Working on dipoles - or semi-dipoles - was a new entry to my hobby and it took my plans into a new direction. Having lived with my semi-dipoles for more than a year, I have a hard time adjusting to any "boxy" sound again. The midrange sounds constrained with limited soundstage and articulation. The obvious next step would be the dipole bass, but I've had my share of electronic crossovers and multiple amps in the past and for the time being I want to keep things fairly simple: One amp and a pair of speakers with passive crossovers. That's all.

The launch of the Stradivari by Sonus Faber has created quite a debate and what has been argued against wide baffled speakers is some loss of pinpoint imaging. Here are some different views on the subject.

Mr. José Victor Henriques from his review of the new Sonus Faber Stradivari speaker:
"The Stradivari sounded like a live performance in a real venue, filling the room with sound and doing the proverbial vanishing act. Maybe less so because you just couldn't take your eyes off them. Given the larger than usual baffle, focus is not to the pin-point accuracy standard of, say, the Wilson Watts Puppies (their natural competition in this price range), suggesting some attention to “toe-in” to further sweeten the "spot". On the other hand I must confess I've never heard live “pin-point focus” as the instruments and singers usually energize the air around them creating a “bubble” of sound not just an “ideal” point source. In this particular aspect of sound reproduction, the Stradivari is closer to reality".

"Live pin-point focus": Has high-end HI-FI forgotten something here? The trend in recent years has been towards speakers with very narrow front baffles, trying to eliminate reflecting surfaces around the drivers, presumably enhancing pinpoint imaging. And with apparent commercial success. What is the trade-off? Because there's always a trade-off in speaker building. Well, by having a very narrow front baffle we will have edge diffraction at a rather high frequency giving delayed response subtracting from the target of pin-point imaging. We all know that a round sphere is the ideal surrounding for a speaker but for large speakers the sphere has to be large, very large and the WAF is close to zero. We never saw a commercial success from a sphere shaped speaker where the very narrow and slim line speakers can be rather deep (to give volume) and still appear smallish = high WAF.

Paul Messenger:
"Narrow speakers with a very wide radiation include more of the listening room and help create an illusion of bringing the musicians into the room; while more directional designs like horns and dipole panels give a precise view onto the recording itself. Neither one nor the other, the Stradivari is perhaps the ideal compromise between the two."

With narrow baffles we also run into serious baffle step compensation issues and we may have to apply large inductors to the basic driver in order to tilt the frequency response to be flat. Large inductors = large phase shifts. With a wide baffle the need for large inductors are reduced as we are dealing with a virtual 2pi situation. The "almost-infinite" baffle.

Henriques again:
"The virtual 2 pi radiating infinite baffle is based on the concept/surface of the “piano armonico” of the violin which allows the midrange unit to reach its lower frequency limit in a more poised and natural way, thus conveying to the listener a sense of better integration with the massive double bass drive units".

With a 66 cm wide baffle the Stradivari mid-drivers hardly need any baffle step compensation. It will be flat down to 200-300 Hz.

Roy Allison (former Acoustic Research and Allison Loudspeakers) in Stereophile, Jan/05) on the development of the Allison Model-1:
"I had emphasized dispersion in order to re-create as best as I could the performance-hall ambiance. I don't want to put up with a sweet spot, and I'd rather have a less dramatically precise imaging with a close simulation of what you hear in a concert hall in terms of envelopment. For that, you need reverberant energy broadcast at very wide angles from the loudspeaker, so the bulk of energy has to do multiple reflections before reaching your ear. I think pin-point imaging has to do with synthetically generated music, not acoustic music - except perhaps for a solo instrument or a solo voice, where you might want fairly sharp localization. For envelopment, you need widespread energy generation."

