I'm preparing to build your "cottonwood" live edge speakers, (with the beryllium dome tweeters) which will become my "ultimate open baffle" speakers. My pair of butternut slabs are now standing in the living room where they will acclimate & stabilize over the winter months. Question: does the DSP compensate for ringing resonances from the parallel backside baffle supports? is that inconsequential, or would they benefit from absorption material on one side?
The Bitches Brew Open Baffle Live Edge Speakers
Stunning Imaging + 120dB Dynamic Range + Natural Wood Finish. Named after Miles Davis’ classic 1970 album.
I’ve been designing speakers for almost 40 years as a hobbyist, including a 3 year stint at Jensen’s OEM automotive division, where I designed speakers professionally for Honda, Mazda, Chrysler and Acura. I’ve built just about every type of speaker and enclosure type that exists except electrostatics.
This system embodies the best of what I know. An “ultimate design” which also aims for simplicity and elegance. 15” 3-way system with 2x SB Acoustics 15OB350 woofers and 1x B&C 15CXN88 constant directivity horn coaxial.
These are made from book matched slabs of spalted birch. I picked out the wood at Schroeder Hardwoods in Harvey Illinois, and Seth Cothron of @studio38designs performed the carpentry. Spalting is when fungi has grown in the wood, and in right proportions dramatically emphasizes the grain.
If you’re willing to spend $25,000 to $100,000 retail you can get some pretty extraordinary speakers. Nevertheless at about $4500 including all parts and carpentry fees, these speakers perform feats no commercial speaker I’ve seen of can do. These put some of the most respected designs in the world to shame.
Achieving flat frequency response is of course important. But just accomplishing that is the junior varsity team. Today with modern crossover design software and DSP, anyone who can’t achieve that isn’t trying. What separates the men from the boys in speaker design is how many other priorities you can simultaneously achieve.
When I have friends over they are in awe at the transparency, realism, impact and imaging. They have the accuracy of a KEF or B&W, the dynamic range of a Klipsch, the impulse and phase response of a Duntech, the transparency of planars and imaging better than all of the above. Most people have never heard sound like this.
Frequency response in real room, mostly at listener positions, averaged across both speakers and 16 locations:
I chose Open Baffle because nearly all “box” designs sound… boxy. Most people have never heard a speaker that doesn’t sound boxy! But once you have, it’s like experiencing “real” Chinese food or Italian food for the first time. After the real thing, you’ll never again be satisfied with Panda Express or Olive Garden. Same thing here: they have a huge sound stage, an open, effortless, spacious image and a palpability that must be experienced.
Open Baffle speakers have almost no radiation to the side and that gives you many interesting possibilities in room placement.
Most people have never heard a speaker that has “dynamic range to burn” - i.e. can easily exceed 120dB even if you never turn the volume above 90. Most people don’t realize that even the most expensive and reputable speakers put out bucket loads of distortion at many frequencies and really don’t have much dynamic range.
You probably haven’t heard drums until you’ve heard them this way. Bill Duddleston of Legacy Audio says, “If you can see the cone moving, it’s distorting.” With 3 15” cone surfaces, you have to get well past 90dB before the cones move visibly. Distortion is incredibly low at any reasonable volume level so compression is negligible.
High efficiency: They’ll play as loud as most people want on a flea-watt amps. My amps are 60 watts per channel (bi-amped, 4 channels total) and I can’t drive them into clipping with any of my albums at any sane listening level.
This design simultaneously prioritizes additional things:
Constant Directivity across the entire audio spectrum. This provides amazing precise imaging everywhere in the room - even when you’re literally standing right next to one speaker! You can walk all around the room and the stereo image is stable. You’re not confined to one sweet spot, and in fact one of my favorite things is to invite a half dozen friends over, sit them all around the room and listen to an album together. There is not a bad seat in the house.
