[JohninCR]

In the No Baffles thread we discussed the problems of circular baffles related
to edge diffraction.  While modelling predicts minimal problems with the 15" drivers
I used for the NoBaffle test structures, we agreed that circular baffles were a
real problem using smaller drivers because the response ripples resulting from edge
diffraction would be very pronounced.  I set out to prove that circular baffles can
be used and may even have a beneficial sonic effect.  I can only imagine how bad
a horn instrument such as a trumpet would sound with a rectangular bell instead
of a round one.

Typically baffle step and edge diffraction are considered synonymous, but I consider
them to be 2 separate issues that are baffle size and shape related, which occur in
the same frequency range.  Baffle step for boxed speakers is where response from
the speaker transitions from radiating only in front of the baffle (1/2 space) to wrapping
around the sides and radiating into full space, and spl drops by 6db during this transition.
A similar thing happens with open baffles, except spl decreases by 6db/oct instead of
leveling off after only a 6db drop.  Edge diffraction results in response ripples during
this transition.  The effect is more pronounced with open baffle speakers because the
pressure change is doubled due to the rear wave opposite pressures meeting at the edge.

The standard approach is rectangular baffles to spread the edge diffraction ripples across
a range of frequencies to net a flatter response.  With this approach the ripples are still
there, but they overlap each other due to the different driver to edge distances.  I believe
a better approach is to eliminate the response ripples, because then we aren't averaging
a bunch of garbage together to flatten response and we are permitted more freedom to
shape baffles however we choose.

Olsen did studies that show how the geometry and depth of the baffle edge can virtually
eliminate the edge diffraction ripples.  I used this info and made a baffle with a cirular front
profile, but a stepped edge to make the pressure change at the edge of the baffle less
abrupt than a square edged baffle and to spread the effect of the rear wave's influence
over a wider distance.  The maximum diameter is 13" near the last layer.  I compared this baffle
with an 8" B200 to a 13" diameter flat baffle and to a 17"W X 33"H flat baffle that I already had.

You definitely don't want to use a round flat baffle with this driver.  The peaks and nulls were
quite pronounced and made for a harsh sound.  While my diffraction ring baffle started rolling
off earlier than the larger rectangular baffle, once I EQ'd them to a similar overall response, I
preferred the sound of the rings.  The single point source sound was much more exact.  In
comparison the rectangle sounded like the sound came from the entire baffle instead of the
center point of the driver.  I also found the higher frequencies to be more focused and defined
with the rings, resulting in sharper imaging.  I also did some tone sweeps and the circular rings
had a smoother response than the rectangle.

I added bass augmenters to the mix to make the bass region nice and full, and despite having
to run the woofer higher for the ringed version, it retained a better single point source sound
Even the bass seemed to come from the center of the B200 vs the rectangular speaker,
where the sound seem to come from the driver and the baffle without a solid focal point. 

I realize that I'm posting this way too early, and I need to experiment with different rings and
edge shapes along with getting scientific with measurements, but I thought that maybe over
the holidays someone else may want to do some experimenting of their own, since this appears
to be an avenue for significant advancement in OB design.

Note that I taped some pieces on the sides of a very narrow boxed speaker, and it drastically
changed the character of the speaker and improved the focus of the higher frequencies.

Here's are pics of one of my "edge diffraction eliminator rings".

[JeffB]

This looks very interesting.
I could not help but wonder about turning the baffle around and making it more of a front firing horn design.
I don't know if there are technical differences between a horn, a waveguide, or simply an attempt to smooth edge diffraction.
I don't know if it would be best to fire forward or backward.
I saw a similar picture the other day.
http://img62.imageshack.us/my.php?image=schssel04small9ix.jpg

This above picture came from the following discussion.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=82116&perpage=10&pagenumber=4
There are a couple more pictures at this link.
See the post by Calvin, page 4, eight posts down.

giorgino1, who posts in the The hornshoppe forum mounted wooden salad bowls from Ikea over the hornshoppe horn dirvers.
I could not find a picture any more, but they look much like JohninCR bowls but in reverse.

[JeffB]

What about putting the driver in the middle of a donut?
I like this idea.

[scorpion]

Harry Ferdinand Olson (1957) ?

