Eight plus One
Copyright 2011 Troels Gravesen

2011 Christmas holidays' experiments.


21W/8555-01 and no-name polypropylene driver of similar size.

A client asked for a Wilson Audio Duette type of speaker made from sliced paper 22W/4851T00 drivers, probably the only pair in existence, and the 710003 tweeters. Hmm... your money, I thought! You know point of crossover has to be taken really low to make this happen? Not really his problem... 
I've been studying the Duette and based on the measurements shown here, it looks like a speaker specifically voiced for book-shelf placement. A broad dip in the 100-500 Hz range and a rather high vent tuning making a decent response down to 50 Hz based on reviews. The vent tuning appears much too high for near-wall placement, but I reckon Duette customers will not only use this on a book-shelf, rather as a stand-mount. Making a speaker ideally suited for both scenarios is almost impossible and a couple of resistors to tune bass and treble is unlikely to be perfect in either situation.
Nevertheless, tuning a similar set-up for stand-mount could be an interesting project, thus a very simple 2nd order crossover was made at 1.7 kHz and after some fine-tuning of tweeter level it was time for evaluation.
The 22W bass driver is a rather "deep" driver, i.e. it has a deep cone and run full-range it has a distinct "cuppy" sound together with the inevitable beaming due to the large diameter of the cone. However, no severe peaking in the 2-6 kHz range. It has to said the sliced paper version didn't measure much different compared to what is shown at ScanSpeak website for the non-sliced version.
Taking the 7100 down to 1.7 kHz was my biggest worry. Despite its wide surround it's still a 1" voice coil and from a simple 2nd order filter it's on a tough job.
Bottom line is that this turned out much better than anticipated and Siri's Killer Note was reproduced as good - if not better - as any other speaker I've made. This was seriously disturbing! A 2nd order filter at 1.7 kHz, not particularly well performing with regard to phase integration! Hmm...Never say never! I more and more think this "note" problem has its solution in lack of beaming and sufficient membrane area. Crossover topology and proper phase integration appears of lesser importance.
I had these disturbing results simmering for quite some time and while cleaning the attic I came across my AT-SW cabs. With their 21 liter net volume they could make a range of eight-plus-one monitors from some of my stock 8" drivers, thus this reporting on some experiments to see what can be achieved from a very simple design using basically 2nd order filters and a tweeter being able to go really low.

Thus, ScanSpeak 21W/8555-01 drivers were dusted off and ready for action. Why not? These paper pulp/carbon fiber cones usually have an excellent midrange, at least the 6" drivers, and I noticed Sonus Faber recently introduced their latest version of the Guarneri speaker still using these air-dried composite cones. Despite the trouble of producing these cones and the frequent inconsistency in performance they can do things few other cones can do. They combine rigidity and decent internal damping.
The 21W/8555-01 has a less deep cone geometry compared to 22W and if not taken above 1.5 kHz we are somewhat below a rather peaky frequency response around 3.5-5 kHz with a 2nd order low-pass filter. Time will tell if this is enough.
Next, NNNN is another driver 8" fitted with a PP cone, an OEM unit and I shan't reveal its origin, only it's one of the best 8" PP drivers I've come around.
ScanSpeak 2904/710003 is one of the most recent in the Revelator series and due to a 4 ohms voice coil and a handful of strong neodymium magnets, it displays a whopping 94-95 dB sensitivity, hardly needed here, but from my client's speakers, I know it works.  


D2904/710003 tweeters


For front panels, 22 mm MDF and 12 mm Baltic birch sheets were laminated.
I like my router and a lazy Sunday afternoon is not badly spent routing some driver holes! 

Hope the pics are self explanatory. The images here shown in sequence of routing.
The damping material is be placed around tweeter is 7 mm thick, thus 7 mm depth of routing.


Don't forget "ears" for tweeter's terminals - and chamfering bass driver hole.

Trying out driver rebates.

Wilson Audio often implements tweeters in ways most diy'ers would never do. Mostly recessed ~10 mm and surrounded by damping material having square holes equidistant to the center of the dome or as here, having a hole in damping material equal to tweeter diameter and further added some star cut felt directly on tweeter's faceplate. I wanted to see how bad tweeters really measure from these surroundings because in terms of frequency response, it's not ideal. The few frequency response graphs we have from Wilson Audio speakers really do no look nice, but Wilson doesn't seem to care and value lack of early reflection more despite diffraction issues.

