Dayton Tweeter Tests and Modifications

The first part of this report showed the initial raw responses. I've now experimented a little. The measurements on this page show the drivers after rotating the dome while it was in place in the magnet gap to try to better distribute the ferrofluid. Others had tried this with some success. It did seem to help with mine as well, though it isn't a big improvement. It is, however, worthwhile, as the fluid really should be evenly distributed.

The impedance plots (graph 1) didn't change a lot in doing this, although the three good ones are now all centered at 950hz, very uniform. The magnitudes are reasonably close as well. The variation at Fs is probably not of much importance given where it is. Below 2khz, the variation may cause some slight difference with the crossover, but I suspect not a lot, unless a low order is used.

Note that above 2khz, they are almost all identical, enough that there won't be any variation with the crossover (as far as the impedance is concerned, anyway).

My earlier measurements had some significant noise in them, though this didn't prevent seeing the trends. Modelling in a CAD program also gets rid of this. My setup is better isolated now, but there is still some noise. I used the MLS signal for these. The swept sine wave (stepped, actually) did not pick it up, I suppose because the frequency steps were not exactly at any multiple of 60hz. Note the fundamental 60hz plus the 2nd harmonic (fairly low at 120hz), the 3rd, 5th, 7th, 9th, 11th and even a very small 13th harmonic. I'm guessing that this is either from the internal power supply or I need to move the PC if I'm to eliminate this. It shouldn't be a problem for speaker testing purposes unless distortion measurements are to be made.

There are two problem areas, these at 350hz and 450hz. They are due to some form of resonance, the origin of which I don't know, probably a resonance in the dome/voice coil combination.

The next graph shows an SPL comparison of the on-axis response of several drivers. They are the original version of the tweeter (Unit C) after dome rotation, an example of the new version of the tweeter and the surprising result of a test I made. I'll be adding a page to show this in more detail, but basically I took Unit D of the original tweeter lot (it had the dome/coil attachment problem) and substituted a Morel MDT-32 replacement dome. Interestingly, the general shape of the curve is the same as for the Dayton Unit C, but it's 1-2 db higher in level and is the smoothest response of any unit I've tested.

The new version of the Dayton (the only one of two I have which I've tested) doesn't look good. This is counter to the accolades others are reporting after listening to them. My unit may just be an anomoly. I will add tests of the second unit later (when I can find which box it's in, as I just moved!).

These Daytons appear to be a hybrid. They look like the MDT-29 motor (no chamber) but the dome/faceplate is very close to the MDT-32. That's how I got the idea to try a replacement dome, to see how much influence there is in it. The graph shows the importance of the dome/voice coil combination. The MDT-32/33 dome is the only one which can be used as it is the only one with the voice coil connectors attached to the dome assembly. The MDT-29 and -30 have the voice coil connectors attached to the motor assembly top-plate.

I've been puzzled and surprised while measuring SPLs. The original Dayton (blue curve above) looks very good for the inexpensive tweeter which it is. On-axis it's shape is similar to some Morels. Three of the original four are consistent. I started to experiment with a chamber for Unit A, so it's not available for this comparison. The comparison of Unit C before dome rotation verses after rotation (below) shows SPL change, but it's debatable whether or not there is real improvement, at least in this unit. Given a driver which was already a good example, not much has been gained in the SPL, unlike the slight improvement in the impedance. In fact, given the larger dip at 7.5khz, it could be argued that it is worse. I may try "burning" this one in (since I haven't used them in a system) to see what happens with them. The improvement depends upon the state of the driver at the start (duhhhh!).

I've wondered for some time what the impact is from the faceplate and what resonances are inherent in the driver alone. Normally a driver is mounted on a baffle. Separating the resonance contribution of the driver from that of the baffle is not possible when it is mounted. However, resonances which are exhibited when not mounted are likely to be partially or wholly damped when mounted. Despite this, I thought that it might be instructive.

To this end, I attached a small plastic-cased accelerometer to the faceplate of several of the drivers. Two conditions were tested, one with the driver unmounted and on its back plate lying flat and the other with the driver mounted in a very small, 1.5" thick small baffle lying on its back. The results for Unit C (an original version) are shown, both a frequency response and CSD (waterfall) plots. The differences are striking.

The impulse graphs show just how much difference there is in the time domain. The yellow graph shows the unmounted driver. The red graph is the driver mounted the small baffle. Both were measured lying flat on a solid table top.

In the first few micro-seconds the signal looks identical. From there the two diverge drastically. At first they just look like variations in resonances. By the time they reach 5 msec, it's obvious that the baffle is damping the ringing which continues in the unmounted situation. The time-limited frequency response curve is shown below.

The frequency response graphs can't be easily interpreted on their own. As the graphs show, in some cases a dip turns into a peak and vice-versa when mounted on this baffle. Part of the problem is that it is likely that if I attached the accelerometer somewhere else on the faceplate, the signature would look different. I made the two measurements without moving the accelerometer in an attempt to minimize this type of difference for the comparison measurements. So what to do?

We can get a better idea of what is happening by looking at the accelerometer CSD (waterfall) plots (on the next page). This will let us see what the resonances are doing over time.