• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Constant directivity speakers (DSP) for DIYers

Jag768

Member
Joined
Jan 29, 2019
Messages
11
Likes
59
I built my own, loosely based on the D&D 8C using a DIY waveguide (plaster) and a passive cardiod enclosure for midbass

IMG_0590.jpg

IMG_0591.jpg

IMG_0569.jpg

IMG_0729.jpg

IMG_0743.jpg

It's quasi cardioid to cover the whole spectrum from schroeder on up

config 4 polar.PNG

2 way, filter is done by miniDSP (cdsp 6x8)

config 4 summing.PNG

IMG_0619.jpg

The end result is pleasing and supplemented by 4 subs (using 4 dsp channels and multisub optimizer)

20181205_123023_phone.jpg

20181205_123139.jpg
 
OP
H

hvbias

Addicted to Fun and Learning
Joined
Apr 28, 2016
Messages
574
Likes
411
Location
US
@Jag768 wonderful stuff. Which equation did you use to come up with the waveguide and just curious if you posted this to DIYAudio?
 

Cosmik

Major Contributor
Joined
Apr 24, 2016
Messages
3,075
Likes
2,180
Location
UK
Can anyone fill us in on the theory of the cardioid cancellation?

The way I see it, such a system can be optimised purely through trial-and-error (with a more scientific name, possibly). You would start out by saying that there must be some arrangement of slots that means that the backwave from the driver meets the front wave at different times and places and cancels out. With trial and error, an optimal set of slots can be found based on some criteria you specify e.g. minimal SPL measured with (possibly simulated) microphones arrayed around the rear of the (possibly simulated) speaker and summed, over some range of frequencies.

Similarly, you could do it actively with drivers at the rear and optimise the DSP'ed signal to achieve something similar. I don't doubt it will work at some level.

But... what I wonder about is the consequences of imperfect cancellation. Without a very special configuration that might not be compatible with real drivers and rectanguloid boxes (yes, it's a real word :)), might there be 'beams' or 'lobes' of non-cancellation, possibly not laterally where the microphones might be, but vertically or in between. Does some of the rear (basically antiphase) radiation escape backwards à la open baffle speaker? What if the speaker is placed near a wall on one side, or a wall at the back? Does the cancellation still work? (The question must surely be: why should it still work?)

In previous discussions it has been clear to me that very few 'objectivists' worry about the antiphase that escapes from the back of an open baffle speaker because it isn't really possible to distinguish it in an in-room frequency response measurement. But I suggest that it can be heard by some people as 'phasiness', nevertheless, . And it must interfere with any time domain psycho-acoustic phenomenon that relies on the ear/brain matching polarities and envelopes of sounds and their reflections. So I would like to know the theory that confirms that "Cardioid cancellation does what it says on the tin and doesn't 'leak' anti-phase into the room".

An ordinary monopole speaker doesn't leak *any* anti-phase into the room, so it is potentially a significant difference I would suggest.
 

HammerSandwich

Major Contributor
Forum Donor
Joined
Nov 22, 2018
Messages
1,137
Likes
1,497
An ordinary monopole speaker doesn't leak *any* anti-phase into the room...
What about ported boxes? And sealed speakers that lack perfect cabinets?

More seriously, the classic BBC cabinet paper showed cabinet inaudibility around -20dB or so, IIRC. That's a long way from no output at all. Aren't the modern cardioids getting 12-15dB attenuation to the rear?
 

DDF

Addicted to Fun and Learning
Joined
Dec 31, 2018
Messages
617
Likes
1,355
What about ported boxes? And sealed speakers that lack perfect cabinets?
More seriously, the classic BBC cabinet paper showed cabinet inaudibility around -20dB or so, IIRC. That's a long way from no output at all. Aren't the modern cardioids getting 12-15dB attenuation to the rear?

A bit OT but to ensure resonances are inaudible, I recommend following the guidance from Sean Olive's study on the audibility of resonances from the JAES many years ago. If you can guarantee -28 dB or better, their work indicates that it should be inaudible under worst case conditions whether pink noise or impulse. Of course its difficult to validate box resonance contribution in isolation from the driver. Accelerometer measurements provide indication of panel vibration frequency and location, but not contribution to the total sound power. For this, cepstral analysis is probably a good start
 

Jag768

Member
Joined
Jan 29, 2019
Messages
11
Likes
59
Can anyone fill us in on the theory of the cardioid cancellation?
The theory is quite simpel. When you place a speaker in a tube, you displace the front and back wave cancellation towards the rear. Only now you have introduced a resonance which affects the amplitude response of the backwave. Hence the damping which acts as low pass filter. The depth of the "tube" behind the driver dictates the resonance frequency but also allows more dampening material depth.