Calling a 50-60 cm wide baffle an infinite baffle is a truth with some modifications. Having an infinite baffle would mean mounting the drivers on the wall, creating a true 2pi environment. By making a wide and curved baffle cabinet, edge diffraction is avoided and a virtual 2pi radiation pattern is produced. Any driver mounted on a baffle will have an f3 = 11,600/width of baffle in cm. A driver mounted on a baffle of 20 cm with will be down 3 dB at 11,600/20 = 580 Hz. Making the baffle 50 cm wide the f3 is reduced to 232 Hz. Enough for a midrange driver working from 300-400 Hz to release its full potential without baffle step compensation. To read more about baffle step compensation and cabinet edge diffraction, try the following links:
http://sound.westhost.com/bafflestep.htm#bafflestepresponse
http://www.t-linespeakers.org/tech/bafflestep/index.html
http://www.speakerdesign.net/understand.html

Ken Kessler writes in his report on the Stradivari that it sounds like a dipole, so whether the midrange here will have to work as a true dipole is to be seen. Prototype baffles will have to be flexible to accommodate both vented midrange boxes and dipole arrangements. The sketches shown here are purely made to start thinking in curved, wide baffled constructions. No dimensions or calculations at this stage. A truly curved front panel is not an easy task and will probably require gluing several layers of thin MDF or plywood sheets over curved internal bracing.

Download Acapella WB file from here

Acapella WB
"Infinite" Baffle Loudspeaker
July 2005

Getting the Acapella WB test cabinets into the living room from the workshop (at the basement) was quite a problem and a sack barrow was needed! These cabs are heavy! And before I go on with this story I also have to tell that the current cabinet size and colour has a close to zero WAF. So, you are warned.

Patricia Barber from her "Paris Live" recording was first on the CD player and I think it's the first time I've heard the French introduction with the acoustic scale the male voice requires and the first time I got a real sense of the size of the room where the concert takes place. These speakers have a phenomenal ability to reveal room ambience. Should you worry about wide baffles having reduced transparency and pinpoint imaging, then you can happily forget about it. You can never really know how a person perceives the acoustic properties of a loudspeaker no matter how many explanations, but I do think I have an idea of what Ken Kessler means when he writes that the Stradivari sounds like a dipole without being a dipole. This is exactly what they do. I have been wondering what he meant by this and thought it was some smart phrasing, but it isn't.

The wide curved baffle has a benign way of projecting the sound waves into the room and I'm beginning to think that if you have to choose between reflection and diffraction, then reflection is the way to go. From a wide baffle you really don't have a delayed response of signals - at least not in the frequency domain that counts, upper mid and treble. And the baffle step really doesn't occur until the sound waves have lengths where they are omni-directional anyway.

As always, excitement is high the first time you connect a new pair of speakers. Will they deliver from all the working hours put into them - and these are just test cabs! If they are good I'll have to do it all over.....
The initial driver set-up was made from a non-standard 10" bass driver, the SS 15M/4531K00 for mid and the SS D2904/7000 for treble. The bass drivers' basic TS data are: Fs = 23 Hz, Vas = 280 litre, Re = 5.44, Qt = 0.37, membrane mass = 25 grams. Some very light-weight cones - and sensitivity is calculated to 90.8 dB/2.8V.
A vent was added to the cabinet 72 (ID) x 100 mm giving a vent tuning of 28 Hz. Quite low. Nearfield response suggests a quite linear response down to 50 Hz - not bad at all when you add the port response. These speakers should do well below 40 Hz.

The very simple series crossover shown in the pdf file was constructed for the MT and a ruler flat response was measured from 300 to 22000 Hz - the upper limit of the CLIO system.
After having found the subjectively best point of crossover between the bass and mid from using a 24 dB electronic crossover, I constructed a similar passive set-up. Not that easy as the mid is a 4 ohm driver and I don't want the impedance to drop below that. Eventually things worked out and a 3rd order (electrical) crossover was in place and all drivers could be connected with positive polarity.