Most people have never heard a speaker that can do this. Constant directivity says that the speaker gives flat frequency response on axis but also as you go off axis, the level drops off evenly (and not only at the top of the woofer’s and tweeter’s range) so the response is still smooth everywhere else. When combined with proper speaker positioning it sounds great everywhere. At 45 degrees off axis, the level drops 6dB and at 75 degrees it drops 12dB. 90 degree side radiation is down 15dB or more.
This is nearfield frequency response at 0, 15, 30, 45, 60 and 75 degrees off axis. (I don't have a good way to measure these at low frequencies, so there is odd behavior below 150Hz in the nearfield that is not present at greater distances):
Some of the finest horn and studio monitor designs achieve constant directivity above 1000Hz, but this system achieves Constant Directivity across the entire spectrum.
Linear phase further contributes to incredible precise and stable stereo imaging. It also delivers incredible attack and slam with percussion because the wavefront is incredibly steep with no phase inversions or delays. This is a huge key to soundstage and it enables you to resolve very fine details in recordings that were previously obscure.
Below is phase response at listener position, averaged at 5 locations around my couch:
Below is impulse response:
Below is Step Response:
As you can see, the impulse and step responses are much better than nearly all speakers.
Specifications -6dB at 19Hz - 19KHz; sensitivity 100 dB, power handling 500 watts per channel; +/-2dB at listener position from 200Hz to 15KHz; +/-45 degree constant directivity; phase +/- 30 degrees from 80Hz to 15KHz.
Sensitivity is 100dB 2.83V/1 meter. Power handling is about 500 watts. Impedance 4 ohms.
Most dipole / Open Baffle designs are mounted on a flat board and the dipole cancellation effect kicks in below about 80Hz. That means by 20 Hz you’ve lost 12dB (-6dB / octave). I added “wings” on this design to counteract some of this cancellation. The top woofer, crossed at 110Hz, operates in dipole mode across its entire range.
Then the bottom two woofers have “cancellation prevention” from the cabinet sides which extends the middle woofer down to about 55Hz before rolloff begins, and 35Hz for the bottom woofer before rolloff begins. This adds significant bass extension while still maintaining the overall dipole sound distribution pattern.
The side wings, then, actually function as a short transmission line, with resonances between 150-250 Hz. By crossing it over at 100Hz I avoid peaks in the upper bass. The low end resonant frequency of the system is about 26Hz.
This is the schematic:
I thought about tri-amping these, which you certainly can do, but I felt it was considerably simpler to use a 2-way MiniDSP 2x4HD (a fantastic product) and bi-amp them, with a very simple 6dB/octave PASSIVE series crossover between the subwoofers and the mid-woofer. It is series instead of parallel because a series crossover has much cleaner interactions with impedance peaks of the woofers.
This however does not create a true 6dB crossover, because the mid woofer is acoustically rolling off at 24dB per octave below 80Hz. So the phase of the mid woofer has to be reversed to prevent cancellation, which you can see on the schematic.
In order to preserve linear phase, I used Eclipse Finite Impulse Response filter software to create a phase inversion “max phase” filter at 110Hz. This is loaded into the MiniDSP FIR function. I also used the FIR filters to linearize the phase of the B&C 15CXN88 woofers at the 1100Hz upper crossover frequency, and linearize the phase of the B&C 15CXN88 horn tweeters at the 1100Hz upper crossover frequency
I have really come to love professional coaxial horns with constant directivity. I used Radian 8” coaxial in my article in AudioXpress and they sound fantastic. A word or two about horns is in order.
I used to hate horns. The first time I heard Klipschorns I thought they had thick, congested bass, ragged midrange, and spitty highs. Yes, I did hear the dynamic range and power, but I couldn’t get past the offensive harsh sound of the horns.
But remember.... that was the 1980's. Many PA systems still sound that way. But horns in the hands of a really good practitioner do NOT sound that way. They sound marvelous. They’re not harsh, and in fact they have much lower distortion than domes and cones. But they are difficult to tame.
