The new version ("New" or "Gen IV" Silkie) Dayton DC28F-8 tweeter, has characteristics of the prior versions that are largely cosmetic. It's really a different driver in many respects. It is still a 1-1/8" (28mm) doped silk dome driver. I had the opportunity to test and inspect a pre-production model, courtesy of Darren Kuzma, and a pair of the production units, courtesy of Shawn A., a denison of the Parts Express discussion board. My thanks to both of them.
|The new Silkie looks nearly like the previous three versions except for to the dome. The doping compound seems to be much more uniformly applied.
|The foam plug that comes with the New Silkie is the one on the left. The plug on the right comes with some Seas tweeters. The visual comparison is a bit misleading. I expected little from the rather small piece of foam, but it does surprisingly well.|
There are a several differences between the old and new Silkies (see picture below). The felt (cotton) ring that is glued to the top of the vent opening is shallower. This does make it less effective. This Silkie also has a true chamber, though a bit smaller than some other tweeter chambers. The last obvious difference is the centering mechanism. The two small protrusions do a good job and make it easier to position the diaphragm assembly as it gives a good visual reference. They make it less likely that the former will scrape the motor as it is placed.
The diaphragm assembly has the look of quality when first disassembled. This is no quarantee of sound quality, but bad visual quality usually, at least in a tweeter, usually translates into poor sound quality.
The felt that is included with the chamber is of good quality and is very firmly attached. My only complaint with it is that it does not fully fill the chamber and would perform better if it did so. A filled chamber with slightly less dense felt does make a small improvement. I'll present measurements of this further down.
The first measurement (below) shows a somewhat varied response with non-linearities often seen in tweeters, but with a couple of aspects that concerned me. However, I want to say up front that I later found that part of it was not a problem with the tweeter. Rather, it was with the baffle cutout, something that really surprised me, as it is a baffle insert that I have used before. I did not see this problem at that time. I can only assume that it is due to differing dispersion characteristics. More on that later.
I think that it is also the measurement taken using the faceplate with the supplied foam attached. The faceplate will easily deform if the screws are tightened too much. I did pay attention to this and made all measurements afterwards with a faceplate I have from a Gen. I driver that did not have the foam. This allowed me to tighten the screws reasonably without deformation. Personally, I think that given that these faceplates are plastic a much softer foam would be in order.
Experimentation with some modifications showed that some improvement can be made. Though I didn't modify the chamber on the new ones, I made a few quick tests on my driver. The dip at 4KHz was not much improved with a larger one, regardless of the stuffing. The graph below shows the response that is possible from this driver with my lamb's wool modification and nohing else. It also includes a fix for the baffle issue. Note how smooth this driver is now, with the exception of the 4KHz dip. That will also be explained further on.
The gap to which I referred can be seen in the following set of pictures. It's a consquence of using a router with the Jasper Jig, but not changing the bit to get a perfect opening diameter. I erred on the side of too large, rather than too small. Most of my dirver cutouts have been a nearly perfect fit, but this one has a gap of about 1mm around it (2mm too large in diameter). This, I had thought, was of no consequence. I had not seen this problem in other driver measurments, so it still puzzles me.
That said, I can't ignore repeatable measurements. I tested this idea by stuffing the gap with some Blu-Tac rolled into a tiny string. This made a rather significant change around 7-8KHz. This falls right into the calculation that has that gap a distance of roughly 1-3/4" from the dome. The calculation is 13,550/1.75 = 7743Hz.
One wavelength for a reflection would normally be constructive interference, rather than destructive. However, since it's a gap, it is acting not as a reflection, but as diffraction. This means that it is initially 180 degress out-of-phase, thus it is destructive interference, a dip in response.
At first I really couldn't believe that it made this much difference, but it was 100% repeatable. The pictures below show the situation, along with my initial, though partial, fix.
|This is how I originally mounted and measured the drivers. The gap is obvious, but appears to be benign.||This was my initial test to determine if he gap really did make a difference. The measurement improved considerably with this. My final measurements used a piece of foam cut so that I could install the driver and fill the gap. It's more difficult than it sounds when done this way. |
|This closeup shows some detail of the gap. It still seems as though it should be benign.|
The chamber provided is large enough to stuff with more felt of a different density. I had on hand some felt removed from another driver's chamber that just happened to fit exactly. The originally supplied dense felt was glued more than securely. I replaced the felt in my sample and measured the impedance. A comparison of the impedance of my unit with one of those supplied by Shawn shows that the lamb's wool doesn't change the Fs much, but the altered chamber stuffing makes a fairly significant difference. A crossover can be more easily implemented due to a lowered Fs and much lower Q. In fact, the Fs was changed from about 1000Hz to about 800Hz in the beta unit. The SPL change has a more gradual knee, but does have a bit of a broad, low-Q dip centered around 1.5KHz. It is not enough to be a big concern and could easily be handled in the crossover. I can't say if it would be different for production units as I wasn't about to tear apart any of the two sent to me.
I've included a CSD plot, though I am less enamored of their usefulness at this point. They are good for finding resonant ridges, but the SPL plot does so as well, since non-linearities are almost always some form of energy storage somewhere in the driver. The rate of decay is, however, useful to a degree.
Next up is the quick-n-dirty chamber I fashioned. It's just a piece of PVC, an endcap, and actually works very nicely, being just the right diameter for most tweeters. Blu-Tac works well to temporarily attach it to the motor backplate. I also doubt that other chambers, larger or shaped differently, will make much difference from this one.
I tried a couple of different kinds of stuffing, first some typical "polyester stuffing" and some lamb's wool. Both work, but I think that the lamb's wool is more effective. The stuffing alone is not useful since it does not effectively diminish the pipe resonance that the vent exhibits. Putting the original felt over the vent then placing the new, filled chamber in place yielded the best results that I could get. The measurements follow the pictures.
Below is a comparison of the beta driver I have with several different chamber configurations with an overlay of an unmodified response from one of the pair of production drivers. Ignore the differences above 8K. This area tends to change at times when the driver is disassembled and reassembled.
The green curve is the production driver response. Note the slight peak at about 1.2K and the response extension. The yellow curve is the reponse with the large chamber stuffed with lamb's wool felt and the vent opening covered with the original felt. The impedance comparison graph is shown again on the right. The SPL and impedance curves with matching colors are for the same driver and/or conditions.
You might think that the larger chamber is not a benefit, given that it seems to reduce the extension. My take is different. Yes, it does reduce the peak at rolloff, reducing the Q. But the benefit may be in the accompanying change in the impedance. This should make it easier to design a smooth highpass, especially if a lower Fc and/or a lower order is desired.