The 6-8 dB "bump" below 1 kHz as seen in most plots of dipole baffles with ribbon drivers was cause for a lot of speculation regarding the origin. In previous testing I discounted "floor-bounce" as the generating mechanism. Since the "bump" effect is function of baffle width, dipole cancellation and reinforcement come to mind. My initial playing with various models show some correlation frequency wise but do not explain so far the magnitude of the anomaly.
Equalising the "bump" out is an attractive first option, but since the "bump" breathes with distance, angle, baffle size and listening height, it is not an "optimal" solution, unless one "restricts" the sweet spot to a small area and has the necessary equipment to measure the system and tweak the curves to flatten this 6-8 frequency/magnitude anomaly. Measurements have shown that equalising the "bump" out does lower the harmonic distortion somewhat, but not to the extend expected. The planar magnetic ribbons show a "distortion floor" based upon the tensioning of the film, the film material utilized and the resulting resonances in the spectrum of interest. Lowering the energy supplied to the ribbons does not lower the distortion any further. This explains why the distortion gain from equalising this particular anomaly has not lived up to the expectations. (The Clearview CXR22 has an uncommitted EQ stage for DIY'ers wishing to implement this option.)
A second option that comes to mind is setting the high pass filter of the crossover such that the bump is "cut" away. (Loosing the bottom two octaves.) The acoustic response of the crossover slope combined with the descent of the bump will be lower than the crossover frequency utilized, but will typically require (additional) mid bass woofers to cover the spectrum which does away with the fantastic spectrum coverage planar magnetic ribbons are capable off. Although this approach is maybe not economical, it has the advantage of removing a resonance rich area for some planar magnetic ribbons with high levels of mild harmonic distortion alltogether.
So the search for a better understanding of the problem continued. Armed with newly gained "modeling" insights I conducted a new measuring session. (September 5 and 6, 1997)
Simple modeling so far has helped me to predict to a certain extend where the "bump" will be and how the dips will ripple through the response based upon baffle size and geometry. Once above the cavity resonance frequency of the ribbons, baffle size and geometry can be greatly discounted. Cavity size and construction dominates the frequency/magnitude response from this point on. Many tests conducted at a fixed distance with many different baffle sizes and shapes have proven this point.
Difficult item in modeling is to model the response of the "virtual" ribbons that exists at the edges of the dipole baffle because of the impedance change baffle/free air. One at each side. This response is hard to predict as it is dependent on many parameters. More math work to be done. But if you can't predict it (yet) , maybe I could measure it ? With this thought in mind I started out on a setup to measure the response of the ribbon/baffle combination in the "dead zone cancellation" area. (Clue: virtual doesn't mean you can't measure it ;-) The plots below show a non-normalized collection of :
The frequency/magnitude behaviour of the two virtual line sources at the edge of the baffle greatly helps explaining the source of the "bump" as seen on mostly symmetrical dipole baffles with planar magnetic ribbon drivers. There are two bumpy sources (white and yellow trace) versus one almost perfect response, center "real" source (red trace). This also goes a great length explaining why the "bump" is breathing with baffle shape, size, listening distance and on/off axis performance. Even proximity of my body with three feet of the edge of the baffle was sufficient to change the virtual line source response. Finding a good spot and angle takes some time since the "dead" zone is small by definition. With a little patience it is possible to find areas where the magnitude goes down another 10 dB.
Which prompted immediately the question: "What if one virtual ribbon is replaced with a real ribbon ?" This reduces the number of sources down to two, a real and a virtual source. In theory this should get rid of the large and wide bump, leaving only cancellations to ripple through the frequency/magnitude response. based on the distance between the two sources. These ripples can be dealt with by "trapezoidal-ing" one edge. (And off course, the image of a certain dipole ribbon + woofer combination came to mind that does exactly the same thing. Ribbon one side, woofers the other side. - I measured them a while ago and I din't pay attention to what went on.)
There still would be room for a bump but a gut feeling told me that this bump would be smaller in magnitude and could be made to disappear by designing the baffle so that the bump occurs below crossover frequency. Since there is only one side to the baffle, a one-sided baffle, similar in dimensions as compared to a two-sided baffle, has twice the acoustical path length for dipole cancellation/reinforcement. Hence the possibility to size the baffle so that the bump (lower in magnitude) resides below the response frequency of the driver. A crossover can assist in further eliminating the anomaly.
The plot below shows an RD75 ribbon mounted at one side of a rectangular baffle (green trace) Measuring distance is 3 meter. The red trace is the same baffle with a trapezoidal taper attached to it. Shows that there is significant room to play with a study of an optimized baffle for this set-up. The test baffles used where ripped out of ply-wood without any finishing nor edge treatment, just a verification of the principle. This is not an optimal dipole baffle yet !(Pseudo anechoic measurements, 14.5 millisecond data with 4096 point FFT, 1/12 octave smoothing.)
This method produces a pretty flat response up to the cavity resonance frequency area of the planar magnetic driver. At that point the cavity size and construction determines the frequency response at a given point in space.
The yellow trace in the plot below shows the response of a baffle with RD75 mounted on the edge. No treatment of the "free" edge of the RD75 although this one could benefit from a halfround smooth edge to minimize edge diffraction. Proper sizing at this end may further reduce dips and peaks. There is no crossover in the picture, so what remains of the "new" bump is visible. It is obvious that this bump can be made to disappear with the right choice crossover frequency, which is exactly what was done in the plot above. (The white trace is explained further in the text below.)
This trick does not work as well however with dual sided baffles, the bump gets wider and wider with wider baffles, influencing the response up to several hundred cycles. A baffle was constructed out of two sheets of plywood for a total width of 101 inches. Result can be seen as the white trace in the plot above. The one sided version (yellow) is significantly flatter than the dual-sided version. Proper sizing of the one-sided baffle will place the first magnitude peak below the cut-off frequency of the ribbon. A crossover will add in "cleaning" it up as illustrated in the previous graph.
Copyright (c) 1997 - Rudi A. Blondia - All rights reserved - Last update September 7, 1997.