My expectations to sensitivity were very high (Tannoys still fresh in memory) and I was initially a little disappointed from this until I realised how high I was playing. This speaker can play loud without apparent distortion and it takes an act of will to turn the volume knob down to normal listening level because the level of transparency is so good that you have to hear all the details on every piece of music.
This speaker does not have the absolute neutrality of the Acapella SE and the ability to portray any piece of music with the same grace and fidelity, but it has some more speed and sparkle that may suit some musical genres. Un-treated paper cones are rarely neutral and the SS 15M mid-driver and 10" paper cone bass are no exception from this rule, but the agility and transient attach is something that coated magnesium cones comes short of and something that the home constructor has to be very much aware of before deciding on drivers for a new project.

These speakers handle female vocals very well. They may not be completely neutral, but they can be played loud = lack of distortion. And going from solid state amplification (electronic crossover + 2 solid-state amps) to the Copland CTA 505 improved this somewhat. Sometimes paper cones and valves go hand in hand and this appear to be the case for the Acapella WB. I'm hesitant to coat the middriver. I know it will take some edginess, but also some vividness and well, colouration if you like. It's remarkable that paper cones have survived until this very day despite numerous alternatives with much better physical properties. But well produced paper cones have a "thing" that any other cone material hasn't.

In my small workshop the bass initially seemed a little low in level and extension, but it is not often you can hear anything from a 31 Hz warble tone, but this one can tell you it's there. And getting the speakers into the much larger living room certainly opened the bass and there was the level and extension. The transient attach of the bass could be better (kick drum) and I have to experiment with a smaller cabinet size and possibly a higher vent tuning before I look for other drivers.
Two other 10" bass drivers were tested, a 4 ohms ScanSpeak driver with paper-carbon cone + double magnet and a ScanSpeak 25W/8567-SE alu cone driver. The 4 ohms driver could almost provide the same sensitivity as the 8 ohms paper cone drivers but with low impedance in the bass region and the alu drivers provided some 3 dB reduction in sensitivity. For the time being I'm staying with the paper cone although the alu cone appeared to work very well.

First time living room set-up. Fresh coffee and a stack of CDs and LPs.

15-08-2005 Update Acapella WB

Seen above is a range of drivers suggesting quite some confusion. The 15M4531 ScanSpeak mid driver is gone; the 7000 tweeter is gone and in comes an AudioTechnology driver and a crummy looking old SEAS paper cone driver from the early Eighties. And a 12" JBL bass driver! There's always a honeymoon with speakers. You may instantly recognise strengths and weaknesses from a new set-up, but it can take days, sometimes weeks to find out whether you can live with the sound from a driver and the way this particular driver will colour the sound, because all drivers do.

The 10" paper cone bass drivers did great and the 7000 tweeter also, but the 15M4531K00 driver will drive you nuts in the long run. Scan-Speak has done a fine job in elimination major break-up patterns usually associated wit paper cones but this does not mean this driver doesn't have some residual "paper" (honky) quality to its sound. This driver is phenomenal in accentuating sh... sounds and every sibilant recording will shred you ears. After some time the 15M driver was coated with a coating material obtained from Speakerbits in Australia: http://www.speakerbits.com/Default.aspx. Actually the glue used for refoaming drivers. This material is well suited for coating paper cones and leaves a nice, non-sticky, flexible surface if done properly. This treatment did not change the frequency response and it improved the sound considerably. For some days I thought this was the cure, but something just kept these drivers from having a neutral presentation. One very noticeable thing is when a high piano note is played and the 15M will smack in your face, but the note doesn't stop fast enough. It's like the 15M is ringing. Does the combination of a high mechanical Q, the kapton voice coil and the paper cone produce this effect? Is a middriver with better damping what's needed?

