Duelund articles
Printed From: SpeakerBuilder
Category: Brancheområde
Forum Name: Gæsteoptræden
Forum Discription: Her inviteres nogle af branchen bedste folk til at lave indlæg.
URL: http://forum.speakerbuilder.dk/forum_posts.asp?TID=44
Printed Date: 20 Apr 2024 at 00:45 Software Version: Web Wiz Forums 9.56a - http://www.webwizforums.com
Topic: Duelund articles
Posted By: Duelund Coherent Aud
Subject: Duelund articles
Date Posted: 05 Feb 2009 at 14:48
Hej
Så dette forum og tænkte jeg ville publicere de af Steen Duelunds artikler, jeg har liggende, såfremt det har interesse.
Mvh.
Frederik Carøe
|
Replies:
Posted By: Duelund Coherent Aud
Date Posted: 07 Feb 2009 at 00:19
Første post er en hurtig film om Steens vidunderlige kabelviklemaskine.
Den blev lavet af materialer, som Steen havde liggende. Inspirationen var vist sket på et teknisk museum. Den larmer helt utrolig og holder 34 cm kabel i minuttet.
http://www.steenduelund.dk/download/25032007.mp4 - http://www.steenduelund.dk/download/25032007.mp4
|
Posted By: Duelund Coherent Aud
Date Posted: 07 Feb 2009 at 00:19
Den drives i øvrigt af en gammel boremaskine.
|
Posted By: Duelund Coherent Aud
Date Posted: 07 Feb 2009 at 00:23
Artikel angående Duelunds indledende tanker om kabel.
Duelund Cables ( evt. download http://www.steenduelund.dk/download/kabler%20intro.pdf - )
The transmission of signals through cables is disturbed in different ways. These disturbances are heavily related to modern time and it’s widely use of plastics and related topics and further of course to the material and form of the wire itself.
Cables were at first of no interest of mine, as my primary hobby is loudspeakers. But two friends Niels and Thorsten had made some convinced me to look deeper into that area. A new search, which led to so many interesting results, was started.
In the sea of audio worms offered on the market, one might loose the orientation and go for the price, assuming there is a connection - the more expensive the better sound. I'm really not to judge between them, as they all are wrapped up in plastics, and I don't like the sound of that material at all. In my ears it adds a fatness and muddiness to the bass and makes treble thin, by apparently increasing its lower and upper end, not to be seen by measurements. As I hear it, it must have a kind of memory - much like the effect known from capacitors, and thereby works as an active element in the dimension of time, where it should be silent. This effect explains why it sounds of more than is measured.
What then? - At first we must make a choice - single core or multicore. Multi-core has, in my ears, a common sound of their own, which I don't find with single core. Single core doesn't sound totally right either, if it is a round wire. This I used in the beginning - 1 mm round pure silver in different insulation, where two materials fell out to be good as isulation: cotton and silk. But there should be others for you to find - flax for instance, which I haven't been able to find in pure form as thread.
All cable experiments were done with interconnect cables, as they transport all frequencies and differences are heard with ease. The form and size of the wire itself influences the way sound is transmitted through it. If you like the sound bounded off in a pleasant manner, you should use round wire around 0.7 mm in diameter. If you choose them thicker, you will spoil the precision in the treble. If you like it very, very precise, you should use thin foil, but then the lower part of the sound register becomes far to thin without weight. Here you have to compromise. I have found a balance I like, when the dimensions of the wire are 0.3 mm thick and 2.5 mm wide. But it is a matter of taste, and the dimensions can be used in fighting wrong tendencies other places, in the whole chain of components used.
As earlier mentioned, cotton and pure silk can be used for insulation. It can be bought as a woven tube. If you are to classical music, you undoubtedly should choose silver with silk insulation, as the most neutral and with a very precise sound stage.
Two main discoveries concerning cables. When the material of the wire is open to air, we have problems: These are oxidation on the surface of copper and sulphonation on the surface of silver. The surface has to be closed from contact with the air. I was working with pure silver, preferring that material, so electrolytic plating with gold seemed natural. Surprisingly it sounded better with gold than without.
This observation led to a new version, silver first plated with copper and then with gold. That sounded even better, so at this point I paused - satisfied. It seems as if the wire should have low resistance in the centre and higher and higher impedance towards the surface. What drove me to my next experiment I can't tell, but I annealed some of the gold plated cable as well as the copper-gold, and big was my surprise, when I took the material out of my little kiln. The gold had disappeared. It couldn't have evaporated, so it had in some way reacted with the silver/copper - so to say, sunk into it - so the whole purpose with the gold plating was lost. The gold version turned white and the copper-gold version turned black. Well! I couldn’t reverse the process, so there they were for closer listening.
The black version was really awful, but the white- Surprise! Surprise! It sounded most promising. The whole idea of protection was cancelled, and a new search began. A series of plated wires was made with gold in varying thickness. 4,6,8,10,15,20,25 micron. Not at once of course, as it is rather expensive process, but all were annealed and gradually listened to always in four versions at a time. A miracle occurred at 10 micron, 15 even better but at 20 and 25 it was as if the sound was effected in a slightly wrong manner.
It is easy to see when the thickness is right, the surface should look exactly like white paper used for machine writing. No metallic shine at all and no cream colour (too much gold) To explain the physical effect, as I see it, we can regard a cable as a very long corridor with reflecting surfaces and even adjacent rooms and the current as waves of sound.
It is then obvious that the dimensions influence on the propagation of sound, as the reflective coefficient off the walls does and as the adjacent rooms - (the connectors and insulation) - do as well. The mixed crystals formed by the silver and gold serves as dampening the reflections from the walls of the wire, seen from within. The surface has changed from metallic and mirror like reflection, to nonmetallic and diffuse reflection - in other words from order to chaos like, also for the electrons to meet progressively. This slows down the speed at high frequency, which is forced out towards the surface.
For protection of the surface of the wire, another solution had to be found. I had suspected the effect of plastic being caused from its ability to hold and deliver a static load. It suddenly came to me, that layer of linoleum on floors doesn't build that up. This material is heavily based on linseed oil, which is a drying oil why that was tried for protection of the surface. At first the wire and later on the wire with is silk insulation was dipped in the oil and dried at 60 to 70 degree Celsius in an electrical oven. It takes 2-3 days.
The listening results were well beyond expectation. I had feared it to have some deteriorating effect - big was the surprise, as it turned out to be even better than the gold plated. It was therefore with misgivings the same procedure was done with the gold plated wire - was all that much money just thrown away? Luckily their better inside treatment of the signal worked clearly through. Used as digital connections I must say the gold plated annealed and with linseed oil treated cable is without competitions - but the price of manufacturing - Ugh!
The linseed oil has a secondary very positive effect, as its affinity to oxygen is very high, whereby the surface of the metal is cleaned effectively from oxidation and kept in that condition. Wires have an orientation created by crystallisation of the metal, much like domains in magnetic material, so it is advisable to mark that, if they are cut from a longer piece of wire. When connected they must have opposite direction to one another. In this way the difference of sound is greatest, when the cable as a whole is turned - if you can hear the difference hidden.
Listen for s, t and f sounds and the whole atmosphere in recordings of choir.
A marvellous recording to judge from is the Philips recording of Misa Criolla by Ariel Ramirez with José Carreras. Philips 420 955-2 Among other thing listen for the choir, you should clearly understand the words sung.
Another record I can recommend, is a Opus 111 recording:
Songs of the world” with Moscow children's choir. OPS 30-157.
Here you are presented children voices and Orf instruments with great clarity, and a marvellous piano, so naturally recorded. In general Opus 111 recordings are very good and worth buying, if you can find recordings of your taste in their catalogue.
