Zaph|Audio - SB Acoustics 2-Way : Tweeter Tweaking
Tweeter Tweak - What is it and why?
Tweeters as supplied by a manufacturer may have some deficiencies. That is, they may not be and often are not fully optimized. This is usually the result of meeting a price point. There may be tradeoffs made or some options may be left out. The SB tweeter is one such case. The construction quality is very good, it has copper in the gap to reduce modulation distortion, but its frequency response is not as smooth as it might be. There are several reasons for this, but all seem to be part of meeting the price point. That said, few tweeters, maybe no others, at this price point have copper in the gap, so it's got that advantage to start.
Two characteristics will be addressed having to do with the motor and in my experience, improvements to these can be made at little cost. We'll examine the motor, the current situation and some changes that can be and have been made. None of this is anything new, by the way, most treatments to tweeters are found in various other models. They've been around for a rather long time. The nice thing is, we can do this ourselves for practically nothing but out own time. If you're really into DIY, I can't see why you would not do it. It just takes a bit of care.
The Motor as Supplied
The changes that can be made aren't really to the motor itself, it relates to the motor structure and the impact the structure has on the sound radiated on the interior, behind the diaphragm. Some designs have gone to great lengths to minimize the impact, for example, the Scan-Speak Air-Circ or the Seas Millenium. These are expensive units.
Cost/performance drivers such as the SB have fairly standard motor structures. This is the typical magnet, top-plate, pole-piece and a pole-piece vent. Some have a rear chamber, large or small, some have none. The SB has a nominal, very small, chamber. Most tweeters have some form of damping at various locations and/or have an actual physical "obstruction" to impede the direct flow of sound to reduce reflections and resonances. The SB has something of a mushroom-shaped fixture (viewed from the side) mounted in front of the pole-piece vent. You can see some of these details in the picture below:
Modifying the Tweeter
The tweak is to disassemble the driver by removing the front plate and the diaphragm assembly. You can some other examples of similar tweeters and some disassembly on other pages here. Do this slowly with some care and it's an easy task. The plastic dome will be obvious on the top of the vent. Pry it off. Insert a small wad of cotton into the vent. A cotton plug from a medicine bottle such as aspirin, Tylenol, Advil or the like is perfect. Cut it down so that it's not compressed much when installed, then press it into place. If you use a cotton ball, be sure to fluff it up quite a bit, it will be way too dense at first, then cut it. Now use some 1/8" thick felt or cotton to ring the top as shown, though this step is a bit less important. It will affect the range around 15K up. I used some felt "rope" 1/8 thick that I had on hand. However you do it, you must ensure that there is no contact with the dome.
The raw response of the SB tweeter, measured on a 2m x 2m baffle (quasi-anechoic) is shown in the next graph. Notice the somewhat ragged response. Possibly more important is the unit-to-unit consistency. You'll see a bit of variation between the two units.
Had it been easy to replace the very small chamber, the bump would not have been so prominent. There's too much risk of damage. The change in tweeter Fs does have an impact on the crossover, but that's the price to pay (in this case) for a smoother response.
Contrast this with the comparison overlay of the treated tweeters below. They are now much smoother and are much closer to each other in absolute response. Another benefit is the somewhat constant slope in the midband. We'll use this to our advantage in designing the crossover.
For a better view of the differences and improvement, let's look at the normalized responses. This is simply a graph of the difference between the two curves. It's a mathematical divide process of the two measurements, one by the other, that leaves only the difference.
The two dominant traits of the difference in the untreated tweeters is the difference due to some resonances around 2.5K and 3.2K. These are differences, not individual resonances. The more egregious difference, however, is that of level. These two drivers differ from about 1.5K to 8K by more than 0.5db on average. This is the best reason to treat them.
Purchasing drivers that are quaranteed to be within +/- 0.5db usually comes at extra expense. We can at times, however, ensure this ourselves. Look at the next graph and contrast it the one above. This one shows the difference after treating both tweeters. Below 12.5K, these two drivers are now within +/- 0.5db. Off-axis, the area above 12.5K may also be closer, since peaking above 10K diminishes off-axis.
I suspect that the difference from 1-2K is from slight differences in the cotton density. It takes time to disassemble, treat, reassemble and re-test, so I have not tried to reduce this. Being at or below the desired Fc, the impact will be lessened as well.
Finally, let's look at the impedance curves. There are two characteristics changes, one good, the other we'd prefer to not change, but with a small chamber, it has to be accepted. The Fs is both higher in frequency and Q. This will slightly exacerbate the crossover work, but in the end the impact is workable. The benefit is easily seen in the elimination of the resonances that reflect back to the impedance. You can see these at about 1.4K, 3.4K and even at 6.2K. There's no impact electrically, this is just verification of the effectiveness of the cotton. These changes are a reflection of the change in the frequency response.
David L. Ralph © 2009