Tests and Analysis by David L. Ralph ©
Updated December 2007
The new 6600 Air-Circ tweeter has generated a lot of interest on the discussion boards. It's unique motor design is supposed to solve some of the drawbacks of a standard motor. That is debatable, at least with the unit I had the opportunity to test. My thanks to Mark K. for loaning this unit for testing. Test results, analysis and a couple of tweaks are presented. The latter help to correct some frequency response issues.
Within two or three days after posting this page, I found that I could remove the diaphragm assembly without damage. If you haven't read through this before, continue here.If you have, you can skip to the new sections at Addendum #1. Some photos in the new section have very large versions that can be opened by clicking on the picture. I also inserted the faceplate comparison graph I had left out in the original section.
I was loaned a second pair of current version 6600s so that I might add my mods to them. These two were much closer to the published specs. The benefit was much less apparent. A full set of SPL measurements is shown at the bottom of the page a Addendum #2.
The basic diapragm assembly is not new, having been introduced in the Vifa DX25. The dome material and doping compound are different, however. What's really unique about the 6600 is the motor. It is composed of six neodymium magnets arranged in a circle rather than there being a single large disc-shaped magnet as is found in typical drivers. This keeps the pole-piece vent of the typical driver and creates open areas between each disc. There is still something of a "chamber" associated with it, but the key is having those six open areas. The outer diaphragm ring section has vents into these open areas that couple with the pole-piece venting into the chamber cover. This is, I assume, the source of the name "Air-Circ".
My understanding is that this unit is not necessarily representative of final production. It's still a good unit, though, and tested well for distortion characteristics. The frequency response is another matter.
I made an initial SPL measurement on-axis to get a baseline on it, shown below. There are some irregularities that are definitely not what one would expect from a unit such as this. The good news is that it can be improved with some minor tweaking by the purchaser. I would expect this to be the case even for the final production unit.
The off-axis response family, shown next, is good for a tweeter of this total Sd. This was a goal for the designers, to have an extended low end while maintaining better off-axis response than typical larger domes. It's not as good as a 19mm dome and I'm not sure that it's better than the domes that are 25-28mm. Looking over some of my other driver off-axis measurements, it doesn't hold up as well and I can't say why. There is one unusual aspect. Typical dome tweeters exhibit a rollof above some frequency, say 2KHz, but below that they're usually not much different. The 6600 measurements I made show that it rolls off throughout most of its passband. I'm not sure whether this is good, bad or inconsequential. The manufacturer's measurements don't show this kind of response. It's the only driver I've measured that exhibited this kind of off-axis response.
The cumulative spectral decay (CSD) or waterfall plot shows the reason for the dip at 7K. LAUD can be stopped during the creation of the CSD to show specific time splices. There is a sharp dip that is indicative of a strong reflection. The source is not obvious, however. I originally suspected an internal reflection, but came to doubt that when the off-axis responses showed the dip diminishing.
One nice feature is that there are six mounting holes even though the faceplate is 4.5mm metal. Too many tweeters either don't have enough holes or they are too thin or not metal and warp. This does have one interesting minor side effect. being symmetric about the axis, they actually will have a small, measureable impact on the on-axis frequency response. In order to eliminate this, I filled them with Blu-Tac after mounting. There's little chance that anyone could find this audible, but I did so for the sake of focusing on the other sources of non-linearities.
Motor Examination and Details
Once I had all of the baseline measurements made, I started opened it up to check out the internals more closely. The motor structure is the most unique aspect. The photos below show the motor from several perspectives.
The first one (repeated from above) shows the arrangement of the six neodymium magnets and the shape of the motor rear plate. This is the key aspect that allows for the "air-circ" characteristic. The second one is a side shot that shows and opening with a barely visible opening just inside of it. This is an opening into the rear chamber arear and is another key aspect. This provides venting of the air under the rim portion of the diaphragm. I suspect that it helps to prevent compression of the air under the rim area. It does have one drawback that will be detailed later. The third one shows the interior under the back plate. There is a copper pole-piece cap that extends all the way into this area. It can be seen (somewhat fuzzy, though) to be protuding from the bottom in the center of the chamber area. This would seem to provide venting of the air around the coil (both sides I suspect) to exhaust into the chamber area.
The driver comes with a piece of cotton as damping material in the chamber. It's fairly effective, but not fully. More can be done, but I didn't come to that conclusion until later. There is also a hard felt disk supported just above the interior vent opening. It's just like that of the D2905/9300, flat on the vent side, rounded on the dome side. The support is from the XT25 piece that the central "bullet" plug snaps into to hold fix the center of the XT25 diaphragm. It's a nice way to do it, but I think that it does have a small bit of reflection that can't be corrected. I'd have to remove it to test this, but I can't because the dome assembly is glued into place.
