SP38/13
ScanSpeak 2-way floorstander with supertweeter
Copyright © Troels Gravesen

Why another 18W/8531G00 sliced paper construction? Didn't the SP95 and the SP98 do well? Why possibly a reduced cabinet volume? And why does USXX predict an optimum 22 litre cab for the 18W driver? Who is USXX and what has LspCAD to do with this?
Well, first of all, this construction dates back to the SP95 where I tried the Scan-Speak D3806/8200 mid-dome from 1600 Hz and the HIQUPHON OWI on top from 13 kHz. I only made a single cabinet from this set-up and never tried in full stereo. The measurements looked good and the sound from the single speaker was promising, so here we go with the SP38/13, a potent 33 litre floorstander providing deep solid bass from the 18W/8531-G00 bass driver. Overall system sensitivity is around 85-86 dB/2.8V and you need a min. 50 wpc amplifier to drive the speakers. The large 3806 dome goes low and releaves the bass driver of handling upper mid/lower treble area, an ideal situation for the bass driver. The cabinets are tilted 5 deg. to provide an ideal listening window and keep driver phase tracking at optimum. Due to crossover topology an even vertical dispersion is allowed.

The Drivers

The 18W/8531G00 hardly needs another introduction and for this application I have compared the damar-coated to the uncoated drivers and I still prefer the damar-coated. However, you may use the un-coated here without any changes to the crossover.
The D2010/8513 does a good job from 13 kHz. This may not be my favourite tweeter when we are talking two-way systems with a point of crossover around 2.5-3.5 kHz, but as a super-tweeter I have no objections. And it's cheaper than the HIQUPHON OWI.
The D3806/8200 driver: Well, this is the driver that makes this construction special. To be honest, few 6-7” drivers manage to perform well in the 2-4 kHz range. Most drivers will start rolling off around 3 kHz or they start beaming when the wavelength equals the membrane diameter and would be better off with a smaller driver taking over from around 1-2 kHz, but in commercial designs, cost is an object and the crossover may be more complicated and final production control more time consuming. Not a problem for the diy people. We have all the time in the world for our hobby. Well, cost may be an object, but the 3806/8200 is no more expensive than the 9500 tweeter and a cheaper super-tweeter could be used, like the Vifa D20TD-05-06 or a ¾” from SEAS. We “only” need to have a linear response between 10 and 20 kHz and this may be easier to find around 19-20 mm domes.
As mentioned in the Acapella files, it’s hard to ignore the history of Spendor. Spendor will gladly use a 1½" mid dome (actually the 3806/8200) from only 3 kHz in the SP1/2 construction. A 19 mm dome will not go lower than 4-4.5 kHz, and so on. The lesson to be learned from this practise is this: Always use drivers capable of reaching preferably 1-2 octaves below the intended operating range (for good reasons this doesn’t apply for the bass driver!).

Looking at the D3806 as a tweeter it appears huge. The pole piece is covered with a thin copper foil, symmetric drive,
to reduce eddy currents and reduce distortion. The voice coil gap is made from machined parts. Not an everyday sight.

Left: The frequency response of the D3806 driver (blue) is not that flat, but it is easily equalised to gain a flat response up to 15 kHz. This driver can almost be used as a stand-alone tweeter. I guess some damping material on the pole piece might do well.
Right: D3806/8200 impedance profile (green). The copper plating of the pole piece appears to do its job producing an almost flat impedance response from 1500 Hz to 22 kHz. The impedance peak around 900 Hz may be taken care of. So much for the drivers.

Box size and who is USXX?

Well, on a Danish diy-website a guy was discussing the possibility of setting up a large system: 2 x 18W/8531G00 + 15M/4531 + R29 (the horrific expensive D2904/7000 tweeter from Scan-Speak).
I shan't reveal the identity of USXX (his discussion forum identity) although he leaves no doubt he is an engineer working at Scan-Speak and furthermore refers to the 18W//8531 being "my baby". And USXX was also the guy from Scan-Speak giving a suitable explanation for inconsistency in production of the 18W/8535-00 drivers used for the 2.5 clones. (Yes, you can get his name here, page 38). So, nice that someone from speaker production takes time to give comments on these pages. First time I've ever seen this.
Interestingly, USXX recommends a 22 litre ported cabinet with a vent tuning of 33 Hz for the 18W/8531G00 driver giving the best step response according to his calculations/simulations - and from first hand experience having a three-way system with 2 x 18W/8531G00 for bass. The SP95/98 takes a 33 litre vented enclosure with a port tuning of 33-34 Hz and I've been pleased with the bass performance from this set-up giving a deep and articulated bass response. Obviously I had to try reducing cabinet volume and target a 33 Hz tuning. The SP95/98 test cabinet is almost made into section by the bracing and it was easy to blank off the lower part of the cabinet giving a volume of 23 litre and a new port (46 mm (ID) x 200 mm) was inserted giving a port tuning around 35 Hz. Not too far off target - the cabinet couldn't accommodate a longer vent. (By the way: USXX recommends a 28 litre cabinet for the 8535 driver with a port tuning of 30 Hz! This is not exactly the 2.5 clone prescription, but well in line with the Point75a construction. Making 25 litre cabs with a vent tuning of 28 Hz gave a more controlled and articulate bass performance.

Left; Green = 33 litre volume, vent = 72 x 200 mm. Red = 23 litre volume, vent 46 x 200 mm.
The vent tuning of the 23 litre cab didn't exactly hit the 33 Hz target, more like 35 Hz but let it be. As can be seen the damping of the 8531 driver is increased from going to 23 litre volume, i.e. smaller peaks (vent too small?) and a better phase profile = easier load on the amplifier.
Sinusoidal measurements (right image above) suggest we have more bass from a 33 litre cabinet - and this is also how it appears when listening to music, but it's very hard to determine which is best. What strikes the most is the difference in sound - there's an enormous difference in sound from the two drivers. The bass from the 23 litre cab may sound more dry and "fast", but I may use the phrase "fast" because I know it's supposed to be fast. The sound of a kick-drum from the 33 litre cab just sounds deeper.
The impact on midrange presentation is hard to ignore. Going from 33 litres to 23 litres has an immediate presence effect on the midrange and I'm not sure which one I like the most. From the 33 litre cab the midrange may sound a little dark and retracted where the 23 litres gives a very forward sound pulling vocals right up to the front of the speaker with subjectively reduced depth. Hmm.... This is tough! Actually I think I would prefer something in-between. Say 28 litre and same vent tuning..... We'll see.

Test cabs aside SP95 (middle)

More on sound

Making A-B tests from a variety of music on the 23 litre and 33 litre cabinets didn't give any conclusive results, but I'm in favour of the sound coming from the 33 litre cab being more round and deep. The upper bass and lower midrange appear to have more weight and remember, we're still dealing with a relatively small speaker here.150 square centimetres membrane area is the determining factor for the scale of sound we can expect from a basically two-way floorstander.
The general sound is just what it appears to be from the SPL-graph shown above. It's direct and there's a lot of detail - and three-dimensionality is indeed very good. As is often the case from speakers with very flat response profiles this is likely to produce excessive sibilance from certain recordings. Sibilance mostly is related to level, not quality, so in most cases this can be taken care of by proper attenuation. There are so many things the shown crossover does right, so I've been reluctant to make major changes to this basic set-up - and I won't even try making a series crossover for comparison. But read on... Plus/minus 1-2 dB in the 2-8 kHz range can make a world of difference to how we perceive the overall sound from a loudspeaker. In the good old days - which were not always as good as we may think - loudspeakers often had two L-pads for adjusting the midrange and tweeter level. L-pads were banished in any "true" hifi connection claimed of deteriorating the sound and maybe they did, I don't know. But depending on the electronics in front of the speakers and not to forget the room acoustics, most speakers often need minor adjustment of overall tonal balance to suit the actual set-up. But it has to be said that these L-pads were mostly used to produce a loudness effect, i.e. turning down the midrange - because the midrange is always the most troublesome area - and turning up the tweeter to give some highs. "Tizz-and-boom" - and without the help of the speaker constructor. But we're not going to use L-pads - definitely not.
Raising the 3806 series resistor from 1R0 to 2R2 removes some of the forwardness sometimes coming from this driver from less than supreme recordings. And the super-tweeter needs further attenuation for matching the upper treble level and 6R8 is replaced by 10R. So, now the overall balance is more in line with the BBC-dip philosophy. However, this may suit my ears and it's easy and no-cost to make your own preferred voicing from making these two changes. Valves - being more tolerable - may prefer the 1R5 or 1R8 to the 3806. Changing the 13-20 kHz region doesn't chance the sound noticeably, but for the sake of order....
I liked the sound from this set-up very much and was planning to make all the measurements needed for publication. I never was quite satisfied from the phase-tracking between the 18W/8531 and the D3806/8200. The SPL response was nice but inverting the polarity didn’t produce the desired dip around the point of crossover. So, time to analyse the whole thing a little more closely and doing so by the LspCAD, http://www.ijdata.com/, revealed that a 4^th order crossover topology might be better suited to meet the target phase tracking.
The "problem" with the LspCAD is that you in a short time can produce dozens of crossovers that has to be tested sonically. The good thing is that when your basic measurements of the individual drivers are done right, the software prediction is usually what you get. For once calculations and the real world appear to join hands.

Predicted response from crossover.

Actual response from crossover, 8513 not included.

The final cab drawings

After having decided on cabinet construction the following drawings were made and the tests cabs were modified to off-set the 3806 and 8513 tweeters as depicted above. Quite some task to modify the test cabs but worthwhile before handing over the drawings to the cabinetmaker. This time I won’t make the final cabinets myself.

SP38/13 crossover schematics.

Measurements
Measurements are not normalised for 2.8V/1 meter, thus do not reflect system sensitivity.

Left: SPL from individual drivers again. Black = 8531. This is a damar coated 8531 showing a dip at 3 kHz and a peak at 3.7 kHz and the lesson to be learned from this is to be prepared to all sorts of surprises when you apply coatings. No problem in this construction due to low point of crossover between 8531 and 3806, but this wouldn't work in a 2-way system with higher point of crossover. Blue = 3806 and red = 8513.

Right: Impedance of drivers, already commented above.

Left: Red = overall system response. Blue = same without supertweeter. Green = inverted 3806 polarity. Right: Wavelengths at 13 kHz are short, very short and moving the microphone just a little up and down easily creates dips at point of crossover between 3806 and 8513. Fortunately this is very hard to hear.

Left: Step response. Right: Vertical dispersion at +/- 5, 10 and 15 deg.

Comments to measurements:
As discussed before, don’t pay too much attention to what is going on above 10 kHz. The wavelengths are so short here that moving the microphone only one centimetre up and down will make a completely new picture.
I was pleased to see that the middle treble response was smoothed by off-setting the tweeters, so apparently there was something to the suspicion on wavelength and cabinet width interfering here. The overall impedance profile looks good. Minimum impedance is about 6 ohms and the phase angles are not severe. A fairly easy load I should say.

Sound:
As always, the most difficult part. This is a very neutral sounding speaker. Due to the very flat response profile a good tonal balance is heard from instruments and vocals. This is not always the case. Only a minor “BBC-dip” here. The bass is deep and articulate – exceeding the 2.5 clone. It’s also a quite revealing speaker and it will tell you right away the quality of your source material.
Getting the 8531 down to about 70 cm above floor level helped in giving the acoustic scenario a more realistic position in my listening room. I always felt a bit uneasy from having the ever so important midrange either too high or too low relative to my listening position.
This speaker has a great level of transparency and throws a wide soundstage and is going to be my reference for my two-way designs – because this is basically a two-way design with added supertweeter.

SP38/13 construction pictures

Here are a number of pictures taking during mounting the vent, adding damping material and mounting drivers and crossovers. A vent tuning discussion will appear at: http://www.troelsgravesen.dk/vent_tuning.htm, as I here have tried various flanged and flared vents to measure the impact on performance.

>
Left: Flaring of vent opening. Right: Alternative vent. Vent tube = 180 mm x 72 mm (ID), Fb = 36 Hz.
Right: Monacor BR-60TR vent: ID = 54 mm at inlet, 60 mm at outlet. Length = 140 mm. This actually works fine and produces a Fb of 34 Hz.

Damping material used. 4 mm self-adhesive bitumen pads added to all internal panels. This adds considerably to the weight of the enclosure. Above approx. 12 mm felt material. This is probably a mixture of cotton and polyester material made into felt sheets. This is what I'll be using for the SP38/13. Alternatively I use 10 mm heavy polyester foam also added to all internal panels.
Last but not least I use this great material - MDM-3 – a mixture of sheep's wool and polyester, also from Monacor. Used for additional damping at top and bottom of cabinet - and right behind the 8531 driver.

Left: One sheet of damping material cut into two, folded and placed behind 8531 driver. This may change during final voicing of the construction.
Middle: ½ roll MDM3 placed at top of cabinet.
Right: A look into the cabinet from bottom. No crossover applied yet. Vent is glued in place

Left: 3806 dome and supertweeter getting in place. Right: ll drivers, vents and terminals in place.

Left: Crossover for bass shaping up. Right: Attaching wires for bass.

Left: Bass crossover in place inside cabinet. Right: 3806 and 8513 crossover at bottom of cabinet.

SP38/13 SZC
July 2008