Gainphile: 2009

S12 OB - Open baffle simplicity

A true open baffle with its simplicity (no H/W/U -frame etc.) I have observed that there is something about them which other variations simply couldn't match due to very uniform polar response.

The tradeoff is absolute maximum bass output. A dipole bass array utilising a total of 8x 10" woofers could be added to overcome this limitation. Then the system effectively become a 4-way as well as requiring more space.

Follow the discussion at diyAudio forum here.

Also similar discussion at Stereo.Net Australia here.

Basic measurement data is available here.

System: Active, 3-way with response shaping and time alignment.
Tweeters: 4x Hivi K1
Midrange: 2x Seas L21 RNXP
Woofers: 4x Jaycar CW2119

System's strengths: Coherency, transparency, very spacious and natural-sounding.

System's weaknesses: Colorations due to L21's metal resonance harmonics are audible on certain material, outright bass output. Back-to-back dome tweeters are never ideal dipole radiators.

S12 Dipole Crossover

Click on the image on the left to see the functional diagram of the active crossover. The crossover can be quickly implemented if you have the budget to procure digital XO such as DCX2496. I use analog op-amps and prototyping board.

Much of the functions are clearly explained at Linkwitzlab website. What is unique is the need to compensate early rolloff of the Seas L21RNXP. The midrange exhibit some coloration due metal cone breakup harmonics and this would be the same even if you use the more expensive W22EX. It is not always audible and only appears in certain recording material.

Seas L21RNXP is an excellent unit aside from the normal metal resonance, however they have been discontinued. Seas L22RNXP is the drop-in replacement. Forget exotic drivers such as the W22, you may as well build the Orion. Why use metal cones then? Well the audible transparency of those drivers is unmatched. They are pistonic until that breakup arrives.

Hivi K1 tweeters exhibited very low distortion and excellent dispersion when crossed at 2.2khz or above with LR4. The tweeters can easily be substituted with something else, as long as the units have excellent distortion and dispersion. With those criteria it's hard to beat the cost of K1 ($12). Dispersion is very important, for example the popular Vifa XT25 would not sound as accurate due to poor dispersion. The venerable Seas 27TBFCG would be a good candidate and the cost is still somewhat reasonable. Better tweeters may allow you to cross lower and help reduce midrange breakup coloration, but not much. Ideally 8" metal cones like L21/22 or W22 should only be used below 800hz.

Construction

Building the S12 is straight-forward, however there are some important considerations.

The side panels are not to increase acoustic path length as normally done by dipoles with "wings". They are for structural reasons, and generally the narrower the better. Although they are very narrow indeed (4.2cm) the midrange section should not be obstructed. I have observed audible midrange coloration if the side panels are simply extended through the midrange height, although I do not see it in the measurement. This could be a very high Q resonance which are hidden by diffraction effects in measurements.

The cavity for rear tweeter must be blocked, otherwise it forms a resonator. Stuffing the cavity using polyfill only helped marginally. It has to be angled to minimise diffraction issue with the midrange.

Link to other designs

Updates:
16/4/2010

Removed comments and pictures about sealed bass, DX25
Added observation notes on strengths and weaknesses

Updates
:17/4/2010

Added Dipole XO diagram and Construction

Updates:
3/11/2010

Implemented DSP XO (MiniDSP.com)

2x MiniDSP kit
1x 4-Way Advanced Crossover

For a total of $210 this is a truly great value for money. Refer to my investigation which led me to this path. The quality is outstanding and they are sonically transparent. Be wary in using digital attenuation for driver level matching. I use trimpots instead.

XO points (acoustics):
W-M: 200hz, 2nd order
M-T: 2khz. 4th order
Built an 8-Channel Class D Amplifier to drive the loudspeakers.
Another value for money. Total build cost for this 8-channel amplifier is $160 (!). Even cheaper than chip based amplifiers and there is no soldering involved.

Dipole Bass Array

Doubling the number of woofers provide further +6db of maximum output. Basically they are vertically-stacked W-Frame open baffle speakers. The arrangements allows the front baffles to be kept narrow.

Follow the discussion at LinkwitzLab forum here. There are some cosiderations that needs to be investigated carefully, for example the effect of gravity to cone acceleration.

S11 OB - Desktop Open Baffle with Seas 8" and Tangband Full-range

Drivers: Seas L21RNXP, Tang Band W3-1364s (bamboo fibre cone)
Frequency: 40Hz-20kHz. Designed to be listened near field (ie. not much SPL).

Build process here.

The speakers are really small as seen above. Very pleasing.