Note: passive cardioid, while very uncommon for speakers, is the de facto standard for microphones (although active solutions are used commonly as well, by combining omni and figure8). For instance, the Shure SM58, the most commonly used of all, is a passive cardioid.
The way I see it, such a system can be optimised purely through trial-and-error (with a more scientific name, possibly). You would start out by saying that there must be some arrangement of slots that means that the backwave from the driver meets the front wave at different times and places and cancels out. With trial and error, an optimal set of slots can be found based on some criteria you specify e.g. minimal SPL measured with (possibly simulated) microphones arrayed around the rear of the (possibly simulated) speaker and summed, over some range of frequencies.
I think it would very well be possible to make this into a mathlab model, which would simplify design. However, as a mathlab nitwit and not being an engineer, i won't be the one making it :p.

I had previously made a damped u frame, similar to the design by John K.

IMG_7341.JPG

It worked, but not as good as my current model, here's the polar of my first attempt. Looks like a guitar right :) ?
Final polar - sonogram.png


For the second attempt I took a good look at the products from Dutch&dutch and made a test model. It took only one test model, it was right on the first take. I tried different densities of absorption (Akotherm D20 and D40), it didn't even matter that much, but D40 won.

20170204_173430.jpg
5d_40_fullspace.png


Looks great right? There's one "but"; the measurements are all indoor because of living in a city in the netherlands, so no large outdoor space to do measurements. My ceiling is 3m, i can use a gate of 5,5ms. However, the mic is at less than a meter distance, which makes it sensitive to variation in distance to the woofer when making measurements at the rear. In the spring, i'm going to take everything outside and do a polar, burst decays and distortion @high spl using different xover settings .

Here's the whole family happily together
IMG_0806.jpg
IMG_0818.jpg
Icepower Rotel amps, the one and only time they "made" class d amps :)
IMG_0619.jpg
The moneyseat
IMG_0809.jpg

Daytime
IMG_0796.jpg
Nighttime
IMG_0790.jpg
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
Excellent work @Jag768 :) I'm sure everyone here is truly impressed.

Have you considered bringing your couch a little forward from the back wall and placing some absorbent material on the wall behind your head? It looks like your ears are going to be extremely close to the undamped back wall reflections.
 

Jag768

Member
Joined
Jan 29, 2019
Messages
11
Likes
59
The picture, an old one of the city i live in, is a hidden absorber hiding 80mm high density rockwool.
Moving the couch isn't really an option, it would block the passage to the kitchen.
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,597
Likes
2,235
I tried different densities of absorption (Akotherm D20 and D40), it didn't even matter that much, but D40 won.

Can I ask where you obtained the Akotherm (polyester wool) and how much it cost?
I've been looking around and the only source I've found is in the Netherlands (I'm in the US).
Have not contacted for pricing (yet).

i can use a gate of 5,5ms. However, the mic is at less than a meter distance, which makes it sensitive to variation in distance to the woofer when making measurements at the rear.

5.5ms gating is a resolution of 180 Hz (for my benefit; I know you understand this ;-) ). Have you tried nearfield measurements, say at a small fixed radius around the enclosure, to get a better picture of the sound field?
 

Jag768

Member
Joined
Jan 29, 2019
Messages
11
Likes
59
Im based in the Netherlands. For once it's the other way around :). Prices of audio gear in the US is generally way lower than here! But we have more vacation to work on DIY projects :p

Mineral wool works just as good, only it tends to sag and it's less precise.

I have tried nearfield. But not sure what it helps. Measuring the front of the driver nearfield will just give you that: the response of the driver itself. Measuring at the "slots"; it's not really nearfield. It's just the sum of front and back.

What you could do, i'm thinking out loud here, is place the driver on an infinite baffle, put the cardiod midbass enclosure behind the driver. Than measure (farfield) the front output and the back output. The back output is the combination of the driver (rear) output + resonance of midbass enclosure + low pass of absorption material. You could than fiddle around until front and back match.
 

dc655321

Major Contributor
Joined
Mar 4, 2018
Messages
1,597
Likes
2,235
But we have more vacation to work on DIY projects :p

Don't for one second think I am not terribly envious :)

I have tried nearfield. But not sure what it helps.

Resolution is what it helps, of course. Assuming the measurements you've shown above are all gated, you have ~5 data points between 100Hz and 1000Hz at your listed gating, and I'm not sure how accurate the first couple would be (up to 0.3-0.4 kHz). You did say you measured under a meter away, so my point may be moot anyway...

Measuring at the "slots"; it's not really nearfield. It's just the sum of front and back.

Is it not helpful to think of the slots as additional sources, the same as any conventional driver?

Please don't take my comments as negatively critical of your achievements: I think your project and results are great.
 

