S16 (M, MS, MSXT, MSTi) - Constant Directivity Dipoles

The S16 being measured in free space condition.
Roger, Paul, and Daniel are seasoned loudspeaker enthusiasts and builders.
Melbourne 15/1/2011

Work in progress to improve high frequency directivity of S12.

Summary of Midrange and Tweeter variants:
  • S16 - 10/12/10
    Seas L21, Dayton 10" wg, D220Ti compression driver
    Also B&C DE250 compression drivers later on
  • S16/M - 27/9/11
    M=Magnesium - Seas W22 woofers
  • S16/MS - 30/10/11
    Symmetrical tweeter response - Seas W22, DX25 + Monacor WG300
  • S16/MSXT - 27/11/11
    Replaced DX25 with XT25
  • S16/MSTi - 27/11/11
    Replaced XT25 with Vifa DQ25SC16-04


Jaycar CW2119

- Two MiniDSP 2x4 Kit
- Advanced 4-way xo plugin

8-Channel Sure Electronics Class D

Follow the build discussion at DIYAudio Forum.

Measurement Data:

Directivity plot, Native frequency response, 1m:

Directivity plot, Normalised to 0deg, 1m:

In-room frequency response (1m ungated), also with reversed tweeter.

Why the fuss ?

You can compare the directivity plots above with typical dome tweeter implementation:

S9 OB:

We see narrowing of loudspeaker polar response, followed by "tweeter blooming". This is an expected response as the tweeters would behave omni-directionally at transition frequency. Humorously known as the "Klingon bird of prey" polar response :)

Recently a combination of large woofer and waveguide are becoming very popular, similar to the S15 Econowave build. While these types of speakers can provide constant directivity down to 800-1khz, unfortunately they soon transition to omni-directional at lower frequencies:

S15 DSP:


There is a certain correctness about the sound of speakers with such smooth directivity from 20Hz to 20khz. It is immediate and dynamic. However with any loudpeakers they have plus and minuses:
Very dynamic presentation. Willing to play effortlessly at very high SPL. Sound is very 'direct' and make dome tweeters sound diffused in comparison. Very natural.

Lack of rear HF radiation (arguably ok or even necessary), some degree of horn and compression driver coloration make them not as smooth as dome tweeters. Flat dipole bass is unable to pressurize the room effectively below 40hz, although these can easily be substituted with H or W frame dipole radiators.

Room response analysis

Elias Perkonen kindly performed wavelet analysis of the in-room response of these speakers.

Native frequency response, left and right:

Normalised, left and right:

Longer time windows, 150ms and 300ms:

As comparison, here are the bark wavelets of Elias' Dipole Line Array, 20ms and 150ms:

Update 21/5/2011
ps. no rapture yet!

Tried Selenium D220Ti compression drivers with B&C DE250. These drivers use Poly cone material and have a smooth but 'mellow' sound. Read about my burst signal test investigation here. (note: update 22/9/11, I prefer the cheaper D220Ti as a more accurate transducer).

The polar response is excellent and comparable to D220Ti although mounting adapters have to be used.

Update 27/9/2011

Replaced L21RNX/P with Seas Excel W22, Magnesium cone

The Excels are interesting. When I tap the cone interestingly they don't sound like typical metal drivers. They sound like polycones! Magnesium are lighter than aluminium and apparently Seas were able to make a thicker cone for the same weight, providing better damping.

Replacing the Aluminium L21RNX/P there were only 2 differences. The cone resonance peak (W22 is about 100hz higher) and region between 1khz-2khz. Frequency response-wise they track within 1-2db. It was also interesting to note the behaviour of the resonance. It's different. The Aluminium has periodic resonance after the main one, while W22 has only one major peak and drops after that. Both L21 and W22 have serious drop about 2khz and really the 1.5khz XO is really pushing it.

Sonics of S16/M

Now, after only minor adjustment, which is the metal cone notch I sat down and listen.... Tonally they are inseparable. They sounded very much the same. However the Excel have this specific character that is quite difficult to explain. Basically it sounded mellow and laid back, but with resolution and transparency. It projects refinement and practically makes any music with overtones sounds......beautiful?? They reminded me of the Orions.

I think, however that the Excels are lacking the gusto and immediate transients of the Aluminium L21 especially playing percussive instruments.

