitude and phase. Passive filters
mostly produce phase errors resulting in
inadequate addition of the sound outputs for
certain frequencies, and degraded impulse response.
- Yes, one needs multiple output stages. That
may look costly. But the actual component cost of an
output stage is some 10's of euros/dollars/pounds.
Why to below 20 Hz?
I do have some recordings, CD as well vinyl, containing
such low frequencies. So I want to hear them. I also use
this system combined with video projection, and quite some
movies / TV programs have very low frequency audio
content.
But more important, and this is valid for any system, if
you want to have a very good impulse response you need a
much larger frequency range than what is actually used. In
the SACD system the frequency response goes far beyond the
for humans audible frequencies, resulting in a better
impulse response in the audible range. The same argument
holds for the low frequency side of the spectrum.
Why radiating all
around?
That is a matter of taste. The -not-so-strong- argument is
that the same spectrum is radiated in all directions, as
do many acoustical music instruments.
Another is the so called "sweet spot", the location
where you have the best audio quality, is very large here.
It does not matter much where you set yourself to listen,
it's ok everywhere.
Sure there are people who prefer a very direct sound and
accept a small sweet spot. They could have profit from
some of the arguments here.
Why concentric?
Well, then radiating around works better. More important
is that the times of travel from each loudspeaker to the
listener can be made equal, especially for mid and
tweeter. Not only in the horizontal plane, but also in the
vertical plane. With conventional setups the distances are
mostly different for different listening positions,
especially height.
Why optimizing the
time response?
The experience of people who did this indicates that every
improvement of the time response leads to better audible
detail.
Why a crossover of
2nd order?
One would like to use higher order filters to prevent
signals to reach speakers for which they are not meant.
However, with filters of order over 2 the time response
cannot be made perfect anymore. (I have this thesis from a
math expert but without prove)
Why Butterworth?
I think it gives a good compromise between on one side to
much overshoot and ripple in the pass band
like Chebychev and on the other side not such a
smooth transition as with Bessel and Gauss.
Why has the mid
channel a 1st order response?
That is a consequence of the subtractive filter. If you do
Bass an Tweeter with a second order filters and the mid
range is derived by subtraction you will end up with a 1st
order characteristic. I do not know a proof of this but
the simulations show it without doubt.
Why not use digital filters? They can realize high-order characteristics without phase shift?
Yes, they can. But a filter without phase shift must have
a symmetrical impulse response, and that is only possible with a non
causal filter, where output is generated before input arrives. In the
digital domain that can easily be done by delaying everything somewhat.
However in the real world every thing is causal, nothing happens before
it's cause. So you may expect strange audible effects from non-causal
filters.
The second objection is that high-order filters with or without
phase-error always have strong pre- and post ringing. We dont want that
either.
(These are also issues in the discussion CD vs. SACD)
Why not for the
mid range also an electrostatic speaker?
This has been considered. But it appears that the mid
range speaker has to do quite some low frequencies, and
that is not possible with an ESL of this size.
Why an mono pole
ESL as tweeter?
I want it only to radiate downwards, and not upwards, that
does not combine with the principle of radiating around
with the double cone.
So it sits in a closed box with damping material. There is
a risk that the membrane will resonate with strong basses
and produce sparks. (as per January 2018 no problem found)
Secondly, ESL's are dust magnets and I don't want to
worsen that by letting rubbish fall on them.
Why remotely
adjust the output levels of the loudspeaker by
software?
I do not want to do that with potmeters which might
be differently set left/right. I found an outstanding chip
for that in the PGA2311 from Texas Instruments and that
requires a microprocessor for the control. Besides that a
micro is also required for the class D power amplifiers.
Why class D power
stages?
Well, the argument is not that strong, other than I'd like
to try that too. From a viewpoint of audio quality there
is no preference for class D or the usual class AB, both
can result in an amplifier with little distortion and
sufficient power.
THE big advantage of class D is it's very high efficiency
so everything can be quite small, no large cooling bodies
etc. But for hifi in the home this rarely a problem, no
need for extremely small.
Jan 2018: The class D amplifiers are operational now.
Based on the TPA3255 from Texas Instruments. Per
system 1 chip in PBTL for the bass and 1 chip in 2 x
BTL for mid range and tweeter
The listening experience indicates less distortion than
before with the STK086 modules, especially heard with choral
music.
Why symmetric
transmission?
Symmetric transmission results in a much better
suppression of disturbing signals like hum from mains
power or ground loops.
Remind that the control amplifier and the
loudspeaker-amplifiers are likely to be powered from
different outlets and that there may be long cables in
between.
Why remotely
switching the power for the final amplifiers?
I do not want them to be continuously switched on and I
also do not want to go there to switch them on every time,
both is waste of energy.
Power switching based on a minimal audio signal is no
option too, something must be continuously powered there,
and I have bad experience with the sets of Philips
Motional Feedback units I have in use, they switch off
when you are playing a bit soft. (I bridged the relays in
these units)
How do you
reproduce frequencies well below 20 Hz?
That can only be done with a
completely closed box. Any other housing would be much to large.
Partialy "open" cabinets like Bass Reflex or other "vented"
housings fall with 24dB/octave below the resonance frequency, a
4th order high pass filter. That cannot be corrected with a
forward filter, all alone because of the phase shift.
Besides that the behavior of loudspeaker is a closed box can be calculated quite easily, it
has a second order slope which lends itself for corrected with a forward
filter. Yes, there will be large membrane excursions, but you will not
lose the pressure through the opening of such a "vented" cabinet.