By Lars Mytting
02 Nov 1996
Printed from SpeakerBuilding.com, 11 Mar 2010 23:27
URL: http:///content/diy/1009/
ScanSpeak and Seas has the last year released new drivers that intends to be in the absolute top-flight of the market. They are priced significantly higher than regular drivers, and all have interesting construction details and parameters.
I set out to test three of these drivers by measuring them and design a three-way with them. The drivers chosen were the ScanSpeak 9900 Revelator 1" tweeter, Seas W17EX002 6,5" midrange/woofer and the ScanSpeak 8565-1 10" woofer.
During the initial process, I made extensive use of the magnificent Clio system. Later on, I used the equally magnificent Calsod 3.10 software, which The Speaker Building Page has received for test by Audiosoft.
Unfortunately, I don't have the whole design process covered with measurements, but all facts on the drivers are available in the measurement-section on this web-page. Here, you will also find downloadable frequency and impedance data which you can use in your Calsod or Leap-program, and you will also find the data provided by the manufacturer.
This baby differs from the other 2905 drivers in having a very large, concave frontplate. According to the manufacturer, the shape is intended to improve directivity. In the official data, it also measures more linear than the 2905/9700, which essentially is a 9900 with standard frontplate.
The Revelator really sounds well, with some margin better than other softdomes I have heard. Most of all I feel that it blends very well with the midrange of the system; it does not emphasise itself; and does the tough act of "not being there" very well. I find it lighter, more weightless than the regular 2905/9000, but with the same good dynamics. It has the same "accent" as the other 2905's; a somewhat dark, easy-floating quality. The driver also has adequate linearity and good power handling capacity. The driver employs a special coating which is applied by hand.
The first set I bought of the Revelator showed quite large differences in the frequency response between the two samples. Using the Clio system, I found that one driver had a +3dB peak at 15,000 Hz. Later, I got a new set that showed far better consistency, and the performance was now remarkably smooth, with a difference of only +1/-0.5 dB in the upper octave. There are no visual change between the "old" and the "new" tweeters, but the new drivers are packed and delivered in pairs, while the former series were delivered one by one.
The impedance peaks on the new drivers were located at 507 Hz at 23.0 ohm and 501 Hz at 26.4 ohm. The impedance peaks on the old drivers were located at 522 Hz with 27.6 ohm and at 537 Hz with 25.4 ohm.
Note that these graphs show the difference between the samples, not the frequency response itself Fig. 1. Old Revelator drivers Fig. 2. New Revelator drivers.
The large front plate looks impressive, and it has a heavy, expensive feeling; very special-purpose and authoritative. It also makes a rigid mounting for the driver. But the diameter (130 mm) forces a large distance to the midrange driver, making phase- related interference patterns a larger problem than with smaller front plates.
The bolts to the magnet system are flush-mounted, and so are the mounting holes to the cabinet. Other 2905-tweeters does not have flush-mounted bolts, and I will devote some frustration to this fact. I really cannot see why ScanSpeak restrict this feature to their top model. Nearly all other tweeter manufacturers, down to the cheapest Indonesian type, flushmounts all bolts.
Neither can I understand why ScanSpeak continue to use their small, Mickey Mouse-terminals. They are far to weak, and break off after a few times with experimenting and resoldering, especially with stiff cables. Yes, tweeter soldering is a delicate art, not macho welding, but take a look at Dynaudio's terminals: Compared to ScanSpeak, they seem intended for war. Of course, neither terminals or flush-mounted bolts affects sound quality much, but why settle for a low standard on some parts of atop quality product?
The driver uses the standard Seas chassis, and has four mounting holes. What may be of interest is that on the latest series of these chassis; there are pre-drilled four additional paths for mounting holes. Just grab your drill and you have eight mounting holes. I guess that in time; Seas will supply four holes for the British market and eight for the American. (The French will probably demand five).
Back to facts. The main concern with this driver is that the magnesium cone has a natural resonance that creates a large peak (approx. 20 dB) at 5000 Hz. This calls for special attention in the crossover. If it is not corrected, it will cause a "ringing" sound. Success or failure with this driver is tightly correlated to the damping of this peak. From the waterfall plot, we may also notice that there is a resonance at 1800 Hz, but otherwise the response is very clean Fig. 3. Seas W17EX002 waterfall.
The two samples showed good consistency, and both frequency response curves were very tight. (See measurement section for graph). What was most exiting with the measurement was the impulse response. It has truly exceptional symmetry. The extreme "oscillation" is probably correlated to the 5000 Hz peak and its implied ringing. The "oscillation" is gone when applying the proper crossover Fig. 4. Seas W17EX002 impulse response.