Horns almost always have a downward tilt in the treble (especially Constant Directivity horns) and they almost always have more dips and peaks than domes. Domes are very well behaved while horns need TLC. You have to coax the best performance out of them.
If you do that, which is best done with Digital Signal Processing, you get extraordinary dynamic range, high efficiency, and constant directivity. This achieves feats of stereo imaging that conventional speakers in principle cannot match.
Combining IIR and FIR filters to get the optimum crossover
IIR filters are the standard Digital Signal Processing feature in the MiniDSP units. FIR (which stands for Finite Impulse Response) requires special software. IIR is “minimum phase” which means that anything you do to alter the shape of a frequency response curve also has a non-negotiable phase penalty. FIR allows you to handle time and frequency separately, but it’s harder to do.
In this design I did all of the crossover and major response shaping with IIR filters and then added FIR to solve minor frequency response and phase issues.
The tweeters have 80uF capacitors to prevent damage. They roll off below 300Hz so they do not effectively factor into the crossover design. The woofer 6dB series crossover has a very high quality, low DCR inductor (<0.6 ohm).
About the B&C woofer & horn
The B&C 15CXN88 coaxial is a 15” paper cone woofer + 3” titanium compression constant directivity tweeter in coaxial configuration. My measurements are actually better than the published factory specs:
B&C 15CXN88 raw woofer SPL
B&C 15CXN88 raw tweeter SPL
As you might expect, I EQ’d the tweeters with the MiniDSP and the Eclipse FIR software to obtain flat response both on and off axis.
I chose the SB Acoustics Bianco 15OB350 woofers for their high Q of 0.7 and 40Hz resonance as they are designed for Open Baffles. In this design they have a final resonance of about 26Hz. With about 10dB of EQ boost they easily reach down to 20Hz.
Below you see a response curve (front side) with the mid-woofer wired out of phase, so you can see a null at the 110Hz & 1100Hz crossover frequencies.
In Open Baffle designs I’ve always felt that having an OB woofer firing on both sides, but tweeter firing on only one side, never sounds quite right. So I added a SB Acoustics H250 horn and 44CD-T driver on the back. Photo:
I added a small passive notch filter to flatten the response around 4Khz which you can see in the schematic.
Rear response curve:
MiniDSP configuration files: https://tinyurl.com/bitchesbrewdsp (Note: These will ONLY work on the MiniDSP 2x4HD and they are based on measurements of my drivers in my room. Use at your own discretion; all modifications are up to you.)
With 100dB sensitivity for the woofers and 106dB for the tweeters, one of the most important things is selecting amps with low noise! I use Adcom GFA2535s with the input levels turned down about halfway which brings the noise down to below audibility.
These particularly excel with percussion, strings, plucked instruments, horns and aggressive bass lines, including electronic music. There is no music style that these do not excel at, from Bruckner and Stravinsky to Rush and Porcupine Tree. The resolution is incredible. You hear everything and the stereo image is huge. It fills the room and has a palpable sense of space.
The design is copyright (c) 2021 Perry Marshall. You may build these for personal use. Commercial applications require licensing.
I'm open to specific questions you may have & can discuss here. I have information on a variety of designs, some similar, at www.perrymarshall.com/speakers
Hello Perry,
I also read with attention your different pdf.
Which of your last designs is the most natural sounding please ? I would like to sort out what is the best trade off sounding between the waved Be of the "Cottonwood edge"or a PA coax ? As EQ is used, did you find noticable difference between different coax size in the final rendu ?
Did you consider a close approach based on the dipole behavior of drivers, bandwidth related to their radius, so 4 ways oriented with littlier drivers (6" + 3" + 1" or open back treble planar) for the mid and treble with open back treble or open back compression driver (Bastanis design comes to mind for the last option) at the price of less efficienty ?
Thanks for sharing you good designs
I also read with attention your different pdf.
Which of your last designs is the most natural sounding please ? I would like to sort out what is the best trade off sounding between the waved Be of the "Cottonwood edge"or a PA coax ? As EQ is used, did you find noticable difference between different coax size in the final rendu ?
