Recently there was a buzz on the boards about some inexpensive, coaxial-driver 2-way, both good and bad. After seeing it and the price, I decided that I'd take the opportunity to check out the coaxial driver, regardless that it was cheap, and have a pair of decent, unusual and interestingly shaped small boxes to use if the drivers proved useless. This page is about the drivers, not about the result of any box and crossover tweaks. That's for another page, link to come later.
The centerpiece of these speakers is a coaxial driver. The woofer portion is equivalent to a typical 6.5" driver. The tweeter is described as a 4/5" tweeter.
To begin, I'll point out that all of these measurements were made with a "square" window, rather than one of the commonly used ones such as a Half Blackman-Harris tapered window that smooths the response to some degree and cleans up the high end. The square window provides the best detail in the low end with the possible introduction of spurious high-frequency anomolies that can occur due to the MLS technique and the window transition. I also did not apply any smoothing. A 1/6 octave smoothing function would provide a response that is closer to what we might perceive the sound to be, but my intent is to provide as much detailed information as possible. The higher the Q of any resonance, the more "hidden" it will be due to both smoothing and a tapered window.
The glossy front baffle is a separate piece that is mounted after the driver is installed. Removing this baffle makes for easy removal of the driver. The crossover connects via spade-style press-on connectors. I removed the drivers to make measurements on my 2mx2m baffle to get quasi-anechoic responses. A pair of typical SPL responses and impedance of each section (woofer and tweeter) are shown below. The impedance is in-box.
The first thing I noted was that the "peak" at 800Hz, the subject of some discussion, is not much of a peak in the driver per se. There is a ridge in the CSD related to this, but it is at 1000Hz in the one used for the impedance measurement. The next thing I noticed is the severe dip above the small peak. This is all indicative of a single resonance, centered close to 1K in these examples that coincides with an impedance glitch centered also at 1000Hz.
The "peak" at 800Hz seen in the in-box measurements is only secondarily a function of the resonance, even so. It is primarily due to the baffle diffraction related step that is always seen in a normal baffle. The response peaks, then starts to roll off below that point. 800Hz is on the downward slope of the step inherent from the baffle dimensions. That just exacerbates the bump. When this is combined with the 5db or so dip that is the true Fc of the resonance, the bump looks that much worse.
Outside of some minor deviation in the 1KHz resonance region, these two drivers are surprisingly close in response out to 10KHz. My goal is primarily to ameliorate the deviation due to the 1KHz resonance through driver modification rather than through crossover patching. Fix ths source, not the symptom, that's my preference. Easier said than done.
Determining the Source of the Problem
I set about testing to see what the primary source of the non-linearity is. I quickly eliminated air gap resonance behind the driver. There was little change to the resonance with any form of damping of the back wave. Had this been a major source, a significant change would be seen in measurements.
The basket is stamped steel, so I tested to see if that was a source of significant reflections . That, too, appeared not to be the case This left me with diaphragm/spider/surround issues. There's not anything that can be done with the spider even if it were somehow involved. But I doubt that it could cause the type of resonance exhibited.
It all came down to diaphragm/surround and/or diaphragm/former joints. This is not uncommon. On reflection, I recalled seeing similar resonance issues with some fiberglass diaphragms, notably the Vifa PL14 and Seas G18RNX/P (carbon fiber) The PL14 has a shelf and the G18 has quite a dip in response that are both, I believe, related to surround termination impedance mismatch. This was shown to be the case for the Insignia woofer with further testing.
The surround is a typical raised, half-roll type, shown in the picture below. It's made of some kind of rubber (likely) or Santoprene (unlikely due to cost, I suspect).
The former attaches to the diaphragm as shown below. There's a bit of non-uniformity in this glue joint. This is so across the board of the four examples I have, one of which is shown below. If you look closely, you'll see that the diaphragm is attached slightly higher on the former and has more glue as well on the area seen on the left side of the picture vs. the right side.
Tweaking, the Nitty-Gritty
Midrange and/or woofer tweaking has been on my mind for quite some time. Close examination of a number of drivers from different manufacturers often shows certain similarities. One of these is an application of some kind of black, moderately tacky material applied along the edge of diaphragms and/or surrounds. Some are underneath, some on top, some both places. The more interesting aspect was that most of those drivers had it applied to the surround only. This stuck in my mind and led me to purchase a material as a doping compound that I hoped was the same or had similar properties. This is shown below.
The Wet Look" sold by Parts Express
Unit 1, the first one I removed to examine, had shown some kind of problem on the low end. The MLS test signal audibly changed when I was running tests on the system prior to removal. The DC resistance is equivalent to that of the others, but the driver still showed some loss in the low end compared to the others when mounted on my large baffle. I decided that this one should be the guinea pig.
My first test was an application on the former/diaphragm joint and the surround roll/flat section "corner". The results were good, but there was an unexpected drop in sensitivity. Further testing showed that the former/diaphragm joint did not provide an SPL response benefit, but was primarily responsible for the drop in low frequency sensitivity. I stopped that method of treatment at that point. However, the application on the surround did provide some improvement.
The mass of the diaphragm may not be very high, so the added mass of the doping compound when applied to the former area seems to be enough to reduce the sensitivity. Keep that in mind if you undertake your own mods. The other mods I made showed little or no such change. The graph below shows the responses with successive applications.