Cables for the loudspeaker
These can be made in exactly the same manner. You may here consider if you want single connection or multi way connections, and balance the dimensions of the silver for the unit to connect.
Single termination:
The dimensions found for the interconnect-cables works fine, if the length is within 3 meters.
Multi-way termination:
Treble: dimension 0.15-0.2 mm thick 2 - 3 mm wide
Midrange: dimension 0.3-mm thick 2.5-3 mm wide
Bass: dimensions 0.4-0.5 mm thick and 4-6 mm wide
To get the cables to look nice you can, after the treatment with linseed oil, put on an extra and thicker tube of silk. Beware of synthetics, when you buy these tubes. They can be pure synthetic or have synthetics within, without the dealer to know. It is advisable to check it with fire.
Cotton shall glow, with no melting when the open fire is blown out. Silk shall be unwilling to burn, but melt, bubble up and smell like burnt human hair. I like the end result to behave in a manner, so you can't feel or detect it to have a metallic wire inside. So I don't hesitate to put it all into the linseed oil once again, and put on a third tube, to get that behaviour from the end product. It sounds, as if, the better the wire is protected from the outer world, the better it sounds. It’s so totally contrary to plastic insulation.
Power-cables.
These very well insulated cables have also been tried used for the main power. Here they showed their true qualities with the most surprising and inspiring results. A recently performed listening test of an OTL vacuum tube amplifier showed the importance of that part. With normal power-cables it sounded very wrong, fat, with loose of dynamic, distorted and unable to establish a steady soundpicture of a single instrument. It was rather funny to hear a single acoustic bass split into two. But a metamorphose occurred, when those treated with linseed oil were used,. Now the amplifier was most interesting, the best of that type I have ever heard.
Further tests with these cables have showed the same effect of cleaning up on all other amplifiers even switch mode and pure digital. This only when silk was used, cotton with its more aggressive character wasn’t good. That the effect is so big is most disturbing. How many amplifiers around the world play with just a part of their capacity solely because of a bad connection to the power plug? It seems unfair that regulative for high voltage is destroying our possibility for enjoying music.
These cables are probably illegal, so don’t leave them connected, when you are not present. I have been presented an idea of protection that may solve the problem by further insulation with a tube of woven glass. This material strengthens the construction and protects it from heat (up to the melting point of glass), but it must be impregnated with anti-static liquid, as glass also is a static material. Impregnation with linseed oil seems to work, but for many people it don’t look sufficiently nice, and it still is a question if it is legal.
Legal or not – it is so easy to change that part, that you can use it when you want to listen seriously, and change it to the legal ones for background music or keeping your equipment stand by. This way of thinking leads to the next chapter.
------------- Pure Duelund www.duelundaudio.com
|
Posted By: Guests
Date Posted: 07 Feb 2009 at 01:29
Fedt, det er herligt at se. Hvis du ligger inde med billeder af nogle duelund højtalere eller lignende kunne det også være spændende at se. Henrik forsøgte iøvrigt at implementere "Synkron Filteret" i Lspcad - Det lader sig godtnok til at være sværere at udføre i praksis end man kunne håbe på.
|
Posted By: Duelund Coherent Aud
Date Posted: 07 Feb 2009 at 15:27
Tak, jeg lægger løbende alt op, som jeg har. Noget er rent flyvsk andet mere konkret.
------------- Pure Duelund www.duelundaudio.com
|
Posted By: Duelund Coherent Aud
Date Posted: 07 Feb 2009 at 17:43
Følgende kommer her:
How to build to the limits of possibility.
Comments
The search of perfection
Loudspeaker impedance correction
The bass response
Comments on peaks in the frequency response and measurement
How to fight peaks in an impedance linear way
How to soften the rear suspension
How to find the mechanical filter’s components.
Method of measuring
The magnet
The voice coil.
The diaphragm and the front suspension
A dynamic tweeter – not a dome and not a cone
------------- Pure Duelund www.duelundaudio.com
|
Posted By: Duelund Coherent Aud
Date Posted: 07 Feb 2009 at 17:45
Artiklen i sin helhed med billeder: http://www.steenduelund.dk/download/HJEM3.WPS.pdf - http://www.steenduelund.dk/download/HJEM3.WPS.pdf
Comments
In this chapter you will be presented problems with a solution but also problems
without.
I hope this paper will create a forum world-wide for discussion of these subjects and if
possible form a brain trust giving a lot of proposals – good as bad – from which we all
can be inspired.
It is time for us amateurs and those of the professionals interested in good sound to fight
against that ignorance, carelessness, cost benefit and resign the manufacturer show,
even when they demand a tremendous amount of money for there production. It seems
as if all energy is put into appearance and design. Would you bye an ALFA ROMEO
equipped with a CITROEN 2CV motor
Thinking back, I will point out the very popular AR-3 as an example of a construction,
where energy was put into its working parts – no matter how it looked behind the front
cover. It must be the ability to recreate sound that counts and not so much how it looks.
If you focus on looks then there are lots of nice looking constructions to choose from.
The search of perfection.
In search of that you must have an open mind and look at all parts in and around the
loudspeakers as bricks in a puzzle that must end to fit together.
The amplitude and phase versus frequency, determined by the crossover, units, cabinet
and room, are most complex, and it will be impossible to make fit together perfectly.
But it is with our hearing as with our eyes, there is a limit for how small a chink can be,
for us to see it.
Our hearing is much more forgiving, as it works in time, where no stand still for further
inspection is possible, but that doesn’t change the fact, that the better the chinks are
filled out the closer you come to the recorded event. When all is perfect, you just have
to close your eyes to see.
Our measuring scale - the dB scale - is a rather coarse one. Normally one would inspect
for linearity of summation as the only possibility. This gives a rough result; as it doesn't
tell, how it is achieved. Every speaker can in broad outline be seen as a vector. What
you measure is the resulting vector and that tells nothing about its origin.
A main problem by measuring is the fact, that you listen with two ears placed at your
head at the top of your body. This enables you to hear the origin of the resulting vector
dimly if you are not trained - clearly if you are.
In order to avoid that, all units must play in phase with each other very precisely
focused at your listening position. You will experience this area, but despite that the
reproduction is strangely still stable outside this area, if you keep your height for
listening reasonably constant.
Therefore we need a complementary technique for measuring.
Using my filter topology and the special cases of it – The Linqwitz-Riley filters - every
parameter is to calculate, so also a magnifying technique, which is straightforward.
Turn the phase of one of the units. Now instead of a 6-dB summation, you have a more
than 40-dB useful nullification, and that shows even the smallest anomaly in level as in
phase. Especially the rounding of the shoulders is most revealing for mismatches.
This magnifying technique is also very good to find the actual acoustic centre, of
greatest importance when the units work together. This addition reach far wider in
frequency than normally looked upon, and if you want filter functions with no overshoot
the slope of cut-off only gradually reaches its maximum no matter its order.
When you shall find these centres, the unit must be filtered completely and take care,
that the two units has the correct level according to theory at their common frequency.
Then you just find the height for the microphone, where the two units nullify each other
best possible at that specific frequency. From your point of listening you then shall see
the microphone pointing exactly at the middle between the centres of the two units’
voice coils. If they don’t you must push the unit apparently nearest the microphone
backwards. When that is in order you measure a sweep on the two speakers
1. In phase
2. Out of phase
And compare the results with the theoretical curves. These curves can be calculated
from worksheets to Mathcad 7 for the 3 and 4 way systems, which are present at these
pages.
The formula for the calculation of curves used with this technique is straightforward.