Did you consider a close approach based on the dipole behavior of drivers, bandwidth related to their radius, so 4 ways oriented with littlier drivers (6" + 3" + 1" or open back treble planar) for the mid and treble with open back treble or open back compression driver (Bastanis design comes to mind for the last option) at the price of less efficienty ?
Thanks for sharing you good designs
Revisiting some No Baffle Sculpture fun!
Finally getting ‘round to playing with my MiniDSP Flex Eight and the 6-channel class D amp I built with @Erica.C modules.
Drivers here are JBL 2217HPL, Focal Audiom 7A2, GRS PT2522-4. Crossover is LR24 at 250 and 2000. PEQ to taste.
Sounds good enough futzing around by ear that I will likely build a couple baffles for these.
Finally getting ‘round to playing with my MiniDSP Flex Eight and the 6-channel class D amp I built with @Erica.C modules.
Drivers here are JBL 2217HPL, Focal Audiom 7A2, GRS PT2522-4. Crossover is LR24 at 250 and 2000. PEQ to taste.
Sounds good enough futzing around by ear that I will likely build a couple baffles for these.
Attachments
Last edited:
Hi @diyiggy - the most 'objectively best' and natural sounding is the Bitches Brew, with the Birch Dipoles coming in a very close second (also featured on the cover of AudioXpress). After I wrote up the Birch Dipoles in AX I replaced the titanium diaphragm with Beryllium diaphragm which is noticeably better. It has a really beautiful transparency whereas the titanium has a slightly "hard" sound. The Birch Dipoles have a slightly better high end. The Bitches Brew obviously have better and deeper bass.Hello Perry,
I also read with attention your different pdf.
Which of your last designs is the most natural sounding please ? I would like to sort out what is the best trade off sounding between the waved Be of the "Cottonwood edge"or a PA coax ? As EQ is used, did you find noticable difference between different coax size in the final rendu ?
Did you consider a close approach based on the dipole behavior of drivers, bandwidth related to their radius, so 4 ways oriented with littlier drivers (6" + 3" + 1" or open back treble planar) for the mid and treble with open back treble or open back compression driver (Bastanis design comes to mind for the last option) at the price of less efficienty ?
Thanks for sharing you good designs
The Cottonwood Dipoles with the SB Acoustics Beryllium tweeter are probably the most accurate measurement wise. They are very transparent and very revealing. But they sound a little bit "clinical" and "analytical" where the Birch Dipoles are a little sweeter. The Cottonwood Dipoles don't have any bass below 35 Hz. I think the Cottonwood Dipoles with an SB Acoustics 15OBO350 would be better than the Lavoce woofer I used in the design. Has 11.5mm Xmax instead of 7mm.
I believe you are asking my opinion of separate tweeters vs. coax. I somewhat prefer the coaxes. They have a better radiation pattern. They are harder to deal with, though. The SB Acoustics tweeter with its waveguide is better behaved. You always get interference patterns in a coaxial / concentric design so you have to decide which interference pattern is more objectionable: The comb filtering of separate midrange + tweeter or the melding of the concentric tweeter with the woofer.
I'd like to show you some measurements of the Bitches Brew design and discuss more of the thinking behind it. (Original post https://www.diyaudio.com/community/threads/ultimate-open-baffle-gallery.123512/post-6682515 )
In this graph...
You see the microphone about 4 feet in front of the left speaker, on axis, (red) and without moving the microphone, the purple is the right speaker, about 8 feet away and 60 degrees off axis (purple). It illustrates the behavior of a true Wideband Constant Directivity speaker. Even the midrange and bass are directional and 60 degrees off axis the whole curve tilts evenly downward.
The only possible way I know of to achieve this is with a huge horn (that would not fit in any normal living room) or else Open Baffle with CD horns that smoothly integrate with the radiation patterns of the woofers.
This is the same curve, but 1/3 octave smoothing instead of 1/12:
...which is more representative of what your ear hears.
