First, it is not that an MTM arrangement "sounds not so good." Joe D'Appolito made this arrangement famous and the design does offer increased sensitivity and lower distortion due to the fact that dual woofers are utilized.
Well, I guess my wording was "not so good" as I guess I should have simply asked why he chose the MTM design over the TMM design given two identical speakers and cabinets and what were the advantages of each.
Let me answer that one. The same sensitivity improvement exists if the design is an MMT vs an MTM, so why go MTM? The answer has to do with what is called "lobing". I do not know if you are familiar with this term or not. But here's how it works:
Any time you use two drivers that have to cross over between each other at a certain frequency, and if the sound from these drivers does not originate from the same point in space, then you have to account for the differences in path length between them in the design.
As a result any midbass - tweeter combo that is designed to sum flat on a particular axis will have changes in this summation as you move off-axis because as the path lengths change so does the phase between the two drivers (due to delay from the change in path length). The result is that all two-way speakers have what is called lobing where the summation is perfectly in phase on one axis and "lobe" to a cancellation on another axis creating a null off-axis somewhere.
Since stereo is based on horizontal separation (L + R) and we tend to move around some horizontally, but not so much vertically, we tend to arrange drivers vertically to reduce the impact of this lobing. You do not want horizontal lobing if it can be avoided because it really messes with stereo imaging, however, vertical lobing is much easier to live with, and this is why nearly all speakers are arranged vertically.
So this brings us to the main difference between an MMT and an MTM. In an MT there will be one predicted vertical pattern of lobes and nulls along the vertical axis. However with an MMT the separation between the two midbasses results in two different drivers both crossing over to the tweeter but with different path lengths. This means that there are now two independent sets of nulls and lobes, which begins to take the form of some complex comb filtering, and can be quite audible in the midrange where wavelengths are getting shorter and the effect is getting greater. If you use a 2.5Way and roll one midbass off earlier, then you may not get the sensitivity you need in the crossover region.
However, with an MTM arranged vertically the path length changes between the upper and lower midbasses will always be the inverse of each other anywhere along the vertical axis, and the lobing will remain symmetrical at all points (mean always the same above and below the tweeter axis). We can not eliminate lobing, but we can make it much more manageable by making it symmetrical. If you are careful with the crossover design you can actually design it so the lobe of one woofer begins to fill in the null from the other woofer and you can really reduce the impact significantly. This was the basis of D'Appolito's 1984 AES paper that started all of this and he and I have discussed it in depth. This control of lobing and the fact that it is always symmetrical is the reason so many speakers use the MTM arrangement, and it is advantage it has over the MMT.
Jeff B.