You just find the value
20*log (H (j w)) for sufficiently small steps of w e.g. 20 to 100 steps per octave.
Remember that w=1 is the centre frequency.
You can also find graphs, which are calculated from the formulae for the 3-way version
with a value for (a) equal 2.828427 except for 2 nd and 4th order 2 way.
These curves are stored in an image file “Antiphase”.
They are calculated from:
H (j w)= +w4 + (a2-2) w2 + 1 .
( w4 -(a2+2)w2 +1) + (-2aw3 + 2aw)j
You just have to change signs on the expression.
Figure 1. Bandpass antiphase (the +sign at (a 2-2)w2 is changed to -)
Figure 2. Treble antiphase (the +sign at w4 is changed to -)
Figure 3. Summation of bass and treble ((a 2-2)w2 is taken away)
Figure 4. Difference between bas and treble (as 3 but with - at w4)
Figure 5. Difference between bass and treble 2-way 2 nd order Linqwitz-Rieley.
Figure 6. Difference between bass and treble 2-way 4 th order Linqwitz-Rieley
If a demand rises, curves for the 4 way even 5 way will be established.
Using this technique, the need of fighting every peak occurring on the roll of down to at
least -40 dB, has been shown. With the single unit you should go deeper, I would go for
-70 dB ore more, but it stretches the demands on the units and measuring equipment
beyond their normal ability.
The problem with units’ range of frequency is by these high demands put into relief.
In some way the manufactures of loudspeaker units have forgotten the importance of
geometry of the diaphragm. New material for the diaphragm pops up, but for me to see,
they all are formed into the same shape, and that can't be right.
It is of course the cheapest way, but given the basket some thought, it should be a minor
task to get possibility for displacement of the magnet assembly, and thereby freedom of
shape for the diaphragm. As an example you could examine the construction of this part
used by “Audio Technology”.
The main problem with the geometry is the roll off towards the treble. This is heavily
related to the slope of the diaphragm and the front suspension as well. Private research
in this area on an 8 inch unit, has shown it possible to construct a really full range
loudspeaker with very good treble reproduction as well, and still have minimum phase
behaviour
I can in this paper only present, what I think right. I'm not infallible, but by many years
work with loudspeakers and related topics, with no commercial interest to stop me at a
given development, I feel to have got a deep understanding or intuition of, what is right
and what is wrong concerning sound reproduction. The whole process seems so
complicated and is yet rather simple. It is the process of doing it all correctly in the long
chain from source of voltage to loudspeaker and further to the listener, that is so
overwhelming.
My last private project will result in a four-way loudspeaker for me and friends to toy
with for the rest of our lives, and give us the opportunity at last to enjoy music, without
the ever irritating noise annoying us through the years. This construction is open, it
means that, it can be changed in a multitude of ways, only the cabinets will be the same,
though their working manner also can be changed from closed gradually to bass reflex,
if that should come into question for some.
This property is build into the construction to be free to try every new components and
loudspeaker units, when ever they turn up, or are made by own hand. As the demands
for every single loudspeaker is determined in all detail, this can be done. All level and
position is to be optimised, the dividing network easy to reach, the important baffle
Side 4 af 25
curvature is changeable and so on. Realising that it's a never-ending work, I call it 'The
No End Loudspeaker'. That construction will hopefully come on these pages in the year
2000, and will be my final statement of a 4-way loudspeaker based on the closed
cabinet.
First now after 25 years work with loudspeakers, I finally believe to know in all detail
how to build this most complicated thing to satisfaction. Former models are still playing
to their owners’ joy. But the real break through where every part is optimised is for now
at a temporary stand still, waiting for the last parts to be made. A real problem for me is
to stop new ideas to pop up, but my friends hopefully will hold me focused at the
project.
Loudspeaker impedance correction.
When you work with a dividing network, it is a must that the impedance of the
loudspeaker has to be corrected to DC-resistance, not just linear as normally seen.
From measurement taken you must determine following.
· DC-resistance =R
· Resonance frequency =Fr
· Resonance impedance =Zr
· Frequency for 2*R (voice coil inductance) =F2
From these you can calculate the impedance, for which you shall find the frequency just
under and over the resonance frequency.
Look for (sqrt(R 2+Zr2))/2 and find Fl(ow) and Fh(igh)
To control your reading sqrt (Fh*Fl) should be equal to Fr
You can now calculate values for the components in following circuit.
To calculate C1:
p=sqrt(3*R 2) / (2*p*F2) and q=R / (2*p*p) then
C1=1000000 / (2* p*q*R) uF
To calculate L2, C2 and R2
L2=1000*R 2 / ((Zr - R)*2*p*(Fh - Fl)) mH
C2=1000000000 / (L2*(2* w*Fr)2) uF
R2=R*Zr / (Zr - R)
If all was perfect, these values should work, but it isn't. The voicecoil is placed into the
magnetic field, surrounded by iron and perhaps copper.
You will therefore probably face mismatches with some or all calculated values, but you
will know about where the values are.
It is advisable first to regulate the voice coil inductance, as this network influences the
readings on resonance impedance. With copper around the voicecoil you must
experiment, as this copper complicates this process.
The C1 works in the lower end and R in the higher end of the rise of impedance caused
by the inductance in the voicecoil. You shouldn't be surprised if you have to change the
calculated value on C1 and even double the R-value. This leads to a further rise of
impedance to be compensated again to look:
The values depend on the unit used, so you have to experiment.
To my experience great care should be taken to achieve exactly DC resistance. If not -
you will have hidden reactance to react with your filter components, sometimes leading
to rise of level, more than that received from the unit without any components attached.
This extra energy is created by the components within the unit together with the
network, and should obviously be avoided. It is a resonance.
Another strange parameter is the need of quality for the components used for this part.
Despite they’re coupled in series with a resistor, they must be of very high quality.
When this is fixed, you should measure again, to get the value needed for correction of
rise of impedance on resonance frequency. Here copper again can interfere, so if that is
present the inductance must be raised by a factor 1.4 and capacitance lowered by a
divisor 1.4. (This value can vary dependant on factory)
You probably can't avoid use of electrolytic capacitors of the bipolar type. If you need
high values you can connect two polar capacitors in series negative to negative or
positive to positive (dependant of can connection) to reach that property, beware of
voltage - that doesn't double and should be chosen high. (100 v or more)
The calculated value using C1 and C2 should be C1*C2/(C1+C2) and remember to take
off their plastic cover.
You can improve the performance of this coupling by adding negative/positive voltage
(dependant on polarity chosen) to their connection. See fig.
C1 and C2 will, if the battery voltage is chosen high enough, never reach zero voltage,
where the memory effect comes forward. It is a cheap trick for you to try.
The coil could be a toroid coil with airgap in the iron, and able to handle high current, or
the former described air-toroid dependent on frequency - iron for low and air for high
frequencies. (Over 400 Hz).
The resistor must be able to handle high effect and its calculated value has to be lowered
with that DC-resistance present in the correctional coil. Some like to place all
impedance inside the coil, but then beware the warmth generated.
But you are not finished yet. You should test how it all behaves under dynamic
conditions. For that you should play some music, at a sound level you normally prefer,
for an hour or two. Then measure again and look for changes due to rise of temperature
in the parts, and correct if necessary.
On could believe this part of the work to end here, but it isn't, you’ll probably have to
return to this circuit to make slight modifications, to optimise the roll off of the unit,
when the filter components is attached. It all is heavily connected, as the output from
the unit still is dependent of its Qt, even if it is electrically corrected. If it comes through
you can lower the impedance in the correctional circuit.
It is a most tiresome work to get it all right, but there are no ways around these
adjustments to get it all in order.