Interesting approach to solve dipole tweeter directivity by roflynn. Using Scanspeak tweeters and Monacor Waveguide WG300

Building the 16/MS (magnesium, symmetrical)
The objective is to provide front-rear controlled directivity of tweeter region, with symmetry. This was not able to be done with Waveguide/compression driver combination.

To the rescue is Vifa DX25 loaded with Monacor WG300. The dispersion is respectable (not as good as 16/M) but more importantly the small dimension allows back-to-back setup.

The polar response is not as smooth as large waveguide in S16, but it is sufficient to match dipole's directivity of -6db @60deg. When moving around the speakers I could hear the level drops uniformly. It is the best compromise so far between directivity and front-rear symmetry.

0-90deg, 10deg increment, normalised.

Replaced DX25 with XT25 ring radiator. The benefit is elimination of 12khz null thanks to the phase plug. This version was shown at Melbourne round-trip GTG.

Deep reverse null showing excellent time alignment and phase tracking. Also displayed is the difference between DX25 and XT25 when loaded with the waveguide. The lack of phase plug caused ~12khz null with DX25.


I wanted better HF resolution which only rigid/metal dome tweeters can provide. A quick experiment mounting a Titanium domeVifa DQ25SC16-04 to the Monacor waveguide resulted in good result. I removed the phase plug of these drivers.

When loaded with the waveguide a 10+dB peak appears. This is excellent as we can notch this peak and resulted in lower distorion (distortion figures of these drivers are already excellent). It also resulted in excellent phase tracking, despite the non-matching diameter between waveguide throat and the dome. To my surprise the lack of phase plug does not exhibit a dip in the response (that 18khz dip is inherent to the driver).

The smallish waveguide loses directivity at 2khz. The polar response is not as good as a proper 10" waveguide, but definitely much better than normal dome tweeter on a baffle.

I prefer the sonics of these drivers compared to XT25 and earlier DX25. They are very transparent and detailed. YMMV. (Update 8/1/2012: They are not nice visually. The dome is light-reflective and draws attention. I liked the black XT25 better and have reinstalled them).

Update 7/2/2012
Experimentation with Linkwitzlab "Watson"

rich31td's build:

Update 31/1/2015
A 'released' build of this prototype is being commissioned: Gainphile R16

S15 - Econowave DSP

The Original S15 Econowave DSP 
Eminence Beta 12a + Selenium D220Ti & Pyle WG 

Econowave is a cost effective controlled-directivity loudspeaker design. DIY folks at AudioKarma had investigated numerous driver + waveguide combination which would achieve the optimum polar response using off-the-shelf products.

Note: You could get lost in the thousands of posts in AudioKarma. But here's my refence links which summarise the variations of the designs: AudioKarma and  TechTalk

For this DSP build, follow my post in AudioKarma and DIYAudio forums.

Topology: Active 2-way with pro-woofers and compression drivers
Radiation pattern: Monopole - Controlled Horizontal Directivity
Directivity: Omni-directional at lower  frequencies, smooth transition to 90 deg beam at 1.2khz
Frequency response: 20hz - 17khz. Critically damped LF Q of 0.5
Sensititivity: ~92dB
Dimension (H, W, D): 57 x 37 x 33cm
Vertical lobe (Listening position) : Tweeter axis, toe'd in 5-10deg
Amplifiers: 4-Channels required (2x stereo amps)
Crossover: DSP (MiniDSP + 2way Advanced Plugin)

Cabinet construction

The cabinet is basically made from 33cm, 18mm MDF cutouts. The offcuts are glued into the inside panels to provide additional strength and rigidity. The finished speakers are very heavy.

Cabinet C1

 Cabinet C2
(future design ...)

A handsaw was used and could not provide precise cuts. Hence the  construction can be made square and flush effectively by ofsetting 0.5mm and sanding.

The primary reason for stuffing is to dampen modes inside the box structure. With so much variations on stuffing material (dacron, acousta-stuff, natural wool fibers, etc.) how much stuffing requires some experimentations. I use dacron (those pillow-stuffing) and starts about this much, very loosely filled. After some listening I removed some. The goal is to have enough to reduce the modes, but not too much that it overdamps and reduce the box volume. It is a competing factor.

DSP Crossover Configurations

The DSP configuration is available below.

I configured the speakers using an older version of MiniDSP plugin which had a bug in it
The effect is that if you use the newer plugin (which is the fixed one) the bass will sound heavy. So please be aware of it. I will revisit these in the future with the updated plugin.