/Erling

[JohninCR]

Jeff,

I saw those circular front waveguide baffles the other day.  Beautiful
but they ignore edge diffraction other than since they're quite large,
the frequencies involved are much lower, where ripples aren't as offensive.
I don't know anything about waveguide design, otherwise I'd try a front
and rear waveguide with a nice thick rounded edge to equally load the
front and rear of the driver and address diffraction at the edges.  Reversed
salad bowls might work, but I've found with winged type designs that a
a straight or increasing expansion of the rear wave retains it's coherency best.

For construction I initially thought about one of those inflatable life rings
for kids, and cover it with fiberglass and resin, along the lines of your donut
idea.  Instead I just used some driver cutouts I had lying around.

[JohninCR]

Harry Ferdinand Olson (1957) ?

/Erling

Yes, thanks for the correction.  pg 23 of Olson's Acoustics shows shapes and results.
The couple of quick tests that I did, showed me that there are substantial gains without
using anything near the dimensions that Olson used.  In fact, with smaller dimensioned
cabs I believe the importance increases due to the higher frequencies involved.  Olson
worked in the old days of large speakers.

As a quick test for effect, I taped 35mm thick pieces of wood with the leading and
trailing edges cut at 45 degrees to my 6.5" wide metronome cabs, so the driver would
see a double truncated pyramid shape for a baffle.  The change in sonics was staggering.

I'll never build another cab or baffle that doesn't take this into consideration.  I believe
designers have just taken the lazy way out and use narrow cabs with 90 degree edges,
so they can rely on comb filtering to help correct their baffle step issue.  To some extent
I understand, because I need to disturb the clean lines of my metronomes, and then I
would need to add BSC because the comb filtering is so significant it negates the need
for BSC.  The effort would be rewarded with better sound though.

[Davey]

John,

That's still a circular baffle.....just bent back.  Frequencies of appropriate wavelengths won't even recognize it's "stepped."

Have you performed any measurements at all or is this strictly a subjective assessment?  I think you're still going to see a strong peak at a frequency where the front/back distance is a multiple of the wavelength.

Cheers,

Davey.

[corloc]

Wouldn't you get horn loading from the rear?  The B200 puts out a lot of higher frequencies from the rear.  Or are you Eq'ing that out?  I'm curious, because I'm designing a u-baffle that would be 4" radius in front with a  normal U in back.

http://picasaweb.google.com/sbweber/Audio02/photo#5004895756463471426

[JohninCR]

John,

That's still a circular baffle.....just bent back.  Frequencies of appropriate wavelengths won't even recognize it's "stepped."

Have you performed any measurements at all or is this strictly a subjective assessment?  I think you're still going to see a strong peak at a frequency where the front/back distance is a multiple of the wavelength.

Cheers,

Davey.

Davey,

No measurements yet.  I need to learn how to do proper measurements first.  I did slow, controlled sweeps and obtained a pretty good idea of what they will show.  It was the same methodology that I used with the 15" driver when I came very close to where the "experts" told me afterward that I would find 2db ripples.

Yes, the small steps on the rings are meaningless.  I did some sanding just to get rid of the sharp edges.

I'm not sure what you mean by front/back distance, but if you are talking about front of the cone to the rear of the cone, that is a common misconception.  If you're talking about the rear wave delay caused by the baffle, the correct dimension is rear of the cone to the baffle plane.  That is the added travel distance for the rear wave.  In this case there are 3 primary dimensions-  The diameter of the flat portion of the front, which determines the pure 1/2 space launch transition point (9.5").  The largest diameter, which is where the wave sees full space (13").  And The added travel distance for the rear wave (approx 9").

While the 13" diameter flat baffle performed as you expect, with peaks and sharp nulls,  the angled edge baffle was clearly different.  By sweeping the edges back, the dipole roll-off point becomes like an 18" baffle.  Offsetting this is baffle step type behavior starting at a point similar to a 9.5" diameter baffle, but because the pressure release is gradual due to the angled sides, the effect is more gradual.  The idea seems to have worked as planned with a smoother overall response, without the comb filtering resulting from a flat rectangular baffle.

Corloc brought up my only real concern going in, potentially horn loading the rear wave.  The cone has to be getting some differential in loading due to the initial constrainment of the rear wave.  Impedance measurements should help quantify that.  I compared the rear waves as well and I couldn't identify a significant difference in the sound although there was some increase in directivity.