Tweeter damping material. 7 mm foamed rubber and 3 mm star cut felt on tweeter faceplate. 


Click image to view large.
My suggested cabinet design based on PE 302-730 dimensions (0.75 cubic feet).
Place bracing across cabinet height as seen on sketch. Please don't ask for further dimensions.
The single vertical brace in the PE cabinet really isn't enough. It needs bracing between front and rear. 


As can be seen the PP driver allows a simpler crossover.

It may be worth while experimenting with C1021. Here simulated response from NNNN + 710003 drivers.
Please read review here on Chario Sonnet and voicing from a crossover of similar design.  

MEASUREMENT 21W/8555-01 + D2904/710003

The bass drivers peak at 3.5 kHz is really nasty and shreds your ear even on the MLS signal. It needs serious suppressing!
Tweeter level here from an average of 4 measurements at 0, 10, 20, 30 and 40o.

Left: Response of final system normalised for 2.8V, 1 meter. System sensitivity around 85-86 dB.
Right: Response at 0, 10, 20, 30 and 40o. What we have here is an incredible wide dispersion field having a significant impact on sound.

Left: Response of final system from an average of responses at 0, 10, 20 and 30o. Not bad at all!
Right: Impact of mid notch filter. Without the filter a much too forward sound.

Left: Response of individual drivers and summed response. Point of crossover = 1400 Hz.
Right: Final system impedance. Minimum = 5 ohms.


Left: The NNNN driver has a much better basic response and an excellent roll-off making crossover work much easier; no notch filters needed.
Right: Extending front panel below bass driver makes an even better midrange response (black graph).

Left: SPL from drivers driven from crossover plus summed response (black).
Right: System step response; this is anything but a time-aligned system.

710003 tweeters' distortion measured at 2.8 volts input equivalent to ~95 dB/1 meter.


For both speakers no discontinuity going from 8" to 1" is noticeable. Smooth all the the way. What's immediately apparent is how easy it is to move around the listening room and not loosing level on upper-mid and treble, which is a key feature of designs like TJL3W, SP38, SP44, Ekta and 3W-Classic.
Bass response is firm and solid from both speakers despite lack of lower octave but as a lot of music really doesn't contain much below 40 Hz this is not an immediate concern. The 220 cm2 membrane area does its thing with no sign of loosing grip on the music.
Both speakers really don't make more than ~85 dB sensitivity for 2.8 volts and coming from DTQWT and Jenzen Accu being around 10 and 5 dB more sensitive respectively, this is something that is immediately felt: The lack of dynamics. Both speakers are power hungry and 100 watts are minimum to get these speakers swinging.

To summarise, both speakers offer solid bass from a small foot-print and a wide dispersion with absolutely no hot-spot. Placed on a book-shelf, the port can be stuffed lightly to make an aperiodic tuning better suited to cope with the inevitable room-gain. Both speakers perform surprisingly well with regard to soundstage depth and transparency despite less than ideal phase integration and lack of time-aligned drivers, features they both share with the Chario Sonnet speaker. A really good tweeter doing well down to 1 kHz is a key ingredient in getting this system right and from the distortion measurement show above the 710003 does exceptionally well down to 2 kHz only showing slightly increasing levels down to 1 kHz, nothing serious with a point of crossover at around 1.5 kHz.

The 21W/8555-01 does surprisingly well, but the suppression of the 3.5 kHz point of resonance is not really good enough for not having a somewhat rough presentation of strings and female vocals from time to time.
The PP driver does significantly better and had this driver been available to the diy community, I would have had no hesitation in its full recommendation.

The study here is only meant for demonstrating proof of concept and with the right drivers it's a viable option for making a potent two-way system from simple crossover topology. Obviously you need good measuring equipment and simulation software to get it right and the final trick is to tune the 1-5 kHz range due to the extreme dispersion of the tweeter; a key factor in voicing the speaker to render a proper perceived balance of basic notes and overtones.