The bass response.
The manufacturer of loudspeakers does not treat this part of the frequency band in a
proper way. It is simply not done right. One problem with bass response seems to be
our brain’s tendency to manipulate with that. What you apparently hear is not what your
ear register. It seems as if, parts of the bass response are overheard and others amplified
by the brain. This could be related to the complicated turn of phase on modern
loudspeakers. Experiments indicate that the capacitive turn of phase is disturbing for the
brain’s work. This also goes for too much inductive turn.
It is a general opinion, that bass under a certain frequency is impossible to point out.
To my experience this is untrue, at least when playing recorded music. I know very well
that deep bass has a tendency to be recorded mono as the wavelength is long, but the
same tendency is repeated played back on your loudspeaker. The bass loudspeaker is
further filtered upwards and thereby delivers parts of sound easily to point out. So for
the sound stage to form in a proper way, you must have a stereo pair of bass units.
Furthermore the level of bass is just as critical as the level of treble.
Lots of energy has been put into the bass reflex solution, with its magnitude of
problems. I started with horns, left them and have tried all the different systems through
the years, with a long stop for now at the closed system.
For me it must be full range, and here it's fare more a question of quality than of
quantity. To my experience the brain will do the rest, when it has understandable sound
to build on. Only the closed box, with slight modifications, has the right potential to
work together with your room for listening. That's a main fact often forgotten.
They do have a bad reputation, based on wrongdoing in the past. This primary in an
attempt at in magical way to conjure bass from small cabinets, which led to that fat
pumping bass response, bringing AR in the hall of fame.
This attempt is now done again with the bass reflex idea.
The real problem with bass is its deep dependency of room behaviour, as the
loudspeaker is a box placed in another box (the room), for the bass waves of sound to
see.
Rooms in general build up anti wise a soft curved 12-dB roll of, which is created
perfectly with the closed box. This build up is dependent on room size and reflection
coefficients; why it's impossible to develop a loudspeaker in general. It can only be
optimised to the specific location. Even though there are a certain line of uniformity in
our living rooms, the loudspeaker manufacturer still develop loudspeakers to play in the
open, if they care of bass at all. It is, I’m sorry to say an overlooked parameter.
The level of bass is really critical, why it must be to regulate both for level and decay
downward in frequency.
In order to have standing waves under some control, you must use great magnet power.
The loudspeaker units also serve as microphones when playing. When standing waves
build up in the room, the bass loudspeaker generates a signal reflecting that, for the
amplifier to see as a rise of impedance, whereby the level is lowered. It is so to say selfrepairing,
but to force them to do that properly, the magnet power should be very high.
This also fights bass output by creating a low Qt. So how to conjure bass again.
There are different solutions for that, but they all demand closed cabinets, very low
resonance frequency and low Qt for the loudspeaker with enclosure.
Units present on the market in general don't fulfil these demands. There are a few that
can be rebuild, but adding a magnet on the backside of the magnet construction might
help sufficiently. For that to be the iron part must be plane, in conflict with a good part
of units on the market.
The simple and the best to regulate
With great magnet power you also should have higher sensibility, if it isn't wasted in
widening the airgap.
It is possible to lower this sensibility, and in same time keep the bass output intact.
The need for that is of course dependent on room and bass response.
The real advantage in the solution, is its ability to be optimised to perfection to fit into
your room, but the gained efficiency and perhaps a little more will be lost.
Connecting a second equal unit in parallel will increase the energy with 6 dB.
This again has a price in doubling the volume of cabinet and halving the impedance so
the amplifier must deliver the double amount of current. This shouldn't create real
problem, as smaller cabinets often are placed on a stand. The volume within the stand
should be the volume for you to use. The impedance is halved, but if you start with 8
ohm you reach 4 ohm, easily for modern equipment to work with. (The actual DC
resistance is lower (6 and 3))
The circuit is straightforward:
The network is constructed to have linear impedance, dampening in normal way (Rs and
Rp). The Ls and Cp parts outbalance each other.
At low frequency Ls is open and Cp closed, at high frequency Ls is closed and Cp open.
Thereby the resistors dampen the higher frequency response and the reactive
components lead the deep bass through.
The phase turn of this network forms an inductively turned low “hill” on the phase
curve, and it is the placement of this “hill” that is critical.
The mechanical parts forming the resonance frequency form likewise around the
resonance frequency phase turns. Under Fr an inductive “hill” and over Fr a capacitive
“anti hill”, these two curves nullify each other at Fr. It is into this “anti hill’s” upper
part, the of the network created “hill”, must be placed, trying to straiten it out.
You have to find this point of your loudspeaker, dependant of resonance frequency and
Qt. According to my calculations it can be found with sufficient precision from the
expression:
Frequency for maximal capacitive turn Fm=Fres * 1.1/sqrt (q) when q is smaller or
equal with 0.5. (That it is 1.1 and not 1 is caused by the influence of the inductive
part)
You should work with low Q-values.
The matrix construction helps considerable on this point. With Q-values over 0.5 you
must increase the magnet power by attaching an extra magnet, or try to soften the
suspensions. The front suspension can be treated with silicon oil, used to protect the
rubber lists on your car. This helps enormously on the dependency of temperature some
modern rubber/plastic types have. Concerning this front suspension, it should be soft,
but regrettably the foam, often used for that, doesn't last, but break down in structure.
Also here this silicon oil helps on its durability.
From measurements in your room you should know how many dB you want the very
deep bass level (under the lowest standing wave) to be risen. You know the
loudspeakers resistance and the frequency of maximal capacitive turn.
With these you should calculate.
Loudspeaker resistance =R
Wanted damping in dB =dB (shall be negative)
Rs= R*(1 - 10 (dB/20))
Rp= R*(R/Rs - 1))
Frequency for maximal capacity turn = Fm
c=1/(1000*R 2)
f=sqrt (1000*Rs/(c*Rp)) / (2* p)
Ls=f/Fm mH
Cp=c*f/Fm * 1000000 uF
Control: 3.2 ohm, -8 dB, and 50 Hz gives
Rs=1.926, Rp=2.117, Ls=9.716, Cp=948.785
The inductor coil must have very low DC-resistance.
With this circuit attached, you further can regulate on the size of Cp, to change the slope
of dampening. If you decrease the value of this part, you can smooth the work of the
network. This could be chosen, if your living room is of the reverberate type. But you
have to experiment with that.
The electronic way.
If you drive your bass unit with an amplifier of it own, there is the electronic solution.
It's called the Linquitz-Grainer network. It's an active part with which, you - within
some limits -can decide the behaviour, your loudspeaker will seem to have concerning
Qt and Fr. But most important it will nullify the turn of phase created by your bass unit.
It has been used to conjure bass up from small boxes with high Qt. That is to use it in a
very wrong manner. It will correct the input all right, but will do nothing to the tendency
of a high Q unit to vibrate at its resonance frequency generated by the standing waves. It
came forward, was used and disappeared again, as it couldn’t correct a 15 inch bass unit
in a 20 litre cabinet, and is to my knowledge rather unknown.
Also for that to work properly, you must have low Qt.
You must search the literature for that. If interest arises I will put it on this page.
An interesting use of this would be to straighten the units all to have the same resonance
frequency and Qt, and then use theoretical filters, this only when electronic filtering is
used,
The charming way .
With this you again must use two loudspeakers. But this time you don't connect them in
Side 10 af 25
parallel in the normal way.
The first of them must be seen as a single unit and treated, as should it be the only
loudspeaker of its kind.
The second unit is corrected for impedance variation as normal to form a resistor
electrically.