Corloc,

Regarding your planned baffle.  I hope you're doing a test baffle first.  I'd caution you against a straight pipe or squared channel in the back.  Plan on tweaking and adjusting, because folding back the sides for fullrangers is time consuming to get right.  I believe it's worth the effort because I believe it is superior to a flat baffle from an aesthetic, as well as sonic, standpoint.  I find it helpful to visualize what's happening with the front and rear waves.    I'll be most interested in pics as you progress.

[Davey]

John,

"A common misconception?"  No, I think not.  You can define the front/back distance however you want, but my point is that it is a factor and will contribute to the overall response.  This is a measurable effect that is easily visible with "normal" baffle dimensions like we're discussing here.  (Obviously, very large baffles will "shift" this effect out of the frequencies of interest and you'll end up with an infinite baffle or something approaching that.)

With your prototype (since it is relatively small) at some frequency X the rear output will come around to the front and add in phase with the front output and generate a peak on-axis.  It will also add at off-axis angles at different frequencies to create a total polar response that is a result of all these factors......baffle shape/size, driver shape/size, driver placement on the baffle, etc, etc.  This should easily be measurable.  If the peak is such that it can be equalized on-axis and also maintain off-axis responses that are well controlled then you might have something viable.

A circular baffle means the front/back distance is constant and for a rectangular baffle it's not.  The result of that change in construction is a significant factor.

All of this will become apparent when you learn how to perform measurements and actually do so.  Speculation is fine, but until you start gathering some empirical data to prove your theories they're just speculation.

Cheers,

Davey.

[konut]

What about putting the driver in the middle of a donut?
I like this idea.

John, when you first proposed the idea of a diffraction ring, I envisaged a donut with the diameter of the donut widening and narrowing to address the standing wave issues. When looking at the device staight on, it would outline an ovoid shape. Is my description clear enough?

[johnk...]

Just lurking here. As you extend the rings back you will introduce a cavity resonance that won't necessarily be noticeable in the on axis response (it may be, depending on the frequency and driver directionality characteristics) which if left undamped may color the sound. The cavity resonance will also make the front and rear response more asymmetric altering the way the front and back sum compared to a flat, circular baffle with similar path length delay.

I would also say that the situation is not really the same as diffraction (though it looks similar). Given that on axis, diffraction in a typical sealed back speaker is the sum of the direct response with out of phase reflections form the edges, the dipole on a circular baffle is the sum of the direct response, out of phase edge diffraction associated with the front response, the delayed rear response, and edge diffraction from the rear response. Over the frequency range were the front and rear radiation are symmetric the edge diffraction from the front and cancels and what you are left with is the summed front and delayed rear response.

As Davey said, you really need to look at some measurement, and not just the on axis response. AT low frequency the response will be similar to a circular baffle but as the frequency rises the response will be significantly different. On axis the response may look better or worse, but you will also need to measure the response off axis and at 180 degrees to see what is really happening.




John k...

Music and Design

[JohninCR]

JohnK,

Thanks for stopping in.  Yes, I definitely need measurements.  Hopefully I can do them
well enough to explain my subjective results.  My intention is to create compact form
OB's that are sonically superior to their flat baffle counterparts.

Regarding cavity resonances, horn loading, etc on the backside, I'm not sure how to
predict them, since it's not as easy as knowing the 1/4 resonances and standing wave
frequencies of a rectangular U baffle.  My hope is that with only a 3.5" depth the low
pass filter of the driver motor assembly would keep the rear output from reaching a
problematic range.

Over the frequency range were the front and rear radiation are symmetric the edge diffraction from the front and cancels and what you are left with is the summed front and delayed rear response.

Is the cancellation of front & rear edge diffraction with symmetrical dipoles something
you have measured?  My thought was that they are reinforced instead of cancelled.  If all
we are left with is the sum of the front and delayed rear response, then wouldn't we just
have a nice smooth response with only the 6db/oct roll-off to contend with?  Since the
dipole roll-off point corresponds with the transition from 1/2 space radiation to full space,
aren't the higher frequency peaks and nulls only applicable to folded dipoles?  If the front
and rear are only each radiating into half space, then excluding reflections they won't
sum.

