Open Baffle Vibrations

[The Prof]

Hi OB guys n gals,

First off, let me say that I am not here to shout what everyone else is doing is NOT right. Rather, the attempt is to try and address the problem of baffle vibration from another practical angle.

An examination of current practices, amateur and pro, shows the drivers mounted onto the baffle, in this case a thick plank of mostly wood, sometimes matrix, concrete etc. Gravity is pulling at the magnet structure and even small mid or HF units will load the baffle with tension. Feed in a signal and add some vibration and we have a sure formula for (unwanted) trouble.

What's the alternative?
Let us have a strong, slim 'backbone' of wood or steel or other material. Attach the drivers by their magnet structures to this spine, of course, keeping in mind the offsets needed to keep their acoustic centres in line--so far as possible. See that the fixture profiles are slim so that they do not mar the rear radiation.

This structure has an added advantage: the weights of the drivers mostly act straight down and so keep the 'spine' (with a substantial base--make it an open tray and fill it with some concrete) balanced mechanically.

Now take baffles (you can have the baffles for the LF, MF and the HF units cut into separate, but 'matched-to-each-other' pieces) and cut the requisite holes for front radiation, with needed chamfering etc. Now obtain some round, tube-like rubber auto door trim (which often has a flat edging also), and fabricate round gaskets for each driver. Judicious application of a sharp craft knife and just one drop of super-glue and a steady hand is known to work miracles. Stick the gaskets to the driver baskets with soft-drying rubber-based glue.

Work out a method of mounting the baffles onto the 'spine' in such a way that by tightening the mounting screw you can adjust how far the gaskets get compressed. DO NOT over-tighten; you are looking for just an air-seal. Be sure to maintain say a small spacing between the individual baffles, say a quarter inch. Stick thick rubber strips to 'link' the baffles and ensure an 'acoustic continuity'.

Mounting the baffles onto the spine is important. And you can device and use 'shock mounts' to minimize the transference of vibrations from the spine onto the baffle. It is easier than stopping the baffles being vibrated with a vengeance by an active driver basket.

Let me know how this has worked for you. This is the only way I know by which you can tame the 'vibes'. You can reach me at ukpanickar@yahoo.co.uk, but be sure to put the subject line 'vibes'. Happy vibes to all OBers.

[LineArray]

Originally Posted by The Prof

...

What's the alternative?
Let us have a strong, slim 'backbone' of wood or steel or other material. Attach the drivers by their magnet structures to this spine, of course, keeping in mind the offsets needed to keep their acoustic centres in line--so far as possible. See that the fixture profiles are slim so that they do not mar the rear radiation.
...

Agreed.

[The Prof]

LineArray,

I like your strong, slim 'spine'. But from the pic, the mounting details cant be seen.

My take is that floating the baffle independently of the drivers is going to give you great dividends. Just try it on one channel and hear the difference!

A few days back I read some complaints about the baffle vibrations, but I was not watching which thread it was. Let us a have a real re-think on the open baffle. Great vibes to all.

[LineArray]

You can view some more pics on the german
part of the website

[The Prof]

LineArray,

Checked out the German area and had a look at the pics. Your idea is good, as natural stone (marble, granite etc) or concrete is the way to go if u want to deaden vibrations. It was my choice also on many occasions. But I guess it is a brute-force method. Design elegance should try to avoid any force, good or bad, I think ;-)

Also, your method of driver mounting is not clear from the photos. By the way, what have you done to overcome the driver interaction artefacts in your large line array? I am curious...

If you are using a set of drivers, large to small, then it is desirable to taper the 'spine' towards the top so that not much occlusion of the rear radiation occurs. A sandwich of stone and wood would give you some mounting ease, and rigidity.

Driver mounting is all important, I guess. Try holding any driver by its front basket and imagine how difficult it would be to counter vibrations, what with the bending moments caused by driver weight acting in one direction--down--compounding the problem. Now try holding it by the magnet structure and see the dramatic change. We can borrow from mechanical engineering and 'shock mount' the driver to the spine easily. Stick the highly pliant tube gaskets to the drivers and mount the baffles as precisely that-- 'baffle' only, and not a 'sounding board'. Compressing the gasket to provide just an air seal is enough. Also, since the baffle/s is/are mounted to the spine, there is no transference of vibrations to it directly from the driver. Here also, with some ingenuity, 'shock mounting' techniques for the baffle can be adopted.

I would urge you to modify one channel so that the drivers are supported mainly at the magnet, and the baffle is 'floating'. Your ears will be 'opened' .... :-)

Happy vibes!!