If the two units share the same volume, the impedance variation must be found with
both units driven simultaneously. If they are identical, you can measure both connected
in parallel and double the read values. If they are different you must use a stereo
amplifier and tone generator, and calculate the impedance using a voltmeter and an
ampere meter for each of them.
The LCR circuit - to correct for the rise of impedance around resonance frequency for
the first unit- compensates and thereby reflects the magnetic power seen from the voice
coil. Thereby it also reflects the decay of level, as current through this circuit. So what
you do is simple. You substitute the resistor with the second unit - also impedance
compensated -, throw the capacitor away, as it will damage the electrical Q of the
second unit, and change the inductor L2 to a type able to handle high current. In other
words with very, very low DC-resistance.
You now have a new single loudspeaker (both seen as one) with a 6 dB lift in the lower
end, if your amplifier doubles current when the load is halved.
This method is filling the decay caused by the great magnet power beautifully. The filter
function used consists of the mechanical capacitor inside UNIT and the passive L2, and
is therefore of first order. You can, if you like, make it to a mild second order by
connecting a further C1 parallel to UNIT2.
This capacitor you may choose with or without. Listen carefully, if you can hear the
difference, and decide from that.
Again there is a price to pay, as the two units are coupled in parallel at the lowest
frequencies. But in normal music there are not much energy there, so you shouldn't care.
I call it charming, because even the smallest detail in the bass response is heard. It
somehow is so easy to listen to, so charming and in my ears so right in its richness of
detail in the bass, seldom heard in reproduced music.
Side 11 af 25
When playing very loud there may be problems. If the two units are too different and
they share the same volume, they are connected electrically and mechanically in the
lowest band. Take care of their equality and you should have no problems, or don't play
very loud - it isn't a disco loudspeaker - just loud.
This solution can be chosen with to different sized unit if your room for listening is
large, as the amplification done by the room, then will lie low in frequency. You also
may choose two enclosures and experiment if they should be sealed or connected with
an airflow resistance. Their equality of Qt and resonance frequency is important and can
be altered by loosing suspensions, adding weight or adjust magnetic power.
Comments on peaks in the frequency response and measurement
Variation in frequency response is normal unavoidable. The unit forms it, but also
reflections on the enclosure play their part. These variations are easy to measure, but
they are not as easy to extract the right information from.
Working seriously with loudspeakers you must have lots of damping material around
treble and midrange, in order to hide the cabinet and it edges from these two
loudspeakers whole/upper band.
That fits perfectly with the need for these two units to be pushed backwards to place
their acoustic centres directly over the bass unit’s, so there is good space to be filled.
Actually the acoustic centres should form an arc of a circle with the distance to your
listening position as radius, and the units should point directly towards you.
Don't listen to these people, who tell you to listen of axes. If they recommend that, you
can be sure something to be very wrong with the loudspeaker. That, not you I hope, can
fool only persons, who weight the measurements done by a microphone.
By measurements you should use different distances between microphone and unit.
When you examine this row of curves you will see peaks/dips which are steady and
others that move in frequency. The “movers” should be judged only from measurements
taken with long distance to the unit, as the movement in frequency is caused by the
geometry in and around the unit together with the distance to the microphone.
You should fight the irregularities by mechanical means, as far as you can, before you
regulate them electrically.
It might be help for you, to find the sources of these, if you look at the transport of
sound near the speaker not as sound but as a stream of air. That it actually is so, you
can examine by the flame of a candle. It is therefore clear, that to avoid turbulence and
reflections the surface must be very smooth. It is also obvious that the front suspension
must be a main problem together with the sudden disappearances of support from the
baffle.
This work is so closely connected to the practical making of a loudspeaker, which you
have to wait to see used on a construction.
It is so common only to concentrate at the part of high level from a loudspeaker, as if
their roll off is of no importance. I’m convinced that to be wrong. Again an opinion
based on wrongdoing.
Mechanical filter parts within the unit you can use as well as electrical ones. You can
Side 12 af 25
even change them magnetically, by adding weight or by softening the suspensions.
The lower roll off and the area, where it works as a p!ston, is normally the most perfect
parts of the loudspeakers response , so why not use that whole. The only problem is the
units’ behaviour at resonance frequency, but that can be solved.
The front suspension has a tendency to form second order distortion by the uncontrolled
deflections of the diaphragm at higher Qt than 0.5 – even smaller. That can be improved
on, and the rest is not an enemy of sound as the level from the bandpass and highpass
loudspeakers hopefully are low there, so also the distortion,
When 12-dB filter is used, resistors must cancel the electric components; to let the
mechanical components take over.
By inspecting the C and L dependency of frequency you can get an idea of the size of
the resistors, where you must stop the capacitor to close and the inductor to open.
With 24-dB filter you should be able to use just two components. It is not so straight
forward, as it sounds, and you probably also must work with the correctional network
on resonance frequency, to make it all add up.
This is another way of looking at the whole process. Start with the roll off part to get
that in the right level with a minimum of parts and then work further with the curved
part and be free to vary the load impedance for these parts to see. In this way you will
be surprised, how peaks and dips are connected and corrected at once. It is really a
troublesome work, but what you gain in reducing the number of components, to achieve
the wanted curve, will reward you in sound quality. To my experience you should start
with the treble and let that part decide how the other units should behave. It is much
easier to form them than the treble.
But again - all this is best illustrated in practise on coming constructions.
How to fight peaks in an impedance linear way
Sometimes you have peaks where your filter components are open or have low
impedance. Then you can't use variation of load. In these cases you have to compensate
the peak in a way for the filter not to see. It must have linear impedance.
Now peaks are seldom symmetrical, so you must examine the peak for that, and weight
Side 13 af 25
which side of the peak you’ll find most important, and then make a drawing of the peak
symmetric, from which you can extract the needed information.
Loudspeaker impedance at DC =Zs
Wanted damping in dB = p (p must be negative)
Then serial resistor Rs = Zs*(1-10 (p/20)) Ohm
Parallel resistor Rp = Zs*(rht/rs - 1) Ohm
You now must calculate a new level within the peak for which you must find two
frequencies: Flow and Fhigh.
Level = -p+20*log (sqrt (((10 (p/20))2+1) /2))
To control your readings sqrt (Fl * Fh) = peak frequency
q=Fpeak / (Fh - Fl)
Wp = 2* p*Fpeak
Lp = q*rp/ wp *1000 mH
Cp = 1/ wp/q/Rp * 1000000 uF
Ls = Rs/ wp/q *1000 mH
Cs = q/ wp/Rs *1000000 uF
The circuit:
The components used in such a circuit must be selected with care. The Lp coil can have
some of the Rp built in. For the Cs and Cp parts you must avoid electrolytic, as they
don't follow theory precisely. The Ls coil must have very low DC-resistance. When the
circuit is built, then test it for variation of impedance, a resistor should replace zs, and
the measurement must give a totally straight line.
You should by all power avoid use of this circuit, but some loudspeakers are tormented
by very high peak level, say more than 6-dB, very hard to silence, then this circuit may
Side 14 af 25
be used. It works miraculously well on the peaks beyond hearing ability from metallic
domes, and should be placed as near as possible to the unit, to work at its best.
How to soften the rear suspension.
This part of the loudspeaker is often causing dips and peaks on the impedance curve and
thereby, when passive filtering is used, also effect the level of the loudspeaker. The
suspension should be much softer than normal seen, but serves also as protection, when
the loudspeaker is used in bass reflex systems.