I want to learn more about actual diffraction at the baffle edges.  You call it out of phase
reflections occuring there.  Linkwitz mentions comb filtering.  Neither helps me visualize what
is going on there.  In another thread, Rudolph says that it's only about pressure change at
the baffle edge, but I'm not sure that explanation is complete either, since there is movement
at the edge.  I believe with an open alignment there must be more movement at the edge,
making edge geometry more important.

[johnk...]

(JPK) All this open baffle stuff  is extremely simple to explain, until real world effest enter the picture. You start with two perfectly flat source separated by some distance, operating out of phase. You end up with the classic dipole response. A peak of +6dB at F = C/(2d) where d is the separation and C is the sound speed. Below F the response quickly begins to roll off smoothly at 6dB/octave. Above F there are additional peaks spaced by 1 octave and nulls between them This is the combfiltering. Such a dipole would only be useful to slightly above the peak. Now we move to a more normal situation, a real driver will become directional at frequencies with wave lengths greater than the driver circumference. If we ignore the asymmetries in the front and rear radiation due to other sources (motor structure, inverted cone shape, etc, and we separate the sources by a distance such that the first response peak occurs at or above the point where the driver becomes directional, then we will still get our smooth roll off below the peak and above the peak, since the rear source will become directional and only radiate into the rear hemisphere, we won't see as deep a first null and the combfiltering will become less and less. Rather than two separate sources we typically use a single source in an H frame for low frequency or a flat baffle mids. The separateion is the either the length of the H or 1/2 the baffle diameter for a circular baffle.


Considering edge diffraction, even with directional driver, provided the response is symmetrical front and rear, and assuming the edge of the baffle is thin, or we are at a distance many time the separation away fromthe baffle, then, even in the frequency range where the driver is directional the same pressure signal will reach the baffle edge from the front and rear. There at then 4 components  which sum form the on axis response. 1) the direct response from the front source. 2) the delayed direct response form the rear source, 3) the diffraction of the front direct source from the baffle edge, and 4) the diffraction of the rear  direct sound from the baffle edge. 1 and 2 are identical (for symmetric front and rear radiation) but 2 is 180 degrees plus the separation delay out of phase the the front direct radiation. This sums to the dipole response. Now conside 3 and 4. For a hypothetically infinitely thin baffle, 3 and 4 will be identical in amplitude but 180 degrees out of phase because they both represent the diffraction of the same signal (symmetric front and rear responses) and are both delayed by the propagation time to the baffle edge. Thus they cancel at the baffle edge and do not contribute to the on axis response.  Thus, if the front and rear response are symmetric, the diffraction from the baffle edges will always cancel and what we end up with is the sum of the front plus the out  of phase plus delayed rear response. Since the direct response becomes more and more directional the final result is a smooth 6dB/octave rise below the dipole peak and a transition to the direct radiation of only the front response as the directionality becomes the major factor. How that transition occurs depends of the baffle diameter (for a circular baffle) or where the dipole peak occurs relative to where the driver become directional. The trade off is reduced low frequency response against smoother high frequency response as the baffle gets smaller.

Of course the caveat is that real drivers don't have symmetric front and rear response and as the frequency rises the front and rear response becomes more asymmetric. What this means is two fold. Once we get into the frequency range where the response is asymmetrical the edge diffraction is no longer identical and 180 degrees out of phase, so there will be an effect of edge diffraction on the on axis response. The second factor is that if the baffle is made smaller such that the dipole peak is above the frequency where the driver is not symmetric front and rear, then these asymmetries will begin to have an affect on the on axis response below the dipole peak and you won't have as smooth a dipole roll off as  you would if the baffle diameter is somewhat larger.

It should also be mentioned that the driver cone itself is in effect a baffle for the dipole. I think you will find that your diffraction rings will behave very similar to a circular baffle at lower frequencies where the effective radius of the baffle would be close to the driver  radius plus the depth of the diffraction ring. At higher frequency you will find you depart from the dipole response, particualrly at 90 degrees off axis, and you may have some caviety resonances to deal with. Since the picture I saw looked a lot like a small U-frame you may want to look at http://www.musicanddesign.com/NaO-II-U-frame.html I discuss U-frame damping there as it applies to a U-frame woofer, but the same ideas sshould apply to smaller systems as well.