For comparement, this is the impulse response of the W17EX001, which is essentially the same driver, but with glass fibre cone. This one is also measured without crossover Fig. 5. Seas W17EX001 impulse response.
More important with this driver is the extremely low resonance frequency. I measured the samples to 18,5 and 20,5 Hz! This, together with sympathetic values for Q and Vas, make a linear response down to the very lowest octave available without going to the very extremes in box volumes.
The frequency response and impedance of the ScanSpeak 8565-1 shows adequate consistency. Worth noticing is that this driver has a +10dB peak at 2000 Hz.
On the two samples, there are some differences in their respective Thiele/Small parameters. This is rather common from many driver manufacturers, and in this case, the values summed up so that there was no practical difference in response in a standard cabinet. But it still something to be aware of. Again, the discriminating speaker builder should demand from a good dealer that they deliver individually measured speaker drivers in reasonably matched pairs. This driver is expensive, and if you want to save money, you could try the Vifa PL26WR-09-08, which share many qualities at half the price.
Then the Thiele/Small parameters. The values given beneath should not be used for final box calculation. The Clio system is extremely precise on Thiele/Small parameters, but I had not available a test box that was 100% appropriate for the drivers. However, both drivers under test met exactly the same conditions, and the values are true with respect to the differences between the samples.
Sample A Sample B Fs 20.5 Hz 18,5 Hz Vas 167 l 209 l Qts 0.37 0.32 Qms 4.25 3.99 Qes 0.41 0.34 Mms 49.49g 48.50 g Cms 1.21 1.51
The differences in parameters may appear to be significant, but in real life, the values interact so that the total difference in behaviour is very small when placed in a cabinet. A "perfect QB3" cabinet for sample A is 112 litres and a 8 cm port of 21 cm length, giving -3dB of 24 Hz. For sample B it is 88 litres with a port of 8x26 cm, yielding a -3dB of 27 Hz.
The official Scan-Speak parameters of this driver is Qts 0.33, Vas 222 l, and Fs 19 Hz. If we use these values when calculating our QB3 box, we get a recomended cabinet size of 103 litres, and a 8 cm port of 22 cm length. And when we place sample A and B in such a cabinet, they both yield a -3dB response at exactly 25 Hz.
The net volume of the bass cabinets are 55 litres. By closing the existing port, I ended with a system Q of around 0.7. This works well and is easy to predict. For maximum benefit of the extraordinarily low Fs, a bass-reflex cabinet is best suited. A good match for the 8565-1 is a bass reflex cabinet of 103 litres with a port of 8x22 cm.
For the 6.5", Gunnar had discovered that the condition for this driver was to use aperiodic damping, simply by stuffing the damping progressively tighter towards a hole in the rear of the cabinet. If you want to use a bass reflex box for the Excel, a size of 8 litres with a port of 5x27 cm will be suitable. The T/S values of this driver is not too cooperative for achieving low bass response, so this combination gives a -3dB of only 61 Hz. If you want better bass response, Excel drivers with smaller magnets and other T/S values are available.
-(filter)---|----- IND | CAP | (RES) ------------|-----
The function of the RLC circuit is to make a very steep and narrow "valley" in the filter response, and the clue is to make this to correspond exactly with the peak of the driver. Crucial for success is to hit the peak exactly. The size of the inductor and the capacitor determines the centre frequency and the deepness of the "valley". An optional resistor is used if you need to change the shape of the "valley" so that it corresponds to the shape of the peak.
Using a large resistor, the "valley" will be wide but not deep. With a smaller resistor (or no resistor), the "valley" will be narrower and deeper. Practical resistor sizes for this use range from approx. 3 ohm to no resistor at all.
All components in the filter will act together, so the shape will change if you alter the size of the filter components before the RLC-circuit. This is definitely a job for simulation software like Calsod or Leap, but you will also get a long way if you use the method described by Joachim Gerhard in the interview with him on this subject.
Underneath is shown the effects of LC circuits of 3,3uf/0,25mh and 22uf/0.05mh. All filters use a 1 mh series inductor and no resistor in the LC-circuit. You will note that both responses are centred at 5000 Hz. The 22uf/0.05mh, however, is far wider than the 3,3/0.25, and is also more damped in the 5-20k range. This is generally an advantage, but it may also make it more difficult to achieve a exact filter slope, like the 4th order Linkwitz-Riley, for instance Fig. 7. LC circuits.