Inside the closed cabinets the air acts as a spring and resists the movement of the
loudspeaker. Therefore the loudspeaker here should have extremely low resonance
frequency in free air, to purify the spring behaviour to be that of the air alone, and in
same time achieve the lowest possible resonance frequency and Qt. The spring character
of the air is further more to be slightly regulated by incorporating an air flow resistor.
The easy way:
You simply massage it with your thumbs to more softness. You could further burn some
holes in it with a solder tip, and let it be with that.
The troublesome but fare the best way:
Don’t try this, unless you are a skilled person for handwork and know loudspeaker
mechanics by heart.
You must take the unit apart, by use of some solvents and patience. You should end up
with following parts.
1. Magnet
2. Basket (if it is possible to take it apart from the magnet)
3. Diaphragm with voicecoil flex-wires and front suspension
From the rear part of the basket you should hacksaw away all parts disturbing the flow
of air and unnecessary for the reassembling.
The magnet and modified basket is assembled again, and supporting parts of wood are
glued on the sides of the magnet and basket to form support for the basket and for the
new suspension.
The mounting holes in the front ring of the basket are used for regulation of the steering
wire, so you must cut a new ring for mounting purpose. This gives you possibility to
create more space at the backside of the unit. Even small compression here causes
problems.
It all are assembled, connections soldered, and a woven nylon wire put into place (see
Fig) and glued to the diaphragm or the voicecoil. If for the voicecoil a metallic form is
used, you should beware the heat built up playing loud, why you must use glue capable
to withstand this heat.
For twisting the wire a screw fitting the hole is shortened down and made flat with a
hole for the wire. This part must reach through the material for the basket and the part
for fastening the unit to the baffle.
You can in this way regulate the tightness of the wire supporting the voice coils
placement in the magnet.
Side 15 af 25
By this method, it is possible to reduce the need of air around the voice coil to as little
as one tenth of a millimetre. My stepson has speakers of that accuracy, he plays very
loud and has had no problem. It works.
How to find the mechanical filter’s components.
This procedure must first take place, when your unit has played for some time, and you
are satisfied with its working manner. Even the slightest change will effect the size of
the mechanical components.
When the unit is corrected for rise of impedance around resonance frequency, that
circuit is part of a system, like that for correction for peaks. Therefore you can calculate
the mechanical components from the components of your correctional network.
R = loudspeaker DC-resistance. Rs is the measured impedance at resonance.
You have Rp, Cp and Lp from the correctional network and find the components inside
the loudspeaker.
Cs = Lp * 1000 /R 2 uF
Ls = Cp * R 2 / 1000 mH
For control: Rs = R*Rp/(Rp-R) Ohm.
From this you can write a transfer function - second order high pass.
Method of measuring.
This is misleading and complicated, if you use a normal enclosure and room, as what
you think you measure isn’t.
Measurement done by the manufacturer is of no use for you, as the difference in basic
conditions is so different. One thing you can be sure of, yours will be worse.
The cabinet, in which the loudspeaker is placed, and the room itself serve as reflectors.
The reflections mix with the sound from the loudspeaker unit, why the graph can be said
to be useless. But there are ways around that
To get an idea of how the unit itself measures, you must place the microphone in the
Side 16 af 25
nearfield (1 to 5 mm) of the diaphragm. This method doesn't tell the truth for the upper
end of the unit, where dimensions and runtime of sound will interfere, but from the
frequency where the ½ wavelength is bigger than the diameter of the diaphragm you get
a very precise trustworthy measurement.
In order to measure the rest of the frequency band, you must silence the cabinet with
rock- or glasswool, and perform a row of measurements from different distances. The
room will disturb, but by examining the graphs, you will get a reasonable good idea of
the units properties.
On order to find the sources of irregularities a calculator must be at hand, and it is
advisable, before you measure, to find the frequency for ½ the wavelength for all
distances form which to expect disturbance.
Irregularities are either stable or moving in frequency dependant of distance between
unit and points of reflection and microphone.
The "movers" are created by change of geometry between microphone and unit and
must be optimised to listen position. These should by all power be corrected
mechanically to be stable by damping points of reflection with SOFT felt or rock wool
placed on the baffle.
The stable peaks need action on the unit, according the principle "Try and Error", and
can be a rather costly affair, considered the amount of unit wasted.
It is less complicated to go for electrical solutions, even if they have a bad reputation. It
is much more a question of quality of the components used. Here you must choose
between the Devil and the deep blue sea. Some like these peaks as they add more pace
and drive to it all. I don't, I find them disturbing.
To get a stable curve to work from, you may measure from a distance of 15 to 20 cm
and find what is missing in its upper end comparing with measurements from distance
(2 to 3 m), and remember to account for that mismatch in your work.
The graph must follow the wanted graph form this distance, but not be measured from
it.
When you believe it to be right, you should measure the unit from different distances
and see if the tendency is invariant of distance.
When this is done for every single unit, you must control how they work together.
You must work with two units at a time, and have your microphone placed equally
spaced from the units acoustic centres. From every measuring distance you must
perform 4 measurements of level.
1. First unit.
2. Second unit.
3. First and second in phase
4. First and second in antiphase
Side 17 af 25
From this curves in comparison with the calculated ones you should be able to get the
two units to work correct together. You must undoubtedly modify your filters, but so it
is aiming at perfection.
You must do this procedure with every
------------- Pure Duelund www.duelundaudio.com
|
Posted By: Duelund Coherent Aud
Date Posted: 07 Feb 2009 at 17:57
You must do this procedure with every two unit in your system.
Reflection on the floor will disturb measurements in the bass. You can be free of that by
placing the loudspeaker enclosure on its side, for the loudspeaker unit to come as close
as possible to the floor, and measure the bass from that position. Raised again you will
still have the problem, but when playing it needs time to build up, so the transient will
come through in correct manner, and that count most..
When all this work is done, your loudspeaker probably doesn’t look nice. The next work
is how to reduce and/or cover the needed damping material. That I’ll leave to you, but
remember to measure and measure again under this process.
The magnet
This part is easy to make yourself, if you have access to a turning lathe and a very
powerful magnetiser.
Years ago, before high-powered amplifiers came forward, much work were done, to
increase the efficiency of the loudspeaker. Lowther fullrange units put into a great hornconstruction,
was a masterwork from this time not to mention Klipsh and Tannoy. The
magnet was very cleverly constructed. But a war in Africa put a spoke in the wheel for
deliverance of cobalt. The alnico was replaced with ferrite, which in many ways is
easier to handle. The precision wasn’t so necessary any more and the row of so-called
improvements placed their deteriorating trace on production. These “improvements”
were much more directed on the yield of the invested capital than of the quality. Onto
that came the closed box’s believed lower demand on Q-value for the bass unit, which
was a helping hand for the bad development.
You can’t get bass from a powerful unit, was the general opinion and still is.
Two very important facts aren’t taken into account.
1. The loudspeakers are placed in a room.
2. The moving mass must be high, and the suspensions soft to keep the resonance
frequency low.
It is really strange to me, the amount of energy put into the reproduction of treble, which
ought to exist hidden, where the bass, which also should exist hidden, is totally
forgotten. It’s the least troublesome to calculate and make.
A room has reverberation governed by dimensions and reflection-coefficients of the
different surfaces but also the frequency of the signal, when the wavelength is
comparable to the dimensions for the room. Then resonance builds up causing great
variation in sound-pressure dependant of position in the room. The summarised effect of
this in the bass area can in effect be compared to a mild inverse 12-dB low-pass filter.
This effect should be taken in account with the loudspeaker’s slope of cut off, why the
closed box is the only useable when room and loudspeaker are looked upon together.
