There are different ways to develop frequency crossovers. One person might take the manufacturer’s measurements and plug them into his preferred simulation program. He puts together the pieces based on the resulting values and is happy that the tweeter plays high notes and the woofer plays low notes. That’s all this approach is good for. Another useful method is to put the self-created measurement data into the simulation, then to disconnect as many circuits as possible and set the program to “optimize.” That at least creates usable results, with a fairly smooth frequency curve. During later measurements, you usually only need to make a few adjustments for the simulation and the reality to match up. Unfortunately, this process often tempts people to remove all of the peaks from the curve, even the small ones. The signal can often get lost among all the components, and then it barely makes it out of the intended membrane. So this kind of crossover development also requires a lot of experience to achieve good results. The great advantage here is in using fewer components whose legs tend to break off just at the wrong moment after you’ve used them a couple of times. Still, we should mention that even this observation, while plausible at first glance, is only half true. In order to gather experience, there’s really no way to avoid building large numbers of crossovers, and that means having all kinds of hardware in the form of coils, capacitors and resistors. Since you also have to take measurements after the simulation, you can also – again assuming plenty of experience – start with the measurements right away and save yourself the trouble of entering the data into the computer.
Even while measuring the frequency curve of the CA 12 RCY, we noticed the distinctive hump caused by the baffle step around 800 Hz, followed immediately by a serious dip at 1.8 Hz (green). We have noticed this quirk for some time now in all of the new SEAS paper membranes.
Now, it’s not a crime to put an anti-resonant circuit in the signal path in order to bring the mountain down to the valley, and then to force a smooth frequency curve with the remaining crossover at the expense of the volume. But we should remember that we’re dealing with a dwarf loudspeaker, one that will sound much too thin if it tries to come up with the whole mid-range. It was by no means the chassis developer’s goal to create a mid-range speaker with the CA12 RCY. The little guy should be able to provide a little bass pressure and undertone warmth, so the loss in volume can actually be helpful. So we didn’t even try to fill in the middle, but filtered it right away with 2.2 mH (red) and 4.7 µF (blue). That also dropped the resonances at 7 and 9 kHz down low enough that they couldn’t disrupt the other tones.
At 3.5 kHz, the narrow baffle board creates a deep indentation in the amplitude of the NoFerro 800 TV, which we think is one of the best calotte-based tweeters (purple). That didn’t worry us, though, because it had already disappeared in the 15-degree curve. First we lowered its volume by about 7 dB, using an L regulator at 3.3 and 4.7 ohms (green). At 4.7 µF (orange) and 0.33 mH (red), I thought the curve dropped off a little too soon at lower frequencies. So we replaced the capacitor with 5.6 µF and were happy with the results (blue).
At the same polarity, the total of the two filters produced a mid-range trough almost an octave wide, forcing the voices back a little and creating the illusion of spatial depth. At the same time, tonal errors caused by exaggerations – which can be especially disruptive in this area – don’t stand a chance of scraping your ears. We can live with the justified complaint that we haven’t built honest loudspeakers this time. Because why in the world should every box be used to analyze the faults of the recordings? Small boxes can quickly become annoying because they’re missing the foundation “down below”; the MS-Micro intentionally has more of that than you would expect.
The history of the MS-D’Appo crossover is easy to tell, because there isn’t anything too special about it. The parallel CA 12s initially provided a prettier frequency curve than the lone warrior in the Micro. The goal here was to keep the sound level at the 88 dB already available at 300 Hz. To do that, we needed a 1.2-mH coil, which naturally had too little effect on too-high frequencies (red). 10 µF helped energetically here, even giving us a volume gain up to a good 3 kHz (blue).
The taller baffle board largely saved the tweeter from the hole at 3.5 kHz, but two resistors with stole about two dB of its sound pressure. Unlike its sister, we connected it a good bit lower, with 2 inductors, which strongly compressed the dip in the middle
Even though the Micro was not designed to fill large living rooms with sound, its 84 dB of mid-range sound pressure and its -3dB point of 70 Hz mean that it no longer belongs to the class of quiet speakers with a 12-cm bass. Even in the 90-dB noise (Klirr) measurement, it has impressively low distortion values. Only below 300 Hz does the K2 increase above 1%, and K3 through K5 are still far from that. We performed the angle measurements at 5-degree intervals from 0 to 90 degrees on a turntable. Unfortunately we couldn’t place the rotation center on the baffle board, so the volume decreases as the angle increases, due to the change in distance from the microphone. However, it does have the advantage of keeping the curves from turning into an ocean of colors; instead they are almost all visible individually. The second waterfall, which in reality shows the clustering behavior, is made up of all the measurements, while the angle diagram only includes the 15-degree intervals between 0 and 90. We will leave the interpretation of the remaining results up to the expert reader. The minimum impedance, at 5.4 ohms, is 6.5 kHz; the maximum, at 33 ohms, is 1220 Hz. Using 33 µF _ 0.49 mH _ 6.8 ohms, it is filtered out for tube operations.
As some of you might expect, the volume level of the D’Appo 4 is not 90 dB, which theoretically should have been reached by doubling the CA12. Still, an average of 88 dB isn’t too bad for such a skinny little guy. The 90-dB noise values are correspondingly lower for this box. At 4.1 ohms, the lowest impedance for the D’Appo is 470 Hz; the maximum of 18 ohms at 1350 Hz is smoothed out by the absorption circuit consisting of 33µF _ 0.39mH _ 5.6 ohms