ScanSpeak tweeters, D3004/660000, D2904/710003, D3004/664000 and R2908/714000

Copyright 2011-13 © Troels Gravesen

Go to on this page: D3004/660000    D2904/710003    D3004/664000    D2908/714000

click images to view large

From left: D3004/660000    D2904/710003    D3004/664000     D2908/714000
Click above to download SS specs.

For future projects, a pair of 6600 and 7100 tweeters were acquired. The basic differences between these two tweeters are the magnet system and the dome, where the 7100 has a ring of damping glue to the dome. The ring neodym magnet of the 7100 tweeter provides a whopping 95+ dB sensitivity, suitable for fairly high-efficiency systems where the 6600 does well up to 92 dB/2.8 volts. The 6600 display some inconsistency in sensitivity, around 1 dB and although I don't regard this as a serious problem, a little disappointing. Only two units were available for the 6600 tweeter. I had another pair of 7100 for test, hence 4 samples on display below. All of these are consistent in sensitivity and only display minor differences between 15-20 kHz, probably due to the damping ring applied. I wonder why SS added the damping ring as the 6600 display less distortion in this area compared to 7100.
Both of these domes appear suitable for two-way systems with a very low point of crossover and from clients' project I know the 7100 can do well down to 1.5 kHz with a simple 2nd order crossover. The 7100 tweeter was tested in experimental designs and proved well down to 1.4 kHz, 12 dB/octave. Very impressive.
The 6640 and 7140 Be domes were added later to my collection.

All measurements done with tweeters flush-mounted on 35 x 50 cm baffle, center placement and 20 cm from top. An IEC baffle may provide better low-end response.

The 6600 and 6640 tweeters present some challenges  with regard to crossovers, because these tweeters go so low the can play midrange! The 7100 with its steeper low-end roll-off due to the very strong magnet system is easier. Taking the 6600 and 6640 down to 2 kHz and using a 12 dB/octave filter makes it necessary to apply impedance linearisation to render a smooth roll-off profile.  


Left: SPL @ 2.8V, 1 m, unit 1 and 2. 
Right: Distortion measured at 0.25 meter, input 2.8V ~92 dB/1 meter, unit 1 and 2.

CSD, 2.8V.
Left: Unit 1
Right: Unit 2.



Left: SPL @ 2.8V, 1 m, unit 1 and 2.
Right: Distortion measured at 0.25 meter, input 2.8V ~96 dB/1 meter, unit 1 and 2.

CSD, 2.8V. Left: Unit 1. Right: Unit 2.


 Left: Step response, unit 1. Right: Step response unit 2.

 7100 dispersion at 0, 10,20, 30 and 40 deg.

D3004/66400 Beryllium dome

Click image to view large. Download specs here.

SPL normalised for 1 m/2.8V, unit 1 and 2. 

Impulse response, unit 1 and 2. 

Step response, unit 1 and 2.

CSD, 25 dB, unit 1 and 2. Very clean indeed!

Left: SPL unit 1 and 2 (not normalised for 2.8V/1m). Right: Impedance unit 1 and 2.
Quite as matched pair!

Distortion from unit 1&2. Measurement at 0.25 m distance, 2.8 V. Black = 2nd harm., blue = 3rd harm.
These tweeter display significant lower distortion compared to 7100 and 6600.

I don't think I've ever had a pair of so closely matched tweeters. They have outstanding performance in my Jenzen ATS speakers.

D2908/714000 Beryllium dome

Click images to view large

Measurements, baffle: 65 x 57 cm, 27 cm from top. All measurements normalised for 2.8V/1m.

My expectations are high on this new tweeter and nothing below suggests I should be dissapointed. The Revelator magnet system is stronger compared to the Illuminator motor, thus an 8 ohms voice coil is used here. On top of this a titanium voice coil former, not seen before on ScanSpeak tweeters. Most noticeable is Qm, 7.3 according to specs. My measurements suggest a little lower, but nevertherless, nothing I've seen before. High Qm means low mechanical damping and some may fear this will make a too lively treble. Time will tell. I've heard reservations towards this 7140 tweeter based on the  response graphs displayed at SS website. The use of a 10-100.000 Hz range doesn't make things look better on paper. Looking at the 1.000-20.000 Hz range alone (see below) display an usually well behaved response and initial modelling suggest an easy tweeter when it comes to crossovers. And let me say that I couldn't care less about the dip above 16 kHz towards 20 kHz. Some people are obsessed about a flat response to e.g. 30 kHz. There are much more to good treble performance than a ruler flat response to 30 kHz. 

Left: SPL at 0.5 m distence, on-axis. Right: Tweeter 1 and 2.
These units are so much alike that only one will be shown below except for cumulative spectral decay and impedance.

Left: Impedance of unit 1 and 2. Right: Step response display better damping compared to 6640!

Left: Horizontal dispersion at 0, 10, 20, 30 and 40 deg., 1000-20000 Hz.
Right: Same at 200-40000 Hz. My CLIO microphone is down 1 dB at 20 kHz and 5 dB at 30 kHz, so please add some level there.
Generally these response graphs look better than those pulished in Voice Coil and Klang&Ton magazines.

CSD @ 25 dB range. These are seriously clean and quite better than those made by Vance Dickason.
I wonder if some work has been done to the tweeter since they were reported in Voice Coil.
Compared to the 6640 tweeter, these CSDs suggest a better rear chamber with fewer resonances in the 500-1200 Hz region.

Left: Red, SPL as-is and red, 6.8 uF bypassed by 3R3. Simple equalisation.
Right: Orange: Test crossover @ 2 kHz from 8.2 uF/2R2+0.47mH in front of above equaliser. I don't think equalising will be necessary in my final set-up. 

Left: Distortion at 2.8V input taken at 0.25 m distance. Green = 2nd harm., blue = 3rd harm. Right: Same at 5.6V input.
This tweeter can safely be used down to 2 kHz - if you can tune the frequency response! Not that easy.

Now, all measurements above were done on a large flat panal, but how does it look when it we use a small baffle like the one I'm going to use for the next ATS 4-way speaker?
A quick cardboard mock-up, approx. 13 x 20 cm, and some midrange cardboard panel below display the red response above compared to the large flat panel (green). Not bad at all! I took the measurement to my simulation software and it looks easy. And no equalising step needed. Looking forward to implementing this tweeter in my next construction.