Tweeter attenuation was tested thoroughly and a new parallel crossover for the MT was constructed but this didn't perform better than the former series crossover. Out of frustration I inserted some old SEAS paper cone drivers (actually from a B&O speaker, S25) and well, these drivers do not have the level of transparency compared to the 15Ms, but vocals just started sounding right, actually very good. Hmm... Seriously, a 25 year old el-cheapo driver! Next I took a Monacor SPH 175 driver, actually quite similar to the SEAS driver. This driver has a coated paper cone and I removed the terrible plastic dust cap and inserted a phase plug from an old SEAS driver. Again, some very enjoyable midrange and good vocal performance.
Replacing the 15M with these new drivers took some extensive crossover modifications every time. There's no plug and play in this business....I've had good times with the Scan-Speak 15W/8530-K00 and -K01, but this 15M/4531K00..... Sorry, not my cup of tea.

So it was time to reconsider the whole project.
A pair of JBL 123A drivers in mint condition caught my attention at eBay and I've always wanted a pair of these and so I had - for a reasonable amount of money. Not often you get two beautifully build, mint condition 12" bass drivers for 150 US$. I knew these bass drivers would have a relatively high Qt and probably would perform best in 80-100 litres closed cabs and they might be worth a try in the Acapella WB. My measurements on the 123A: Vas = 225 litres, Fs = 25 Hz, Re = 4.4, Qt = 0.49.

JBL 123A Signature. This driver has a remarkable frequency response. From the wide baffle cabinet they perform +/- 1.5 dB
from 200-6000 Hz! No serious break-up at all. No wonder this driver can run full-range in the JBL L100 Century monitor helped
at the top by the LE5-2 midrange and LE25 tweeter. And the 123A drivers are sensitive, 90-91 dB/2.8V.
A good starting point for the Acapella WB.

Next test set-up.

The midrange driver:

Midrange - the ever so troublesome midrange! I always thought that the Acapella should not be a "true" clone of the Stradivari with regard to driver selection - it would be too easy to take 2 x SEAS W26EX001, a 4" AudioTechnology mid and the 7000 tweeter to get as close as possible to the Stradivari. But the target is to maintain 8 ohms impedance and so it will be.
But having some 4H52-06-13 SD AudioTechnology drivers, these were an obvious choice for a new set-up. The 4H52 drivers are very broad-banded and can easily perform up to 3500 Hz and the Fountek JP3 ribbon tweeter is feeling good from this point of crossover (p.o.c.). Thus a new set-up was in place as depicted above.
The 4H52 drivers used are not standard drivers; they have an underhung voice coil, only 6 mm wide in a 13 mm magnet gap. The membranes are made from 500 my polypropylene and they have a concave centre "dome". And they obviously have symmetric drive (SD).

The 123A bass driver was tested in a vented and closed version of the cabinet and vented so far sounds the best although the difference is small. Vent tuning is 25 Hz. A three-way, 24dB LR electronic crossover was inserted and with the CLIO measuring system a reasonably flat FR was obtained with points of crossover at 400-450 and 3500 Hz. Finding the right point of crossover between bass and mid driver is the most tricky part and as the electronic crossover has continuously adjustable points of crossover, it's a good experience to slowly change the point of crossover from 100 Hz to 800 Hz while listening to vocals. Taking the p.o.c. down to 200 Hz, this set-up starts sounding like a pair of satellites with added sub-woofer. Not good! Moving up to 600-700 Hz has a negative impact on vocal performance as well, although it's difficult to describe how. Vocals just don't sound right from the 12" driver. Looking at some large classical British speakers like the IMF, Radford, etc., we will often find a p.o.c. around 300-475 Hz from using a 10-12" bass driver and a 5" midrange. (The 4H52 really is a 5" driver having a cone area of 95 cm^2).

Listening to a wide selection of recordings, this set-up proved worthy of constructing passive crossovers targeting these points of crossover. From the electronic crossover the bas and mid produce a good frequency response, but the individual phase response wasn't too good. Something that may be more easily managed with a passive crossover and produce an even better bass-mid integration.
Frequency and impedance files were produced from all the drivers; the 123A measured at ½ meter distance and attenuated 6dB. This usually goes well for modelling. The benefit is a much lower reliable FR for the bass driver for modelling.