[LineArray]

Hello,

Originally Posted by The Prof

...
Your idea is good, as natural stone (marble, granite etc) or concrete is the way to go if u want to deaden vibrations.
It was my choice also on many occasions.
But I guess it is a brute-force method.
Design elegance should try to avoid any force, good or bad, I think ;-)
...

Both the marble "spine" as an inertial vibration damper and the baffle are no
massive structures but two layered sandwiches utilizing a special sealant as
a contraint layer.

The construction does not rely on the properties of a single material solely.

The spine and the rigidly spine-mounted drivers , make up a standing "Pendulum"
of defined subsonic resonance frequency and damping, which is achieved by a
bearing in the foot of the speaker decoupling spine and baffle from the foot.

That bearing/damping is mainly made from cork and is fairly invisible.

Originally Posted by The Prof

...
Also, your method of driver mounting is not clear from the photos.
...

The pics on the site are no documentation of mounting ...
but see below.

Originally Posted by The Prof

...
By the way, what have you done to overcome the driver interaction artefacts
in your large line array? I am curious...
...

The construction you see is an answer to that question, the question has been
asked by myself before construction.

With rising frequency the output power is shifted towards the upper
- non equidistant - driver trio.

Interference patterns vary very smoothly with frequency and/or the listeners
vertical position, movement horizontally is even less a problem.

Subjectively very little change is observed when moving your head
vertically. The arrangement is superior to an equidistant arrangement and
has in fact much less interference problems than common 2-ways.

I know that this may be (very) hard to believe until you do some
simulations and finally listening tests and measurements which such
an arrangement: The driver distances have been optimized excessively.

It is not even peaking - comb filtering - if you put the microphone
on the bottom in 1m distance to the baffle ... which is not a typical
listening position.

There is of course a loss in brillance if you listen sitting on the
bottom and too near.

But the high quality listening area ist quite large with that speaker.

Originally Posted by The Prof

...
I would urge you to modify one channel so that the drivers are supported mainly
at the magnet, and the baffle is 'floating'. Your ears will be 'opened' .... :-)
...

I am 44 Years of age now, but i guess up to now my ears are quite open still ...

The baffle is attached with only one of its constituent layers to the baskets
using damping sealant and grummets.

The second - rear side - layer of the baffle is "floating".
This way 3 things things are achieved:

1) High damping in the frequency region, where the baffle could
resonate from being accelerated by "rest forces" acting through
the baskets (Though the system is very quiet at mid to high frequencies).

2) Utilizing the stiffness of the spine to have a rather thin
(important in this concept) baffle, which is nevertheless rigid
(by beeing suported) enough to stay calm due to excitation by
low frequency sound pressure.

3) Having a baffle of less than 1/2 thickness near the rear
part of the basket, which yields maximum openness for
the rear basket side by having staggered driver cutouts in both
baffle layers.

To summarize: Since i have this speaker as a reference my own
standards of what is possible especially in non resonant bass to
lower midrange region, which is a problem with many cabinets
or open baffles, have moved into regions which are
not achievable with standard construction methods.

Concerning "clean" bass to mid region, this speaker truly opens
the ears of almost any listener ...

The listening room is the weakest part concerning modal
behavior in the bass region.

Kind Regards

[LineArray]

btw. the spacing of the upper 3-4 drivers was inspired by
"golomb ruler" spacings, which have an application e.g.
for the positioning of antennas in an array used for
radio astronomy. At least this was the starting point.

In a golomb ruler non repeating distances are used
for all marks (objects). The distances used for the upper
drivers in the "dipol 08" array are unique and as an additional
property mostly prime to each other.

Concerning bass output an equidistant spacing like the
larger distances of the bottom drivers would have a slight
advantage, but it would also exhibit serious comb filtering.

A correction network helps padding down the equidistant part
of the array, when entering the frequency range where comb
filtering would show up in an ugly manner.

The plot attached shows FR on the BOTTOM in front of the baffle
in approx 1m distance. Since the height of the reference axis of the
speaker is betweeen the 2. and the 3. driver from above, this means
the measurement is from ~45 degrees of axis vertically ...
The plot is unsmoothed. Above 500Hz the level is reduced due to
the directivity of the array in vertical direction, but the typical
"line array comb filtering" does not occur : It occurs neither on- nor off axis.

For those interested in golomb rulers:

Golomb ruler - Wikipedia, the free encyclopedia

[The Prof]

Line Array,

Thanks for that detailed reply. I can see that you have done excellent work. I am 57 and it looks like my ears have been opened! The appearance of your OB led me to believe that the drivers were bolted to the baffle, and surely even with gaskets and grommets, it would excite some unnecessary and noticeable vibes, and hence my humble suggestion for some mods. But your 'sandwich' technique has licked it I guess.