Contrary to many others, I enjoy crossing a third woofer a bit high to compensate for this; as I feel that it adds more muscle and impact to the music. I am satisfied using the 10" ScanSpeak for this, but a safer solution with respect to midrange quality is to use two identical mid/woofers, and to roll one of them off at approx. 200 Hz. The other woofer is crossed to the tweeter as usual. (Be aware to solve some complications with phase and impedance).
The alternative is to use a larger series inductor (often around 2 mh) to "eat" the higher sensitivity in the upper mids, but this technique requires that you lower the total sensitivity. Since you cannot easily add gain in a passive system, you have to bring the sensitivity of all drivers down to the "weakest point" of all drivers. Usually this is the mid/woofer sensitivity at 100-200 Hz, where it may be 84-85 dB, resulting in a total system sensitivity of this level.
--------------- (crossover) | 1,7 mh | 64 uf ScanSpeak 2905/9900 | 6,5 ohm | --------------
------1.35 mh----|--------------------- + | 3.3uf | 0.25mh W17EX002 | 1ohm | -----------------|-------------------- - -------3.3 uf----|-----3.7 ohm ----|------- + | | 0.47 mh (opt. RLC) 2905/9900 | | -----------------|-----------------|------- -
The measured response from 200 to 20000 Hz is shown in Fig. 9. SPL response.
The waterfall plot for the complete system in Fig. 10. Waterfall plot. Some room reflections affect the response to 1k. No smoothing is used on the lines in the plot.
------ 1.5mh ------ 0.65mh -------------- + | | | 0.4mh | | 18uf 2.7uf W17EX002 | | | 2.7ohm | | ----------------------------------------- - |- 1.5uf-| | | ---- 6.8 uf-------|-3.9ohm |------------- + | | | | 0.35mh 15ohm 2905/9900 | | | | ----------------------------------------- -
I am very satisfied with the overall sound from this filter, and it is the best choice for most systems. It sounds more coherent, not so aggressive, and is easier to implement with subwoofers and "hard" rooms.
The Calsod-simulations in Fig. 11. Filter function, crossover 2 and Fig. 12. SPL and Power, crossover 2 show the electrical response of the filter and the total SPL and power response.
I enjoy to cross a bit high, often around 250 Hz, to get more "snap", "bang" and energy from the 10" in the lower midrange. For experimental purposes, I use a active crossover that rolls off at 12 dB/octave. This blends fairly well with the rest of the system, but for maximum performance, the crossover should be closely optimised to the lower rolloff of the mid/woofer.
To achieve better power handling, I have placed a simple highpass filter on the input of my power amps. This filter consists only of a series capacitor. The formula for this is the same as for a speaker 6dB highpass filter, only that you substitute the speaker impedance with the input impedance of your amp. For a 200 Hz filter on a 10 kohm amp, you need a 0.08uf cap, for the same on a 50kohm amp, you need a 0.016uf. Note that many amplifiers already use input capacitors to block DC, and a good solution is to replace this cap (usually large) with the value giving the required rolloff. Many preamps also employ output capacitors for the same DC-blocking purpose. Use the best quality you can find for this capacitor.
My amplifiers have 600 ohm input impedance, and this graph shows the rolloff caused by 0.47uf and 1.5uf capacitors Fig. 13. Passive line filter.
The soundstage is very airy and open, with good imaging. Before, when I have played classical music with many violins, I have often missed the certain quality which was well described by hifi-writer Knut Vadseth as "the sound of violins like a bee-swarm". The system provided the necessary detail and micro-energy to provide a full measure of this. So, given that you manage certain crossover problems, a true reference system is within reach.
---------|-------------|---------- | | IND 1 IND 2 |----RES 3----| | | CAP 1 CAP 2 | | | | RES 1 RES 2 | | | | ---------------------------------
Still curious about the term "hylo-idealistic" used in the beginning? Well, full understanding of the term is difficult, but the following link contains complete reference literature:
http://www.mk.net/~dt/Bibliomania/Fiction/wilde/stories/oscar3.html
Happy listening and happy reading!
Lars Mytting
Fig. 1. Old Revelator drivers
Difference between two samples of old Revelator drivers. Scale 2 dB.
Fig. 2. New Revelator drivers
2. Difference between two samples of new Revelator drivers. Scale 2 dB.
Fig. 3. Seas W17EX002 waterfall
Fig. 4. Seas W17EX002 impulse response
Fig. 5. Seas W17EX001 impulse response
Fig. 6. The Nighingale
Fig. 7. LC circuits
Fig. 8. Tweeter impedance correction
Fig. 9. SPL response
Fig. 10. Waterfall plot
Fig. 11. Filter function, crossover 2
Fig. 12. SPL and Power, crossover 2
Fig. 13. Passive line filter