About the bass... the optimum compromise between distortion, listening level, and excursion is 40Hz. That's pretty deep, dipole or not. The woofers are working hard. Some 30Hz can be felt but not respectable. Below 30Hz is just asking for trouble. These metal cone makes horrible sound when bottomed. It's simply just physics and there is no way I could get more, except putting again a pair of 15" on my desk.

The distance to rear wall is about 35cm. I could not hear any faults. Phantom image is convincing, almost like the speakers making no sound and the music comes out from behind the wall.

I have small clip taken using my mobile phone in YouTube.

Update 14/7/12
- Migrate ASP to DSP
- Baffle strengthening

Next:
- To document the project

S10 OB - Seas L21RNXP on Open Baffle

After completing my last iteration, I really don't feel the need to experiment with different speakers / OB topology. Any improvements I thought, would come from the drivers themselves which unfortunately is about $$$, not engineering. I should also start to make a PCB and perhaps paint the speakers or something like that :)

But when I saw someone is selling a pair of Seas L21RNXP really cheap on ebay I couldn't resist having a go at rigid piston driver. The Seas drivers are beautiful indeed, and this is not even the magnesium series. They were old types of drivers but never used. The newer ones are L22RNXP.

What I would like to gain from this one are:

- Ability of the midrange to cross lower than P13WH. Let's say 120Hz, presumably better integration with the woofers.

- Transparency of rigid cone drivers. Is the claims true?

- Much higher SPL level

And some drawbacks I would expect/investigate:

- The need to cross the mid to tweeter lower, possibly straining the tweeter.

- Would the increase of baffle from 19cm to 24cm affect the polar response in a really bad way?

- Metal cone resonance

Build process is discussed at diyAudio. See interesting finding here. Measurement data available and discussed here.

S9 Open Baffle Speakers

Early version of S9, circa July 2009

This build is to revisit the previous design which sounded most natural and smooth. So much that the further effort to obtain dipole bass quantity had to be scrapped.

Build process is discussed here.

Driver-baffle relationship is very important. If the inherent response is not smooth, no amount of EQ would result in natural sound.

The smoothness of this design can be summarised by differences of frequency response between on and off-axis measurements:

Update 23/8/2009
Finished dipole H-Frame woofers:

Update 10/10/2009
Tidy up the prototype XO

Update 14/6/2010

Updated active crossover topology

W-M : 300Hz, 2nd order

M-T: 2.1kHz, 2nd order

Update 17/1/2011
- Converted Analog XO to DSP
- Changed H-Frame to W-Frame subs
- Measurement from recent GTG:

Normalised plot at 0 degrees, showing directivity pattern. Here we can see that the back to back dome tweeter is the achilles heel of such setup:

Update 29/5/2011
- Converted analog XO to DSP (MiniDSP)
- Use W-Frame woofer configuration. This opens the possibility of stacking 4 woofers/side.

Update 8/6/2011
Completed S9R, an "R" or rigid-cone version of this project. Fountek FW146 + Vifa DQ25. The phase shield of the tweeter needs to be removed to avoid resonance and stored energy.

Non-normalised sonogram of S9R, 0 - 90° taken with 11° increments.
I moved these loudspeakers in to my future "audio room" here: http://www.youtube.com/watch?v=RnaTs8oK8tA.
The iPhone does not record bass well unfortunately.

S8 Open Baffle Speakers

4x HiVi K1 Tweeters
2x Vifa P13WH-00-08
6x Jaycar 10" woofers

Fully dipole radiation. LF equalized using Linkwitz transform to -3db at 20Hz.

Build process here.

System compromises :
- 5" mid still acoustically too small for 30cm baffle, thus the first dip is 1kHz
- HiVi K1 can only crossed comfortably 1.8kHz upwards. A 1.4kHz tweeter like Seas Millenium would be nice. But at $350 each they're way out of my league.

Update 20/7
The first compromise, P13 being acoustically too small for 30cm baffle is more serious than I thought initially. I was unable to get smooth midrange response despite multiple stages of equalization/notch. I have since scrapped this design and returned to S7-styled baffle, yet with further improvements.

S7 Open Baffle Loudspeakers

The S7:
- Fully dipole radiation from 35Hz - 20kHz
- Vifa P13WH midrange
- Eminence Alpha 15"
- Fully active 4th-order Linkwitz-Riley crossovers M-T at 1.5kHz
- 2nd order LR W-M at 300Hz
- Time aligned using phase correction
- Tri-amped using 3 pairs of LM3875 Gainclones

Follow the design process here.