I noticed the unequal spacing of the drivers, but didnt know the methodology of your approach. Your expertise and knowledge in other fields (Golomb etc) has indeed come in useful. I am, like many ought to be, curious of the way you solved the interaction artefact problem. Do publish a more detailed approach so that all of us can benefit further.

Are you epoxying the driver magnet structures to the spine, or is that a kind of holder for the magnet?

All said and done, from the excellent results you have achieved by now, it is plain that a more detailed presentation is going to benefit the fraternity, I am sure. Especially as we are keen about your crossover techniques. And what is your experience about the rear radiation, its desirability, level, frequency domain etc. Looking for a more detailed presentation from you,

Sincerely, as always.

[LineArray]

Hello Prof,

yes, magnets a are glued directly to the stone using a mix which
stays flexible a bit ...


To provide a "generalized" aproach or algorithm for the design is not that easy, because
it is a strategy mix and the actual cocktail depends on what you want to achieve.

The "strategy mix" used here has nothing new under the sun concerning its ingredients,
but the combination of strategies may be a bit uncommon.

Weighted line arrays are already common, and this one falls for sure into that category.

The weighting of drivers in a line array can be by power distribution between the drivers
or by distance between the drivers used. The "acoustical weight" rises where the drivers
are mounted closer or are supplied with input power increased relatively compared to their
neighbour drivers.

---
The first idea for "dipol 08" was to combine a frequency dependant power weighting and
weighting by distance, to make the array shorter with rising frequency more effectively.
This also helps in getting along with low order filters and smooth phase shifts while
doing so.

Dipol 08 is not meant to be a virtually infinite line source and is not a
"nearfield line array". I mention this to circumvent the "this is not a line array"
discussion which had evolved sometimes ...

Since the driver distance cannot be made frequency dependent, there is a "pre weighting"
by distance. Of course the relative to wavelength driver distance varies with frequency.
So up to some hundred Hz there is no real difference between dipol 08 and a fully
occupied dipole array of same height in behavior concerning directional radiation.

---
The second idea is to have the power weighting in a way that you get "nested" line arrays
which shrink in size (and number of drivers) as frequency rises.

In this case we start with all 6 drivers in the bass and end up with 3 in the highs.
The design can be consequently altered to reduce the array to a single driver in the
brillance region. Since the dipol 08 crossover network is a passive one currently, this has
proven impractical, the drawbacks are larger than its benefits.
(KISS keep it stupid simple ...)

Using an active configuration with each driver having its own power amp, reduction to a single
driver could/would be a consequent step. Then the reduction of drivers with frequency would go
like this (driver numbered from top to bottom):

654321
...321
....2.

The second driver from above would be the acoustic center for high frequencies.
De facto the current approach is already very close to that. The sound is subjectively
centered somewhat below the second driver from above, the array does not sound "smeared"
along its height:

654321
...321

Power tapering or "shading" is also not new entirely, the usual approch is to pad some of the
outer drivers down with rising frequency, so the array shrinks at its ends with frequency rising.

Since i wanted the acoustic center at ear height, i chose to get rid of the lower drivers as
frequency rises. Also there is not only downpadding, but the voltage transferred to the
upper drivers also rises above ~4Khz. This is to compensate for the missing contribution of the
downpadded drivers on the one hand, but also for compensation of the non sufficient energy radiated
in the highs from the fullrange drivers themselves.

I think you see, how things interact: The crossover network compensates also the driver
and has to be aligned for the particular driver used, as things are for every other loudspeaker
concept.

---
The third idea in the cocktail can be fuzzily described as follows:

When the driver distances come into the order of wavelength i want "golomb like"
spacings to "soften" the interference patterns. So the nested array active at a certain
frequency should have a "golomb ruler similar" spacing.

This is the idea behind it, not the real object which resulted during development.

As i said golomb rulers were a starting point in using unique distances.
For the upper 3 drivers a golomb ruler of 3. order would have the distances

0-1-3 position on ruler (marks)
1 2 3 driver number

for the drivers 1,2,3 when counting from the top.
Hence the distances between the drivers are

dr1 dr2: distance 1
dr1 dr3: distance 3
dr2 dr3: distance 2

A golomb spacing ensures unique distances for acoustic centers of each driver pair.
Broadly speaking: If two drivers interfere destructively at a certain listening
position, the third one will probably fill the notch.