The room was compared to a mild inverse filter, why the bass loudspeaker also must
have a mild decay - a mirror image of the rooms builds up. To achieve that, the magnet
power must be very high. With the loudspeaker placed in a closed box the decay
probably still is too quick, so further regulation is needed. This is described in “The bass
Side 18 af 25
response”.
A further argument for great magnet power for the bass unit is the simple fact that a
loudspeaker also acts as a microphone - also described under “The bass response”
The power parameter - talking loudspeakers, is the Bl product. “B” the magnet power
and “l” the length of wire in the magnet gab is what creates the force factor. Lots of
other parameters interfere, but let’s stick to these two for now.
To increase the length of the wire you must lower the amount of insulation to squeeze
one or two windings more on the coil – seen on single wounded coils with flat wire.
But there is more to gain by increasing the diameter or height of the coil present in the
magnet gap. That leads us to the construction of the magnet as the main point.
The normal production of magnets for loudspeaker is too much guided by low expenses.
It will cost more to make them better. Modern technology with CNC-steering makes it
now within economical reach to make the parts softly shaped to avoid sharp edges. The
nature of the magnetomotive force is to establish a magnetic field reaching far out.
Ferromagnetic material attracts this field to be concentrated in the airgap. The iron
normal used should have high permeability (ultra low level of carbon), but again that
costs - so the demand for this product has sunk so low, that production to my knowledge
has stopped. We have to be content with, what we can get. Therefore the iron must be
thick. It is normal to use 6-8 mm plates – It is far too thin, I use 20 mm.
The centre tap is and must be the keyhole in the magnetic circuit, in order to establish as
many field lines in the airgap as possible. Therefore the dimensions of the centre-tap
dictates the greatest height of the airgap.
The cross section forms a circle with the area of r 2*p.
The receiving band of the front-plate with a gap of 1 mm is (r+1)*2* p*height of iron.
The maximum height is therefore: r 2 / (r+1)*2
Examples: r=20 mm, height =9.5 mm r=25 mm, height=12 mm
As a rule of thumb the height can be set to a quarter of the diameter of the centre-tap.
That the height of the front-plate often is seen thinner, is an attempt to saturate the iron
around the voicecoil. For that other types of magnets as Alnico or Neodymium must be
used. Again the price is high
Side 19 af 25
Figure: The magnet construction
As seen from the figure, a construction of copper-rings fights the rise of impedance
caused by the voice-coil. Likewise these rings fight variation of the permanent magnetic
field. I’m not sure if more could be reached by expanding them to follow the inner
curve on the front-plate and cover more of the top of centre-tap. I’m neither sure if the
iron structure could be even better. It would be better, if we could get the corners of the
magnet embedded into the iron. But the variation in dimension for the magnet makes it
very difficult in a production. It should be possible if the iron parts aren’t cleaned for
the rest of iron always formed by the cutting process, until after assembling. Then this
rest can be bent over the corner of the magnet and thereby be adjusted to the actual
magnet. It is handwork, risky for your fingers and therefore expensive.
A further investigation into this, demands access to an expensive computer program, I
don’t have.
The magnet must be seen as a foreign body placed where pressure is high, why its form
isn’t trivial.
I have used this construction for some years now, and have gained 4-5 dB by that.
The voice coil.
For this little part, which is placed in the heart of it all, there seems to be a lot of general
misunderstandings.
The winding form was earlier made out of paper, but by modern time’s higher demands
on effect-capacity other materials have come into use.
1. Aluminium – this material is formed from a rectangular piece.
At the point where the two ends meet, the form is open – a problem.
As it is a metal - vortex current arises, which effects the free displacement of the
diaphragm.
2. Kapton - this is also formed from a rectangular piece.
Besides it is plastic, and that I avoid by all power.
3. Fibreglass is used on professional units, but it is heavy and too thick.
The material for the winding form should be light, thin, strong, be able to resist heat, be
an isolator and possible to wind to a thin-layered cylinder to avoid the open assembling.
The only solution, I can think of, is thin paper impregnated and wound with water glass.
Side 20 af 25
This material could also be the glue that fastens the wire. To secure best possible
cooling, the coil should be wound the way Lowther uses - one layer inside and one
outside for better balance. The open suspension I suggest and the copper rings coloured
black will help too.
If that is the right solution, I don’t know for sure, but it’s a possibility. Another material
could be aluminium oxide with diamond structure or glass but how that can be
manufactured I really don’t know.
Another problem with the voice coil, I have stumbled over, is a missing ability to
deliver energy to the diaphragm at a certain high frequency. This phenomenon is related
to the length from the bottom of the voice coil to where the diaphragm is placed on the
form. It seems as if this point forms a node, unable to deliver energy. If the distance is
e.g. 4 cm, then nullification will occur about 4 kHz. At 1.1 cm it will happen at 15 kHz.
Lowther seems to have had the same problem, as the assembling diaphragm/ voice-coil
is unique. A simpler solution is needed. It will not be possible to construct a good
middle tone loudspeaker in the normal way, if this problem remains unsolved.
The last problem is the wire itself .
It has positive temperature coefficient. This means that when playing, the impedance
vary with warmth. The passive filter is disturbed, as is the amplifier, why transients in
the music are dampened. Solutions involving the amplifier have been suggested, but
why don’t go directly to the cause of the problem and develop a wire material with a
coefficient around zero?
Modern science should have no problem to develop an alloy fulfilling that demand, it
has succeeded with alloys for resistors, so why not try for conductors. The price
shouldn’t count much, as the amount of wire for this particular purpose is low per unit.
A solution, in which I personally believe, would be a mix of copper or silver and
graphite/nanotubes.
The net temperature coefficient would be very near zero – done right.
Graphite fibres could be woven to at thin string, and the metal put on/in by an
electrolytic process. The technique works, but it is impossible to develop to the
perfection for a single person and besides I don’t have the knowledge of metallurgy
needed.
Side 21 af 25
I really would prefer it the other way around – copper/silver with a layer of graphite, but
I don’t know if that is possible.
From literature I know of another possible way to do it. By nano-technology it is
possible to create nanotubes. These are developed from a in 1985 discovered family of
carbon crystals called buckyballs or fullerenes. Much research has been done into this
area to find practical use of these new forms. It is possible to bind atoms to this
molecule as every single carbon molecule is connected with two single- and one double
binding. It should therefore be possible to develop a great variety of conductors.
Hopefully one of these would have low resistance and small temperature coefficient.
The future will show if it is possible. The market for such a product is present
worldwide.
So much for the voice coil, which by its mechanical connection lead us to the next
problematic part – the diaphragm.
The diaphragm and the front suspension
For those who really seek the true sound, these parts have to come into consideration.
The practical and mathematical skill for you to have is now far beyond normal. But let’s
give it a try.
Nature has a solution for nearly anything, if you can find it - so also for the loudspeaker
diaphragm. It is said about the diaphragm, that is must be stiff – and light, but that is a
qualified truth.
The weight is dependent of the force moving it, the position of the resonance frequency
and its Qt and the efficiency wanted.
It is your wish of the final result that determines that parameter. It is a question of
balance and not just one parameter.
The stiffness is a parameter on which I can agree but disagree with how it is practised.
It is normal to use stiff material, but that causes problems at the first break up point,
where the stiffness energises that break up beyond possibility of regulation.
NB! By precise adjustment the first break up of the diaphragm and the nullification
from the voice coil could be placed at the same frequency and by care probably
outbalance one another.
For me to see, the stiffness must be created in shape as well as in the material, which
therefore must have high strength of stretch but not necessarily be stiff. This property is
characteristic of kevlar and carbon fibre. (Fibreglass could be considered as a cheap
material to use for experimentation.)