Jag768

Member
Joined
Jan 29, 2019
Messages
11
Likes
59
No thank for the suggestion :). I know the resolution is low, and measuring close makes it also prone to variation in distance between mic and driver. Therefore I'm planning on doing the measurements outdoor in spring. For nearfield measurements at the slot, I just wouldn't know what information it would give me. Even if you would assume it to be a separate sound source, you still have the output of frontside.
 

Jag768

Member
Joined
Jan 29, 2019
Messages
11
Likes
59
As a big fan of his work, I'm very curious to know what @Floyd Toole thinks of using cardioid midbass, thereby extending the controlled directivity to the transition zone. The importance of smooth directivity for high frequencies (ie above transition zone) has been convincingly shown. I would expect the directivity to be important in the transition zone as well, since there is no sharp demarcation among those zones. But i haven't been able to find any controlled listening experiments comparing omni midbass to cardioid midbass in a (normal) reverberant room. Does anyone know any?
 

Juhazi

Major Contributor
Joined
Sep 15, 2018
Messages
1,717
Likes
2,897
Location
Finland
^Not me, just some articles in magazines and forum comments. A cardioid radiation pattern helps only for front wall cancellation, sidewall get as much energy as a standard monopole loudspeaker gives. Dipole pattern or horn helps for sidewalls, but a dipole of course suffers from front wall... A bass horn must be terribly huge.

Choose your poison...
 

andreasmaaan

Master Contributor
Forum Donor
Joined
Jun 19, 2018
Messages
6,652
Likes
9,399
^Not me, just some articles in magazines and forum comments. A cardioid radiation pattern helps only for front wall cancellation, sidewall get as much energy as a standard monopole loudspeaker gives. Dipole pattern or horn helps for sidewalls, but a dipole of course suffers from front wall... A bass horn must be terribly huge.

Choose your poison...

It depends a bit how things are oriented. The sidewall reflection point may well be >60 deg off-axis, in which case it would be effectively reduced by a cardioid radiation pattern. I’m also not sure I’d call sidewall reflections a “poison” in two-channel stereo, as there’s much evidence their presence is perceived by listeners to actually be beneficial.
 

Jag768

Member
Joined
Jan 29, 2019
Messages
11
Likes
59
^Not me, just some articles in magazines and forum comments. A cardioid radiation pattern helps only for front wall cancellation, sidewall get as much energy as a standard monopole loudspeaker gives. Dipole pattern or horn helps for sidewalls, but a dipole of course suffers from front wall... A bass horn must be terribly huge.

Choose your poison...
Thank you for your reply. Im referring to the transition zone, say 200 - 700hz, rather than bass frequencies (modal zone). The question then is; does the spectral content of reflections matter in the transition zone as it does for >700hz.
 

Juhazi

Major Contributor
Joined
Sep 15, 2018
Messages
1,717
Likes
2,897
Location
Finland
I was mainly discussing first reflections, which cause a huge dip in response when freguency/wavelength and delay make cancelaltion. We must remember to consider also the back wall behind the listener! These are most difficult around 100-400Hz.

Transition zone 200-700Hz is difficult to analyze. In in-room measurements we see many many first reflection nulls with short gating and also many other wiggles that may come from speaker's edges/directivity problems. Also multiple order boundary reflections are mixing together, with long gating or RTA. I use to overlay multiple measurements from different mic locations or longperiod averaged RTA with mic in my waving hand (MMM)

I want to believe that highish horizontal directivity from 200Hz up is for good and also not too difficult to achieve. A 15" cone+horn crossed around 900Hz can do that, also most cardioids and dipoles.

My copy of Gradient 1.3 speakers has monopole bass crossing to dipole mid around 180Hz LR2 and horizontal directivity is cardioid 120-250Hz, dipole above it. I have done hundreds of measurement both in- and outdoors and this is the nicest looking horizontal 360¤ sonogram, outdoor 12ms gating, other normalized. Room response with 500ms gating shown too, my room is quite wide. This sounds nice, if I lift 100-300Hz higher it starts to sound boomy. Bass is sealed downfire with dsp, -6dB is around 12Hz in room response!

My build thread at diyadio.com https://www.diyaudio.com/forums/mul...aborative-speaker-project-48.html#post4231049
 

Attachments

  • ainogneo83 vx out polar 10¤steps 12ms.jpg
    ainogneo83 vx out polar 10¤steps 12ms.jpg
    96.3 KB · Views: 314
  • ainogneo83 vx out polar normal 10¤steps 12ms.jpg
    ainogneo83 vx out polar normal 10¤steps 12ms.jpg
    96.3 KB · Views: 356
  • ainogneo83 conf3 room spot vari 500ms 112.jpg
    ainogneo83 conf3 room spot vari 500ms 112.jpg
    111 KB · Views: 344
Last edited:
Top Bottom