It Can Be Tweaked
As good a driver as it is and separate from the fact that it has good measured distortion, there are still those non-linearities. The curves that I and others measured don't quite match the results provided by Scan-Speak. This may be addressed to some degree in the production versions. I'll be very interested in seeing what they do for the final production.
The first thing I did was add some small pieces of somewhat loosely packed cotton into the spaces between each of the neodymium magnets. I fluffed up some cotton balls (stretched them out evenly) to get a medium density. Then I tore off small pieces and cut them to a size that allowed them to be positioned into the gaps without much pressure. It's necessary to ensure that none of it overlaps the outer rim where an air-tight seal is made. The change in response is shown in the next graph. Blue is the original response, yellow is the improvement with the cotton plugs.
The dip at 7K remained, so I next tried adding a damping cylinder between the supplied cotton packing and the vent dome support. This provided more improvement than I expected. This one came from a Seas tweeter. As far as damping foam, those found in Seas are the best I've encountered. It becomes squashed to a very small shape once the chamber is replaced.
I also tried shaping the felt in the gaps so that it partially covered the small vent holes. This actually had a negative impact. This seems to indicate that the 7K dip is due to a small resonance associated with these holes and would explain why the dip decreased off-axis. This would be due to the phase differences as the mic is moved off axis and due to some partial occlusion of the ring area that would be "behind" the dome in the off-axis. At this point I just can't conclude what the real source is for the 7K dip. I'm inclined to think that it's the small hole vent resonance.
The conundrum is that Scan-Speak has measurements that appear to totally lack any resonance at 7K. If it's the faceplate rim area reflection/diffraction, it would appear in their measurements. That couldn't be altered except through dome damping that alters the off-axis output from the dome. If it's the vent resonance, they'd have to be damping it internally somehow, but that damping is non-existant in this sample 6600. If I find the answer conclusively, I'll update this page.
My next attempt in the damping area was to use a piece of cotton of moderate density in place of the foam. This probably pushes a bit of if into the area of the support that holds the felt dome.
The final tweak in the damping area was to add a piece of self-stick 1/8" medium density felt (cotton will do) in the recess for each of the six magnets. There is a gap just deep enough for this thickness. The 7K dip is lessened, but surprisingly, the frequency response is leveled down to about 2K as well. I'm not sure of the reason for this boost, but it was consistent in later measurements.
I haven't found the reason for the small step at about 9.5K up to about 12K. It may be related to positioning and/or density of the damping material. When it's removed and replaced, the shape is altered a bit since it's just a piece of cotton that I formed. The difference between the foam cylinder and cotton is too small to matter much, but the cotton seems the better overall. That's convenient, since good foam such as this is hard to come by.
The next graphs overlay the original response with tweaked responses for comparison.
The final CSD is shown below. This is a 40db scale and shows very good decay even up to 20K with only some minor resonance ridges.
The next graphs are presented adjacent to each other simply to show the comparison of the original response to that of the manufacture and scaled to match. I'll let you draw your own conlcusions.
It's often easier to see the difference in two measurements by showing one normalized to the other. This shows the difference only. The graph below shows the SPL of the modified version normalized to the original measurement. This in essence shows what changes were made by the modifications.
I can't say what the audibility is due to the differences above 5K, but that change is one that is not normally correctable. However, the broad change about 900-5000Hz should be. What would be more interesting would be to see some distortion measurments before and after.
Motor Influence without Damping
All motors have their own characteristics with regard to air spaces. The best motors are those that have optimally designed volumes and/or those that have optimal damping in the volumes that exist. I investigate the response of drivers with no damping when I can to see just what resonances exist. The 6600 is quite different in this respect. Typical tweeter motors have nearly the same type of resonances and repond similarly to damping materials. The 6600 is unique due to the unusual, somewhat oddly-shaped chamber area. The measurements below were taken with the chamber off and no damping other than for the hard felt dome just beneath the diaphragm dome that cannot be removed without removal of the diaphragm that is glued into place.
Faceplates create interesting differences. Sometimes they appear to physically similar, but there can be surprising differences. The DX25/XT25 faceplate (same on each) looks to be very close and can be mounted onto the 6600. I was surprised at how much difference it makes. The central opening looks to be very close, but there is quite a change in output. THe 7K dip is reduced quite a bit while a gradual droop occurs from 8K up to the high end peak that's almost always the same. The low end also veers wildly, peaking a bit at 800Hz, then taking a nosedive. Keep in my in throughout this part that I'm only referring to the on-axis response. I didn't take time to make a series of off-axis responses with the DX25 faceplate.