February 2006:

After a long break from the wide baffle speakers, I finally put together the crossovers for the JBL 123A + AT 4H52 + Fountek JP3 ribbons, based on the simulations I made half a year ago and initially the points of crossover were 300 and 4500 Hz. 4500 Hz is really high for the mid but I wanted to hear how the 4H52 would perform from this. The 4H52 can manage well up to 5-6000 Hz, but treble produced from a 5" cone isn't my cup of tea. The 4500 Hz p.o.c. went reasonably well despite minor honkyness, but transparency was excellent. Actually so good that I decided to go for this solution. Polypropylene cones may not reveal the absolute ultimate level of transparency but this is so close to "it" that I decided to pay Per and Eivind Skaaning a visit to have them make me two middrivers with some differences to the 4H52 FlexUnit here. The drivers will have 52 mm voice coils of 15 mm length wound on a kapton former in a 6 mm deep magnet gap, standard convecs domes and finally it will have the H magnet. With the given choice of polypropylene thickness, a sensitivity around 91 dB should be the result. The driver will have the C-Quenze chassis and it will be named 15H52-15-06-SDK. I very much thought that underhung voice coils would be the way to go here but the transient attack is better from overhung voice coils and the frequency response produced from overhung voice coils is better suited for midrange. Underhung voice coils may produce a more flat frequency response well suited if the driver also has to play bass - despite limitations in Xmax.
So, while waiting for the drivers being made it's time to think about the whole project again. The test cab "as-is" is definitely too large to render a green light for living room placement. 45-50 cm width is OK but height has to be reduced to between 100 and 110 cm. Depth is not a major issue, which is nice as we get a lot of volume from increasing depth even a little. But the JBL bass driver? Good question. The JBL 123A drivers do fine with the current 80 litres volume but a smaller cab added Variovents may do fine too. Having Per Skaaning make 4 x 8" drivers suitable for a total of 60 litres volume would be easy - and costly! And we could order 10 ohms voice coils to produce an impedance that wouldn't kill our valve amps.
A very well-known high-end system having two 8" bass drivers in an approx. 60 litres cabinet get along from using some rather unusual drivers: Fs = 35 Hz, Vas = 41 litres, Re = 5.4 ohms, cone weight = 35 grams and Qt = 0.40-0.45 (depending on voltage applied). This makes an F3 around 40 Hz in a 30 litres vented cabinet and the room does the rest - I hope, because the company claims a 25 Hz extension! Well, well... These bass drivers have a rather rigid suspension. Feels like drivers for subwoofers used in cars actually. But who cares - if it works. I have some 10 ohms Dynaudio 21W54, but will two of these go deep enough? Only experiments can tell. If the new suspensions they're having will reduce Vas a little and increase Qt a bit - it might work.

March 2006, middrivers have arrived:

C-Quenze 15H52-15-06-SDK. Kapton voice coil former.
Actually the label says Flex-Unit, however a C-Quenze chassis is used.

Frequency response of 15H52 driver from various baffle widths:
Red = 20 cm, blue = 30 cm, green = 40 cm, yellow = 50 cm, purple = 60 cm.
Baffle curved.

May 2006:
Evaluation the Acapella WB project

After testing a wide range of bass drivers, the JBL 123A still outperformed the whole lot and for various reasons, I don't want to use this driver in the final version. My preference would be two 8" drivers with 10 ohms voice coils, suitable for a 60 litres cabinet volume, but I don't have these drivers and for the time being I'm not prepared to have them made either. Too costly.
The 15H52 middriver will most likely make a fine match with the old SEAS 33FWK-A bass driver and this will be the next thing to try.
Another thing is that these very wide baffles are unlikely to ever fit in with our living room. So a more conventional cabinet with a semi-wide baffel for the mid and tweeter will be tried.

Maybe like this.......Not a particularly small cabinet, but visually much smaller than the Acapella WB.
Project closed until.....