Practical OB design practices

What has been learnt. Slapping a full-range driver on a wide baffle just doesn't work, unfortunately.

Drivers

Operate all the drivers at their pistonic properties as much as possible. This is roughly below their beaming frequency which is 1/2 wavelength of the effective cone diameter (0.5 * 345 / diameter) where 345 is the speed of sound in m/sec.
Baffle size

- All dipoles will exhibit a dipole peak. The 1st peak can be tamed using notch filter, but the next peak-and dips are not. The idea is to use the response before and just above the dipole peak. The consequence of this is usually the need of tweeter which can be crossed low, at about 1.4kHz.

- The baffle size should be less than 2.2x effective midrange cone diameter. This way the dipole peak will be pushed higher in the frequency response, ready to be crossed with a tweeter. Using very wide baffle will push down the peak (more bass) but also put the peak-and-dip in tweeter crossover region.
Dipole peak equalization

- Get 1m gated impulse measurement. Then select using 10ms time window. Notice the 1st peak. (Edit: I have since trusted outdoor, ungated measurement more. Raise the speakers to at least 2m from the ground and measure from 1.5m).

- Obtain Fo, Q, and depth of the notch filter. This can be really easy using JohnK's spreadsheet , but not entirely required.

- Build the active notch filter. The components must be as close as possible, and better yet simulate them using Spice.
Time alignment

- Measure on-axis the woofer and tweeter, both at exact same distance from baffle surface e.g. 30cm. Measure using impulse response.

- Select same time start and time window, e.g. 5.5ms with window of 6ms. Compare the phase of the woofer and tweeter, obtain the degrees difference. Then calculate:

e.g. xo freq = 1500, difference = 43.6 degrees. Then:
For LR4 the difference should be 0
Phase = 43.6/360 = 0.1211
time = phase/xo = 0.1211/1500 = 0.0807ms
equal distance = speed of sound * time = 345 * 0.0807 = 27.855mm

then create the circuit based on equal distance.

Open Baffle (OB) Loudspeaker Articles

Collection of well-written articles and investigations about Open Baffle Loudspeaker design:

Advanced Articles:

Measured Monopole and Dipole room response
By Elias:
I did some measurements with monopole and dipole bass in a room.
Particularly I try to pay attention to temporal behaviour since as I believe steady state measurements in this case are pretty useless when considering human perception. Also I try to match the measurement excitation signal to represent the final situation, that is music.
Music is no impulse nor steady state sinusoid, thus I'm using short tone bursts with shaped envelope. It will give quite good visuality to what is happening in the room due the reflections.
Monopole vs Dipole bass
Elias gives emphasis on temporal fidelity of the reproduced signal. Real music signal contains time domain energy variations that must be reproduced accourately before high fidelity is achieved. This is also true for bass.

Past Projects

Various speakers I built in the past which I remembered to take pictures of, not in any particular order.

Why build so many speakers? Because only transducers and speaker topologies makes acoustical difference to a hifi sound. Not CD players, not DAC, not Amplifiers, and most definitely NOT cables!

S11 Bookshelf OB speakers.

S9 with H-Frame woofers

Two 10", Vifa P13, and tweeters in dipole configuration
Sufficient output for a desktop OB!
Sept. 2008

Same thing as above before painting

Building Pluto Clones.
May 2008

Trapezoid 3-way OB speakers.

One 10" woofer was not sufficient in terms of bass output

How would dipole-monopole hybrid sound like? In short...not good.
May 2008

Experiment with U-frame dipole woofer.
Good output but not so good sound, most likely due to distance to the wall.
April 2008

What the H-frame setup look from the front.

Trying out H-Frame baffle.

Also not too good due to distance to rear wall avoid dipole radiation pattern.
April 2008

Small dipole desktop speakers. Passive XO. There's no bass output.
March 2008

The small desktop dipole from the front.

The woofers are taken from $12 dicksmith bookshelves.
They eventually used in the Pluto clones.

My first amplifier build for almost 20 years of HiFi hibernation.
Oct. 2007

My first ever Active Crossover.
Dec. 2007

"Number One". First dipole speakers!
It was finished in Christmas eve of 2007.

Very simple crossover.

Number Zero. The first time I ever heard of dipole speaker was this one, and not even a pair. I knew right away that this is 'the way'.
Dec. 2007

The drivers were from kerbside rubbish and the baffle was from unused kitchen cabinet.

Eminence alpha 15" + $4 Jaycar midrange.
Nov. 2008

Investigation on influence of driver offsite (time alignment).
The experiment was failure as the resonances kept from actually making good judgement.
Oct. 2008