At first I thought that something was wrong. I tried adjusting the faceplate, but the same thing occurred. I suspect that the upper end above 8K is strictly due to the curvature of the faceplate opening. It's not as shallow and has a more complicated curvature. The 7K dip is harder to pinpoint. I noticed on mounting the DX25 faceplate that it pressed down on the diaphragm right at the edge. However, the 6600 faceplate has a slightly angled cut to it. It's a small outward slant on the backside such that it provides a small amount of additional clearance and matches the slant of the outer edge of the ring of the diaphragm, preventing contact of the faceplate with the moving portion of the diaphragm. The DX25 faceplate may inhibit some movement of the ring, even if only slightly.
Another difference between them is that the 6600 faceplate is a very shallow horn overall. It's not flat anywhere. This is a bit like the D2905/9900, only less pronounced. This would primarily benefit the low end, but it also allows for the inner area to be thinner than one that is nearly flat to the center, such as the DX25, so this affects a broader range and may be part of the reason for the high frequency differences due to the thinning towards the center. The roundover of the inside rim is shallower.
The diaphragm can be removed with care. Rather than remove the clear chamber that covers the sides of the diaprhagm support, leaving it on provided a small ledge for leverage with a screwdriver just underneath of the connectors. The reason that I initially thought it to be glued on was because of the gasket that is positioned between the diaphragm frame and the top plate. It is pressed tightly with a fairly large surface area contact, so it tends to hold if it's been there for some time.
The 6600 motor gap and top plate are different than any others I've seen in one important way. The outer portion of the gap, the top plate section and viewable in the picture below, has a bevel cut into it that is significantly larger han others I've seen. Most of those have seemed almost insignificant in comparison to the point that I made little note of them. This bevel will create quite a different flux field. Exactly how it adds benefit I can't say with assurance, but I believe that it increases the density in the gap between the flat section and the pole piece. I don't think that this design is under-hung, though, so the flux above it will also be altered. There is also copper in the gap on the pole piece of the driver. The copper ring on the section below the bevel coupled with the copper cladding on the pole piece give the 6600 its excellent distortion characteristics.
The gap with the copper cladding can be seen a bit better below. Note as well the relatively large holes in the top plate. These vent the area underneath the ring section. There's a closed volume of air that is part of the diaphragm assembly that acts as a closed chamber as designed in the DX25 and XT25. The 6600 frame is identical. These vents must improve the response, but they do seem to add a vent resonance that I think is responsible for the 7K dip. I'm not sure how Scan-Speak has eliminated this dip, but it's not evident in measurements I've seen in newer versions of the driver (thanks, zaph).
The dome is the same as the one used in the D2905/9300. It's glued onto the plastic support and provides an opening into the chamber. This method of construction makes for more uniformity between samples. The 9300 and 9500 (with dual mirrored felt domes) weren't always positioned the same since they were glued with what appeared to be silicone only. Gaps weren't always the same from one sample to another.
I cannot always be sure that I've accurately determined the source of a non-linearity. The reason is that without another design that changes the true source of the problem with which it can be compared, tweaks I make may be treating the symptom, not the source. The difference made by changing the faceplate is a good example. It provides a bit of a sanity check when trying to analyse the issues.
There's one other somewhat minor attribute that I never see discussed. The 6600 comes with a gasket pre-installed. This is a nice touch, but in this case there's another detail. The faceplate mounting holes are offset. That is, they have a small washer-shaped protusion that prevents fully compressing the gasket. The gasket is thicker than this protrusion, so a good seal is make without destroying the gasket. Removing and replacing the driver multiple times is easy and still maintains a good seal. Scan-Speak considered all aspects of a driver in this one.
Despite the neodymium discs, the 6600 is more evolutionary than revolutionary. It is, though, a prime example of taking advantage of every last bit of that evolution.
The new drivers are clearly better than the earlier sample, though I couldn't see anything visually to indicate the reason for the improvement. My guess is that it is in the dome and/or doping compound application. In any case, the differences are significant.
The graph below shows the on-axis response using both a Half Blackmon-Harris and a square window for an interesting off-topic comparison. The two were both generated from the same impulse response. It's a demonstration of the uncertainty due to both the frequency resolution and the values generated with the window used. The range below about 1KHz can and is likely to be less than accurate. What is smooth with one window has what some think may be resonances when another window is used. In essence anything a the low end of any typical MLS measurement should be considered suspect.
I added my mod to the two drivers sent to me. The graphs below show a full set of on- and off-axis SPL measurements for one of them.