But as the distances have common factors, there are still interferences causing peaks
and deep notches under certain angles although things are already better than using
equidistant spacings. *

To get a first solution, the arrangement was optimized using a software simulating frequency
response for different angles. Optimization was aimed at having the frequency response
varying as smoothly as possible in the range of vertical angles relevant for sitting and
standing persons at desired listening positions.

The minimum listening distance assumed will determine the range of angles, which are in the
main focus of optimization. I chose about 1.5 meters as far as i remember. The driver positions
have been varied randomly but keeping tuple distances unique.

First the whole array was undergoing optimization - it would also work rather well
without frequency dependant power tapering - then the "nested" array was "fine optimized"
for the range above ~2 Khz, since the uper 3 drivers dominate the presence to
brillance region.

Furthermore the 12 drivers have been selected according to serial tolerances:
Every driver knows his individual place in the array. A driver with some more
irregularities in the top end - or a rather "dull" candidate - qualifies as
"bass" not as "tenor" ... it is that simple.

Left and right speaker have to be made as similar as possible, this was also taken into
account for selection.

This was some effort, but i am convinced it made things a lot easier in the end, and
contributed to a - rather - simple correction network/crossover, because the
charcteristics of individual drivers were used to balance each other to some extent.
But this is really going into fine detail.

---
As i started out i could not say whether this approach would result in a homogenously sounding
speaker, although i was confident because it was not the first time for me going into that
direction.

Many listeners do the following test with that speaker intuitively without
being asked to do do it (its amazing!):

They stand up at their listening position and start to "knee bend" ...
up and down slowly - sometimes a bit faster - and the usual comment is:
"There is no change!" This is not what is usually expected.

In fact there is change ! It is fun watching the spectrum slowly
wiggling while leading a microphone over different vertical angles.

But it wiggles softly enough, it is sufficiently good for being
undetected subjectively while just listen to music.

Also knee bending is allowed (at home) but not necessary while listening ...

Like in every speaker i would do some changes for the next time.

The height of the reference axis is little bit too high, so small persons
- especially when sitting close on a low sofa - have a slight drop at the
top end.

They mostly do not burden, but i know its there. Usually i tilt the speakers
very slightly forward for them, then it is fixed.

Maybe the next time i will downsize it a little bit ("Lady Edition") ...


Kind Regards


-----------------------

* There is a rule of thumb also mentioned in a paper on nearfield line arrays
by Jim Griffin, that the coverage of the line array with radiating
surface should be at least 80 percent in an equidistant array to resemble
an ideal line source in a sufficient manner.

Unfortunately the paper does not seem to be available on the net currently:

Design Guidelines for Practical Near
Field Line Arrays
James R. Griffin, Ph.D.


It is obvious, that using fullrange drivers 80% occupation cannot be
achieved for high frequencies, even when mounted basket to basket or
slightly "overlapped".

This is because the radiating area of the cone breaks up and narrows
effectively with frequency rising. To approach the needed coverage
of the area, typical nearfield line arrays use a separate tweeter
line source.

My goal was having fullrange drivers solely and no crossover frequency
like present in a 2-way line array, which is likely to cause
irregularities in the horizontal plane ...

In the design patterns introduced by the US architect
Christopher Alexander

Christopher Alexander ? Wikipedia

it is mentioned as common solution, that a design constraint that cannot be met,
should simply be dropped. This often/mostly yields better results than trying
to meet the constraint in a half-hearted way.

With "Dipol 08" i did not even try to achieve that occupation of area, resulting
in a reduced number of drivers needed for the height of the array (about 1/2).

In turn also the "large number of cheap drivers" conflict is avoided governing
the cost of a conventional line array. The idea is to invest the budged saved
into better drivers if possible.

[LineArray]

There are two further aspects, which should be mentioned:

When moving down with frequency, the drivers grow closer together acoustically
due to distance getting smaller compared to wavelength.

That compensates partly the falling radiation resistance with frequency
due to the open baffle design.

There is still need for some compensation, but much less is needed than usually
would be necessary for such a narrow baffle.

The lower cutoff is somewhere between 65 and 80 Hz according to the room
and position. I use the speaker as satellites with a mono subwoofer,
high passing them at 80Hz 1. Order.

They have no "dipole peak" in the upper bass which would need compensation.

During design the bottom reflections were taken into account by
"mirroring" the array at the bottom:

There is a second "dipol 08" under the floor, as an acoustic mirror image,
you just don't see it.

Optimization for low frequencies took the mirror image into account
while varying the driver spacings.

Kind Regards