Important: do not forget to implement  L-PAD for the tweeters.

Woofer DSP crossover  

Woofer W1: Eminence Beta 12
Cabinet C1 
  • XO1250hz, LR2
    Crossover point is chosen by careful consideration of distortion, group delay, power handling, and polar response. See my investigation here on dispersion pattern of various xo points.
  • EQ1:  Notch Filter
    PEAK 618hz / -2.8db / 2.6Q
  • EQ2:  Linkwitz Transform Biquad (LT)
    LT: Select "Advanced", Copy and paste below, and click "Process".
    The woofer is equalised to F3=20hz with critically-damped Q of 0.5. This means -3db@40hz and -6db@20hz. Please note that these values are driver and cabinet volume-dependent. If you change the cabinet volume drastically or change the Woofer, then you need to recalculate and remeasure the transform. No you can't just punch in factory-provided TS parameters.

    Read more about LT Biquad here and here, and also a calculator.

    Raw woofer measurement (nearfield)
    LT Biquad (simulation)
    Woofer with LT, Notch, and XO (measured nearfield)
    Green=raw, Black=Equalised woofer, Red=Target frequency response
  • EQ3:  Baffle step compensation
    HIGH-SHELF 206hz / -6db / 1Q
    Baffle step consist of shelving highpass with -6db/oct gain. It compensate the transition from 4pi (omni-directional) to 2pi (forward radiating) of box speakers. Read about baffle step here.
    Woofer response with Baffle Step (measured, nearfield)
    Woofer response from 1m
    Black=with baffle step
  • Ignore values of EQ4.. etc. 
  • Attenuation, Phase, and Delay
    With CD1: Attenuation 0, Polarity NORMAL, Delay: 0ms
    With CD2
    : Attenuation 0, Polarity NORMAL, Delay: 0.06ms

Tweeter DSP crossover

CD1 (Selenium D220Ti)
Compression driver: Selenium D220Ti screw-on
Waveguide:  JBL 1" thread-on Progressive Transition, Part# 338800-001, or Pyle Pro PH612, or Dayton clone. In Australia: DJCity YD-L033
Directivity pattern measured from 0-90deg, 10deg increments:

  • XO1250hz, LR2
  • EQ1:  Waveguide loading notch
    PEAK 1956hz / -4.8db / 2.6Q
    Raw and equalised Compression Driver + Waveguide (from 80cm)
    Green=raw, black=equalised.
  • Passive L-PAD
    Passive L-PAD is used at the compression driver to provide main attenuation for woofer level matching. The benefit is that noise from Amplifier will also be reduced. The remaining few dB is adjusted from MiniDSP.
    L-PAD-11db using 5.6 Ω  and 3.3 Ω 5w resistors
  • Attenuation, Phase, and Delay
    With W1: Attenuation -5db, Polarity INVERT, Delay: 0.04ms
    Measurement from 1m at Tweeter level

    Ignore lower frequencies.
I have a quick iphone capture of these loudspeakers in youtube:

CD2 (B&C DE250)
Compression driver: B&C DE250 + DJCity adapter (photo)
The adapter needs to be enlarged to 1 inch, so better ones may be found from PartsExpress.
Waveguide:  JBL 1" thread-on Progressive Transition, Part# 338800-001, or Pyle Pro PH612, or Dayton clone. In Australia: DJCity YD-L033
Directivity pattern measured from 0-90deg, 10deg increments:

CD3 (Celestion CDX1-1745) 
Compression driver: Celestion CDX1-1745 (photo) + PE Adapter
Waveguide: As per CD1 and CD2 (Pyle) 
I configured and listened briefly this compression driver at recent GTG and was pleasantly surprised as how similiar the frequency characteristic and sonics is compared to B&C DE250.
A longer-term experience and listening may be needed to confirm, but there were three of us who felt the same during that brief setup. We even set it up playing both drivers together left and right.
(update: this has been confirmed by another guy that CDX1-1745 is a poor man's DE250!)

Please follow the exact configuration of CD2 (DE250) for this compression driver.