There is a huge payoff to this for stereo imaging, because I have hosted listening parties with 20-30 people and EVERYONE hears a great stereo image. I arrange the chairs in a big 2/3 circle that stretches from the outside of the left speaker all the way to the right. The photo below is a listening party at a hotel, with a different dipole project that I hope to write up someday:
You can stand right next to the left speaker and hear the right speaker equally well, across the room ten feet away.
The other issue I want to raise is flat phase response. This is the phase response curve of the Bitches Brew design:
Very very few speakers come close to this; nearly all speakers with passive crossovers have polarity reversals and phase rotations. A very few manufacturers of passive speakers (Thiel, Vandersteen, Duntech) attempt to address this. The rest just ignore it. I work around the phase issues with DSP FIR filters. But what I want to focus your attention on is the low-end phase response, which runs only about 45 degrees positive in the main bass and only tends towards 90 degrees below 23 Hz. A reflex speaker will be at least 180 degrees out of phase at the tuning frequency.
Most people say you can't hear this, but you can prove otherwise to yourself if you have DSPs and do a little experiment.
Play "Fear of a Blank Planet" by Porcupine tree, whose bass drum has a lot of high-end slap but also delivers a very low fundamental around 20-30 Hz. Set a 20Hz high pass (infrasonic) filter and then try changing the slope from 6 to 12 to 18 to 24 to 36dB.
You will easily hear that with steep filters (=longer phase shift) you hear the slap first and the "boom" of the low bass comes LATER. As in, the two sounds arrive at different dates. Try it with a set of headphones if all your speakers have steep low end rolloffs.
Open Baffle speakers have 6dB per octave rolloff from the baffle loss and 12dB per octave roloff from the drivers, for a total slope of 18dB per octave.
I used low Q notch filters to create a 20Hz bass boost for the Bitches Brew woofers which you can see here:
The low-Q EQ boost preserves phase information. In well-recorded music with lots of bass and percussion, the timing is impeccable.
Also I believe that near-zero phase lag in the midrange and treble (and corresponding good impulse response, which you can see in my original writeup) further enables the design to deliver fantastic imaging.
All of the measurements here are done in my room with no gating.
In this graph...
You see the microphone about 4 feet in front of the left speaker, on axis, (red) and without moving the microphone, the purple is the right speaker, about 8 feet away and 60 degrees off axis (purple). It illustrates the behavior of a true Wideband Constant Directivity speaker. Even the midrange and bass are directional and 60 degrees off axis the whole curve tilts evenly downward.
The only possible way I know of to achieve this is with a huge horn (that would not fit in any normal living room) or else Open Baffle with CD horns that smoothly integrate with the radiation patterns of the woofers.
This is the same curve, but 1/3 octave smoothing instead of 1/12:
...which is more representative of what your ear hears.
There is a huge payoff to this for stereo imaging, because I have hosted listening parties with 20-30 people and EVERYONE hears a great stereo image. I arrange the chairs in a big 2/3 circle that stretches from the outside of the left speaker all the way to the right. The photo below is a listening party at a hotel, with a different dipole project that I hope to write up someday:
You can stand right next to the left speaker and hear the right speaker equally well, across the room ten feet away.
The other issue I want to raise is flat phase response. This is the phase response curve of the Bitches Brew design:
Very very few speakers come close to this; nearly all speakers with passive crossovers have polarity reversals and phase rotations. A very few manufacturers of passive speakers (Thiel, Vandersteen, Duntech) attempt to address this. The rest just ignore it. I work around the phase issues with DSP FIR filters. But what I want to focus your attention on is the low-end phase response, which runs only about 45 degrees positive in the main bass and only tends towards 90 degrees below 23 Hz. A reflex speaker will be at least 180 degrees out of phase at the tuning frequency.
Most people say you can't hear this, but you can prove otherwise to yourself if you have DSPs and do a little experiment.