Successive Compound Applications, Unit 1, Expanded Scale
The improvements in some areas do not necessarily come at no detriment in other areas. A possible drop in low end senstivity is just one. Another may be changes higher in frequency. The mechanical impedance at the surround/diaphragm junction or within the surround roll corner itself is not purely resistive. It's reactive, meaning that it is frequency dependent. This can be seen in the modified responses shown. Examine the frequency range from 3-5KHz for example. Note the change from a linear slope to a "stepped" slope. This could slightly complicate a crossover, but in comparing to the other three drivers I have, they all had this shape new, so it would seem that my guinea pig driver had some anomoly different than the others. Be sure to keep this in mind as well if you ever modify a driver. I plan to use this mod on some other drivers, but the whole bandwidth has to be examined with each application to ensure no excessive detrimental impact occurs in area outside of the intended one. The alteration in this one was easily acceptable.
I initially found this "step" to be rather unusual, but the explanation is in the impact of the tweeter, that I demonstrate later in this document. The downward slope from 1.5-3.2KHz is also directly related to the existence of the tweeter.
Next I added doping compound to the flat surface of the surround that is directly attached to the diaphragm. This appeared to do nothing. Adding the doping compound to the surround roll where the "corner" is on the diaphragm side is where the improvments can be made.
What I found to be optimal is to apply two coats directly onto the surround roll "corner" and a third application applied only to roll at the corner. This moves the application up onto the roll a bit more. I stopped at this point.
1st Application Appearance on the Surround
How to Apply the Compound
I experimented with a couple of different brushes for applying the doping compound. The best results were achieved with a size 000 artist's paintbrush, purchased at an Ace Hardware store. The compound starts to dry quickly, so cleaning of the brush between drivers (if more than one to do) is highly recommended. This is quick since it is cleaned with water and soap.
The photo below shows the appearance of the first coat. It's not going to look as aesthetically pleasing as what is applied by a manufacturer, but is relatively unobtrusive.
Was it worth the effort?
Without doubt, this is one of the most successful and useful tweak results I've obtained. The benefit cannot be had any other way, short of digital equalization and has little in the way of negative consequences. The final results are shown below for the pair that I modified after I exhausted all experimentation on the guinea pig driver.
Final Results, Units 3 and 4
The resonance at 1KHz was not totally controlled, but the max/min delta in Unit 3 went from 11.9 down to 5.3db, a reduction of 6.6db . In Unit 4 it went from 8.8 to 4.6db, a reduction of 4.2db.
What impressed me even more is how close these two drivers now are from the 200Hz to 3KHz area now. The scale of the normalized graph below is 1db/div showing that the difference is within +/- 0.5db in this critical range. This should provide good imaging.\
Final Results, Units 3 and 4, Normalized
The response of the best raw unit still shows significant improvement when the before and after responses are overlayed. There is the additional, undesired peaking at 3.5KHz, but this is still acceptable given the improvement elsewhere.
Unit 4, Before and After
The CSD for the worst case, Unit 3, improved rather dramatically as seen below. The ridge associated with the 1KHz peak is very distinct in the new driver. It is now reduced and much less distinct.
Initial Response, Unit 3 CSD
Final Response, Unit 3 CSD
The tweeter measurements show a horrible response. I could not determine what was due to the tweeter driver and what was due to the horn loading of the tweeter inside of the woofer. So I removed the tweeter. This was even easier than it sounds. The tweeter wires were unsoldered and the screw holding the tweeter in place was removed to provide access. THe glue attaching the wires to the plastic cover were carefully unglued. Then I inserted a small metal piece (a small Philips screwdriver in my case) was position inside and tapped increasingly hard until it popped loose. It actually stayed in position due to the magnets, but it was easily fully extracted afterwards.
Woofer with Tweeter Removed
The next graph shows the influence of the tweeter on the woofer response. It has more impact than you might expect. What was most apparent to me was not the huge dip at about 2.7K, it is the difference between the green and yellow curves from 1-3K. The departure of the original green curve from the linearity exhibited in the yellow curve (with a dust cap), is almost exlusively due to the tweeter. There's a peak at 1.5K and a trough at 3K. What's frustrating here is that if the tweeter is to be used, no changes can be made to improve this response.
Woofer Response with:Tweeter in Place (New) - G, No Tweeter (Treated) - B, Dust Cap (Tteated) - Y
My goals were to improve the native raw response of the woofer section in this coaxial unit in whatever way possible while making it easily replicated. I think that I've succeeded to a greater degree that I had thought possible as evidenced in the comparison of the two drivers and their initial and final responses. Another goal was the collateral benefit of being able to create a crossover that yields both a smoother response with a smaller parts count and/or smaller values. That, too, should be possible.
The Wet Look can is enough to do maybe a hundred drivers. The current cost for this change is approximately:
|The Wet Look||$12.70|
|Paint Brushes||$ 3.00|
One last option would be to remove the tweeter permanently, attach a dust cap fashioned as I did, but permanently, then install the tweeter as a separate unit in a conventional 2-way. The coaxial benefits would be gone, but the raw responses of these two drivers would provide an arguably good system in that configuration.
The response when the driver is mounted in the box, especially with the glossy front baffle installed, will vary quite a bit from these. The normal baffle edge diffraction coupled with the diffraction from the glossy front will change it, but that appears to have an impact only on the woofer. The tweeter is totally occluded from the baffle edge apparently. That will become apparent in my page (not ready as of this writing) on the crossover that I implement.