Optional variations

35hz Linkwitz Transform
If the woofers (W1) are equalised to 35hz, Q=0.5 (instead of 20hz above), then further +5db can be obtained and that raises the loudspeaker's sensitivity to ~97dB. This also relieves the strain of the woofers when playing loud. In order to have this:

  • W1 EQ2 biquad needs to be changed to 35hz, Critically damped Q=0.5
  • W1 EQ3 Baffle Step SHP needs to be changed to Shelving Lowpass
  • W1 EQ4 biquad of -1db is added.

Second-order Linkwitz-Riley crossover topology (LR2) was chosen for group delay, however you might want to play the loudspeakers at very high SPL continously. In this case a 4th order crossover providing 24db/oct slopes may be a better approach. This can be done by selecting Woofer and Tweeter crossover to LR 24db/oct and changing the Tweeter's Polarity to non-inverting.

Using Subwoofer
Subwoofers can greatly relieve the stress to the pro-drivers which has highish Fs. I would recommend at least a pair to minimise excitation of room modes. The optimum scenario is to cross with LR2 slope (Q=0.5) and this can easily be done by modifying EQ2 (LT) biquad values. For example, if you would like to cross with the Subs at 80hz the biquad values for the LT would be like this.

"Geddes" Foam Plug
Earl Geddess popularised and patented the 30ppi foam plug approach on his GedLee line speakers. Many users reported improvements in the sound quality. I also  tried it and posted my experience which was different. The foam will attenuate the high frequency. So you need to increase the tweeter level by 2dB. 
In Australia you can buy them from FlexFoam. Ask for "30ppi Mericell foam". They are also referred to as filter or reticulated foam.

Loudspeaker Toe-In
There is a view that these speakers are best listened at extreme toed-in angle, as much as 45degrees. It does seem to provide uniform coverage across listeners but with soundstage depth compromise. I find about 5-10deg off the normal triangle is satisfying.

Room response analysis

Elias Perkonen kindly performed wavelet analysis of the in-room response of these speakers.

Native frequency response, left and right:

Longer time windows, 150ms and 300ms:

Update 11/2/2011
Zilch (Evan) who is the champion of this design passed away. He will be sorely missed.

Update 3/2/12
Production version of original S15 documented and  posted.

Update 4/2/12
B&C DE250 variant posted
Celestion  CDX1-1745 variant posted

Update 14/9/12
Added information on stuffing

Update 16/9/15
Check out these youtube videos :)

Update 19/6/16
Biquad values will be different for newer 96khz boards such as 4x10HD or 2x4HD. Use these two biquads instead. The first biquad is to shape the original response to 35hz with Q=0.7, and the second one is 35hz Q=0.7. Together they produce acoustic shape of 35hz LR4 highpass.

Try to remove the baffle step compensation first.  I did not need it in my room.



S14 DSP - 4 Way Open Baffle with DSP XO

A work in progres ...

These are 4-way Open Baffle loudspeaker design with 3-way main panel and inverted W-Frame subwoofers.

- Low midrange distortions, by assigning separate drivers for low-mid and high-mid duties
- Low vibrations due to W-Frame subwoofer's opposing forces
- Small footprint and height

- The 5" high-mid drivers may proven to be too small for the baffle width
- Designing 4-way crossover will be more complex due to phase interference
- Back to back dome tweeters' known limitation. "dipolar" response, instead of dipole

Tweeters: Vifa DQ25 with phase shield removed. Back to back.
Mid-high: Fountek FW146 5" aluminium woofer
Mid-Low: Seas L21RNXP, 8" aluminium woofer
Subs: 4x Jaycar 10" Paper cone

Two 2x4 MiniDSP Kit
Advanced 4-way crossover plugin

Sure Electronics ClassD 4*100 watt

Follow the design and testing process at HTGuide Forum.

Target frequency response of S14

S13 OB - Fullrange dipole tweeters

The back-to-back dome tweeter used in typical dipole setup is the achiless heel of the system. The on and off-axis frequency response is not smooth.

Many diy-ers investigated alternative approaches for the problem. Basically to keep uniform dipole radiation to very high frequencies, the baffle needs to be either very small or naked drivers are used. Having a pair of Tangband fullrange speakers I experimented with them.

The resulting result is smooth polar response up to 10khz.

While it looked promising, after a while I felt the whole system did not sound coherent. I have not understood the reasons behind it but it is perhaps due to:

- The transition of baffle size
- The fullrange driver unable to deliver the resolution required for high frequencies

I have sinced moved back from this design. The full investigation can be read here.