Play "Fear of a Blank Planet" by Porcupine tree, whose bass drum has a lot of high-end slap but also delivers a very low fundamental around 20-30 Hz. Set a 20Hz high pass (infrasonic) filter and then try changing the slope from 6 to 12 to 18 to 24 to 36dB.
You will easily hear that with steep filters (=longer phase shift) you hear the slap first and the "boom" of the low bass comes LATER. As in, the two sounds arrive at different dates. Try it with a set of headphones if all your speakers have steep low end rolloffs.
Open Baffle speakers have 6dB per octave rolloff from the baffle loss and 12dB per octave roloff from the drivers, for a total slope of 18dB per octave.
I used low Q notch filters to create a 20Hz bass boost for the Bitches Brew woofers which you can see here:
The low-Q EQ boost preserves phase information. In well-recorded music with lots of bass and percussion, the timing is impeccable.
Also I believe that near-zero phase lag in the midrange and treble (and corresponding good impulse response, which you can see in my original writeup) further enables the design to deliver fantastic imaging.
All of the measurements here are done in my room with no gating.
Sorry @kd4ylq I only just now saw this. (I think you have to put @perrymarshall in your post for me to get notified by email???) I don't think I would put padding on any of the interior surfaces. Let's say the padding was 4" thick, it would only absorb frequencies with wavelengths 4X the 4" or above ~800 Hz. And I think it would interfere with low frequencies.
I think the ringing at these frequencies is inconsequential. You may notice some squiggles around 200-300Hz which is probably transmission line behavior of the "U" baffles but padding as described above will not solve any of that. If you feel it needs attention then it's easy to fix with DSP.
I think the ringing at these frequencies is inconsequential. You may notice some squiggles around 200-300Hz which is probably transmission line behavior of the "U" baffles but padding as described above will not solve any of that. If you feel it needs attention then it's easy to fix with DSP.
Hi Perry,
Marry Christman.
Your Bitches Brew looks interesting. And a few claims that are very interesting, including
"it's the only speaker [you've] ever seen
that is constant directivity from 20 to
20k it's a true figure eight
radiation pattern from bottom to top..."
...the phase response is flat from 50 hertz
on up the impulse response is nearly perfect
the step response is beautiful i do not
know of any other speaker in the world
that does all of those things at the
same time and 100 db sensitivity"
That's a great accomplishment, if true.
Do you have any full anechoic polar response, CTA-2034A or spinorama type measurements to share?
Have you considered sending them to Amir in Seattle to measure?
Marry Christman.
Your Bitches Brew looks interesting. And a few claims that are very interesting, including
"it's the only speaker [you've] ever seen
that is constant directivity from 20 to
20k it's a true figure eight
radiation pattern from bottom to top..."
...the phase response is flat from 50 hertz
on up the impulse response is nearly perfect
the step response is beautiful i do not
know of any other speaker in the world
that does all of those things at the
same time and 100 db sensitivity"
That's a great accomplishment, if true.
Do you have any full anechoic polar response, CTA-2034A or spinorama type measurements to share?
Have you considered sending them to Amir in Seattle to measure?
Last edited:
I don't have anechoic or spinorama, I do have these taken in my listening room:
In the above graph, I suspect the "bunching" around 100 Hz disappears at greater distances. This was taken with the speaker on a turntable in my room at 0 15 30 45 60 75 degrees.
Because of the limitations of my room, accurate low frequency measurements are difficult. Sending them to Seattle would incur some expensive shipping. But there is no question that the polar pattern is a figure 8 dipole. Do you know of any other speaker that is constant directivity from 20Hz-20Khz?
In the above graph, I suspect the "bunching" around 100 Hz disappears at greater distances. This was taken with the speaker on a turntable in my room at 0 15 30 45 60 75 degrees.
Because of the limitations of my room, accurate low frequency measurements are difficult. Sending them to Seattle would incur some expensive shipping. But there is no question that the polar pattern is a figure 8 dipole. Do you know of any other speaker that is constant directivity from 20Hz-20Khz?
Last edited:
