Waveguide for ScanSpeak D2904/710003 tweeter
Copyright 2012 Troels Gravesen

Modification of waveguide   Measurements           


ScanSpeak D2904/710003 tweeter.

I have a box of Monacor waveguides cut to various depths for a quick test on tweeters to pick those suitable for applying a waveguide. Just throwing in a waveguide on a tweeter mostly doesn't work. Some tweeters may require a shallow waveguide, some tweeters a deep one. 
The initial part of the waveguide, the throat, is critical in the way the tweeter couples to the air and loads the "horn". There's no clear distinction between a horn and a waveguide but if a waveguide is deeper than 30-40 mm, I would start thinking of it as a horn. With the TW034 tweeter a very shallow waveguide was possible and counts for its lack of any "horny" sound. The shape of the dome loading a long tube may cause phase irregularities and count for aggressive treble performance, thus compression drivers are often fitted with phase plugs meant to load the horn properly.
A horn provides an acoustic impedance match between the driver and free air and improves the efficiency with which an e.g. dome loads the surrounding air. As seen from the measurements below we have some 6 dB increase in response around target point of crossover, thus when equalised properly, we have a significant reduction is distortion.

The ScanSpeak tweeter is a high-sensitivity tweeter displaying an impressive 95+ dB/2.8V sensitivity due to the low-ohmic voice coil and strong neo magnet, thus a potential candidate for high-efficiency speakers. 
Initially I picked a shallow waveguide taking advantage of the 710003 faceplate's initial rounding towards the dome. This turned out really bad and produced a serious dip around 10-12 kHz. Thus, a deeper waveguide was tried and produced the results seen below. Really nice!
The D3004/660000 tweeter has a different faceplate and does not take the here shown waveguide, just be ahead of that question. You could use the 7100 faceplate on the 6600 tweeter, but I don't think you can buy one from ScanSpeak. In that case you will have to modify the 6600 faceplate to fit the Monacor waveguide. I have no measures for this and no plans for doing so.

Before you rush to buy a pair of 7100 tweeters and waveguides, think about implementation with your other chosen drivers. Making a crossover for a waveguided tweeter is much more difficult than a properly placed tweeter on a flat baffle or properly faceted baffle. The latter may produce a nice flat response easing crossover design. For the 7100 tweeter here the response at 3 kHz is some 6 dB higher compared to 1 and 10 kHz, thus good measurement data and simulation software is necessary to get it right. Quite often it means we can use a much smaller series capacitor, not unimportant when we apply for super caps. Another possible feature is that the acoustic depth of the tweeter is increased the same as the height of the waveguide, e.g. 20-35 mm, which may be the same acoustic depth of the middriver, thus allowing experimentation on phase and time-coherent crossover implementation.


Modification of Monacor WG300 waveguide

Monacor waveguide, item #WG300, designed for the Monacor DT300 tweeter.

We need to make two spacers to accomodate the ScanSpeak D2904/710003 tweeters.
Here I used 5 mm MDF. Outer and inner diameters are 116 and 50 mm. 5 mm plywood works as well.

Either saw off the threads or make four 9 mm holes to allow the flange to rest on then ribs.
Fill cavity with filler, e.g. cheap acrylic filler or Superfix as used here.

Smooth filler and press flange towards the ribs. Apply filler round the throat and smooth with finger as seen on photo.

Make a gasket from acrylic felt or similar fabric and mount the driver. Use washers not to damage the rubber gasket of the tweeter.
I used 4 mm wood screws. Drill 2.5 mm holes before fastening the tweeter.


On my prototypes I used 5 mm plywood and shaved the treads to mount flange with four countersunk screws. 
Actually works a little better. Screws are better with plywood compared to thin MDF.



All measuments performed on IEC baffle at 0.5 meter distance, response normalised for 1 meter, 2.8 volts input.

Let's slowly zoom out over the next four graphs and start looking at the 1000-10000 Hz range to the upper left, the range where things really matter because most treble energy is in this area. Well, not much to comment on because this all look smooth and easy as should be. 
These two graphs are an average of 0, 10, 20 and 30 deg. responses.
To the upper right I have included the 10-20 kHz range and averaging the 10-18 kHz range we see a response of ~94dB/2.8 volts. 
Overall, quite a potent system for high-efficiency systems.

To the upper left we have the 2-40 kHz range at 0, 10, 20 and 30 deg angles. Right: 40 and 50 deg. angles included.

For those interested, the impulse and step response.

CSD @ 25 and 50 dB scaling. Note the unusual clean 25 dB presentation. 
50 dB scaling is hardly relevant as we're getting into the noise floor, but it may give hints of trouble zones, 
like the tiny wrinkle at 6 kHz, but it's so low it doesn't matter.