[JohninCR]

Interesting stuff JohnK.  Have you done measurements with real baffles that reflect the cancellation of front/rear diffraction.  We see more and more use of narrow flat baffles for the mids combined with different types of bass augmentation, and I question these small dimensioned baffles without addressing edge diffraction.  I still don't have a firm handle on edge diffraction, so I could be way off base, but to me diffraction effects would be quite directional, preventing the opportunity for cancellation.  I've used folded sides, and I hear significant benefits from large radius roundovers on the front edges, and your explanation helps me understand that better.  You make me wonder though whether a flat thin construction is better. 

[johnk...]

I don't think there is any question that edge diffraction for a flat baffle dipole system cancels between the front and rear as long as the baffle is relatively thin compared to the wave lengths in question. At higher frequency there is some evidence (or analysis) that diffractions sources are not omnidirectional. The bigger problem with very narrow baffles, say less that twice the driver's effective diameter, is as I said, the dipole peak frequency is above the point where the response is symmetric between front and rear. Thus the diffraction effects will be present to some degree below the peak frequency. When the driver looses front-rear symmetry, diffraction becomes basically the same problem it is for a boxed speaker.

Adding wings creats a very different problem because the diffraction sources for the front and rear are not the same anymore. I was reading the thread on Open top U-frames and made some quick measurements comparing the rear radiation from a flat baffle dipole (11" wide) and one with 8.5" deep side wings, straight back. These are pretty deep wings, but the point is that once you start adding wings things become more complex, for better or worse.  Green is rear w/o wings, red with wings.


.

[JohninCR]

Thanks again John.  Can you help me get my brain around edge diffraction and how it relates to
dipoles?  I'm referring to the ragged response through the baffle step transition demonstrated in
Olson's work regarding monopole cabinet shapes.  In the frequency range where sound starts to
wrap around the baffle, the abrupt transition at the edge of typical speakers with 90 degree sides
creates a ragged response.  I visualize this resulting from the rapid back and forth motion across
this relatively sharp edge and the corresponding pressure changes, creating what you referred to
as "reflections" there.

Now in the case of a flat baffle dipole, as a compression wraps around the edge it meets a corresponding
rarefaction from rear.  To me, it makes sense that this will accelerate or increase movement at the edge
and increase the problem, not create equal and opposite problems travelling in the same direction resulting
in cancellation.  What am I missing?

[Rudolf]

(JPK) ... There at then 4 components  which sum form the on axis response. 1) the direct response from the front source. 2) the delayed direct response form the rear source, 3) the diffraction of the front direct source from the baffle edge, and 4) the diffraction of the rear  direct sound from the baffle edge. ... Now consider 3 and 4. For a hypothetically infinitely thin baffle, 3 and 4 will be identical in amplitude but 180 degrees out of phase because they both represent the diffraction of the same signal (symmetric front and rear responses) and are both delayed by the propagation time to the baffle edge. Thus they cancel at the baffle edge and do not contribute to the on axis response.

johnk,
regarding edge diffraction cancelling out at the baffle edge - this would be too nice to believe. While the components 1 and 2 are travelling in the same direction (along the baffle plane) components 3 and 4 are travelling in opposite directions (perpendicular to the baffle plane). You can´t think in amplitude and phase only but have to consider the impulse vector too IMHO. Otherwise there would hardly be anything like a reverberation time in rooms, since every reflected wave would cancel with the next oncoming wave in next to no time.

So things (sadly) stay as described by SL http://www.linkwitzlab.com/faq.htm#Q8
 

Rudolf
www.dipolplus.de

[JoshK]

Where did I see it demostrated that an open baffle is the most extreme case of diffraction at the edges for a given baffle shape/size?  Might have been Mark's speaker measurement site.

[JohninCR]

Rudolph,

Thanks for making me read thru SL's FAQ#8 a few more times.  I think it helps me understand why OB's
sound so different, with the rear wave arriving at the same time as the wave resulting from the pressure
change as the front wave reaches the baffle edge.  I also now understand the "reflections" at the baffle
edge, which are actually a new sonic event originating at this point of the pressure change.  Again, this is
all a baffle step type of view and doesn't explain the ragged response in this transition region demonstrated
quite clearly by Olson.  There is more going on at the edge (turbulence from the oscillating change in pressure
or something of that nature) that I am trying to bring into discussion when I used the term "edge diffraction".