These fibres can be bought in various qualities and densities in one - two- or tree-axial
weaving. It can be formed in any shape you like and kept there by use of hard hardening
epoxy. The hardening process must happen slowly and the amount of epoxy left within
the weaving kept to a minimum. This material can also be found mixed with thermal
plastic material to be formed easily by warmth.
The shape of such a diaphragm must be guided by mathematics.
In the same way, that the shortest distance between two points is a straight line, closed
lines will form an area between them, a surface, which can have many shapes. One of
these surfaces is the smallest, the “minimum surface”.
A loudspeaker diaphragm can be said to consist of two circles with a connecting
Side 22 af 25
surface.
The minimum surface is by nature formed automatically by a film of soap, which
always will form a surface as small as possible when the gravitational force is accounted
for. The ideal form can be described mathematically:
Radius voice coil: r
Radius rim: R
The searched curve can now be drawn on a millimetre paper with the voice coil centre
placed in point (0,0). The height follows the x-axe and the growing radius of the
diaphragm follows the y-axe.
The amount of points (height, radius of diaphragm) = (h, rd) can be calculated from:
Hmax = r*LN (R/r + sqrt (R 2/r2 –1))
Rd = r*cosh ((h*arccosh (R/r)/Hmax)
This curve is idealised as the angel between the diaphragm and the voice coil is zero,
whereby the transfer of sound to the diaphragm should be perfect. But the speed of
sound in the material for the diaphragm is a major factor to be cared for.
You have the freedom to use any point of height as place for the connection of the voice
coil. Then the radius of voicecoil and Hmax are uncertain. This leads to another set of
formulas, where R, r and Hmax are fixed.
Rd = c*cosh (h/c + arccosh (r/c)) where c is found from the equation:
Hmax = c*LN ((R/c + sqrt (R 2/c2 – 1))/(r/c + sqrt (r2/c2 – 1)))
The speed of sound is dependent of the material and the epoxy, but to make it even
more complicated it is further dependent of frequency as well. The speed is somehow
inversely dependent of frequency by a law I don’t know of, but that shouldn’t disturb so
much, as it further again is a complex mix of longitudinal and transversal waves, that
you can’t avoid measurements. You most learn from them.
It is clear, that near the voice coil where the curvature of the diaphragm is steep the
higher frequencies with their lower speed should be transferred to sound. Adding an
extra and smaller cone normally does that. This solution is catastrophic, as the
loudspeaker thereby looses its minimum phase characteristic. One can - as done - forget
it or stop caring, but striving for perfection I see no way without. The possibly right way
to do this came from research on quite another matter -
The distribution of mass in the diaphragm.
When a cone-diaphragm is energised by a voice-coil, the energy is spread out in the
diaphragm to a higher and higher mass formed by the greater area and constant
thickness, and that can’t be right. The amount of energy per volume unit should be
constant. This would be fulfilled if the thickness of the material were inversely related
to the radius. Further the momentum seen from a hypothetical centre also should be
constant, so we again must have mass (thickness) inversely related to radius. The
thickness of the membrane must therefore be inversely related to r^2, for the diaphragm
to be in balance with the force inserted. You know the thickness at the rim, and from
Side 23 af 25
there you can calculate the thickness inwards. Now the question arose how to follow
that rule, when the material was equally thick. The idea of producing a sandwich of
differently sized 8 point star-shaped parts on the back or front of a Circular base, which
in all followed that rule, was executed and measured.
Figure: star-shaped cones drawn with a compass. It’s not all mathematically correct, but
it will give you an idea of its construction.
Big was the surprise, when this construction of the diaphragm showed to have extended
frequency response , despite its heavier mass. This was most satisfying for the bass-mid
units I worked with. I just often used the construction for bass and didn’t think further.
Years later a friend asked for my opinion on the possibility of making a full-tone
diaphragm to use with a Lowther magnet. I considered it possible, having the star
construction in mind.
The construction was changed to have the sandwich of three star shaped diaphragms on
the front to form a four cones unit. All four cones are equally shaped, so the speed of
sound is the same. The possibility for the high frequency to couple to the air is changed
by the many endpoints of diaphragms. It was to be used with a PM2 magnet, but that
was changed to PM6 and also a too high voicecoil was used. Despite that, its frequency
response from 20 to 20 kHz was smooth with a minor dip from voice-coil nullification
and within -10 dB at the end points- a beautiful bandpass function.
This result was marvellous from an 8-inch unit and it should be possible to scale down
to 4-inch. A friend and I are in the middle of that experiment, but lack of time together
Side 24 af 25
prolongs the process. The need for a bigger voice-coil than the calculated 20-mm
disturbed more than expected and for now we have a deep nullification at 15 kHz. These
problems will be solved, but measurements on this unit without front suspension have
directed my attention to that part. The sound of this unit was so different, so clean
compared to the same unit with front suspension, that this part simply must be silenced
one way or another. Its deteriorating effect on sound is far greater, that I’ve even dreamt
of.
I was unlucky also to work on some dome tweeters at the same time, and really it’s a
bad construction as well – even the expensive ones. So also that construction needs a
closer look.
The front suspension.
The experimental 4-inch unit with the star-diaphragm had a very linear frequency
response but with a dip around 1 kHz and nullification at 15 kHz without front
suspension. It was beyond recognition with that part attached - still the null at 15 kHz
wherever the microphone was placed. Therefore it must come from the voicecoil itself.
Many different solutions for a suspension were tried and one fell out to be most
interesting. A woollen string filling the space between diaphragm and basket left the
upper part nearly untouched. It wasn’t sufficiently airtight and silent. A mix of Merino
and Angora wool was silent, and a tube was knitted. It has the advantage that, it can be
knitted end to end, so it can’t be seen or felt. The sound is wonderful, but it still isn’t
airtight, and every attempt to get it that until now, has punished us with disturbed
frequency response. I hope it can be fixed one way or another, we just don’t know how
for now.
Shouldn’t we succeed in this, a totally new way must be found. I have an idea of a flat
suspension with a special willingness to move.
This new idea has now been tried with surprisingly good result. It has to be assembled
and for that purpose soft glue used for diving suits works fine. This suspension is at start
high but forced flat and held into that position by the gluing to the diaphragm. It has a
tendency to seek out from this placement, but hasn’t the power to move more than itself
and the glue. It seemed to be a present from Heaven. It was by measurement really a
surprise to se an untouched frequency response and at the same time have a tight
suspension.
The material for this suspension, is foam of butyl used to tighten roofs with boards of
Eternit. The adhesive side can easily be cut away with a warm thin wire to get a wanted
thickness of 1.5 to 2 mm
This solution is most interesting and far the best to my experience, but from there to get
a manufacturer even to try it is a long way, but here it is for you to try – it works. The
Side 25 af 25
diaphragm and the basket must be glued to the in- and the outside of this suspension.
A dynamic tweeter – not a dome and not a cone.
I got this idea an evening after a longer discussion about the compression in domes,
which a manufacturer had postulated his version not to have. That was a lie – of course!
But was it possible to make at all?
Speculations led in different directions. As the discussion was started on a dome
construction the thoughts was focused on that.
It should be possible to part its area to both sides of the voice coil, forming the wellknown
ring-radiator. This construction has been used for horns by JBL. But haven’t
been seen as an open construction.
This idea is presented as a sketch, but it should work and be without compression more
than its closed chamber presents and can be chosen freely.
This version is inspired from the Vifa solution, Salute to them for daring try this
other way of thinking. But shouldn’t they improve it?
Fig. Ring radiator. The light blue supports are radial plates, so freedom for air is
undisturbed.
------------- Pure Duelund www.duelundaudio.com
|
|