@polkalover The most difficult aspect of audio system performance is imaging. There have been many articles written on the subject. Relatively few audiophiles have heard a system image at the state of the art. Most systems can image well to a point, about 90%. It is that last 10% that is difficult to achieve. @mikelavigne made some great suggestions, limiting the number of crossover points and room symmetry are very important. The speakers have to see identical, but mirror image environments. I would also like to add a few points and suggestions. Early and even some late reflections can ruin imaging. There are three ways to optimize limiting reflections, sitting closer to the speakers has it's limits, choosing speakers with controlled dispersion and finally acoustic treatments of the room. Horns and ESLs are examples of speakers with controlled dispersion. ESLs can be formatted as full frequency line sources which IMHO is the speaker type with the best radiation pattern of all. They send virtually no sound to the side walls, ceiling or floor. An example would be any of the 8 or 9 foot Sound Labs speakers. Omnidirectional speakers will never give you the ultimate image in any room and you have to use a lot of acoustic materials to get even close.
Next is an issue you will not see covered often and that is identical sonic performance between the two channels. No two speakers are exactly alike and you can not place them in exactly the same location. They will have different amplitude response patterns or rather, they will not have identical frequency response curves. When one speaker is louder than the other at any given frequency the image is smeared towards the louder speaker. Thus the image can be smeared back and forth at various frequencies killing the 3rd dimension, that last 10%. The third dimension is not instruments at different distances away from you, any system can do that and much of it is artificially created by the mixing engineer with echo. The 3rd dimension is the sense that and instrument or voice is a 3 dimensional object in space. If that space is full of reflections, echo and amplitude smear you will not be able to delineate the 3rd dimension at all. This is the state of most systems including some incredibly expensive ones. The only way to achieve sonic symmetry is by measuring each channel independently then adjusting amplitude over the frequency range. I limit adjustments to between 100 Hz and 12 kHz. You do not have to have a flat response, you have to have identical frequency response curves. Flat actually does not sound so hot. Studies have shown that the best response curve in most residential situations is boosted bass below 100 Hz with about 2 dB/oct attenuation above 1000 Hz.
Imaging is always going to be best at the listening position and on a line perpendicular to the speaker axis through the listening position. There is no such thing as a "wide sweet spot". However, it would be nice to be able to hear the far speaker clearly and again line source speakers are best at this because volume does not fall off near as drastically as point source speakers when you move away from the loudspeaker.
As Mike Lavigne suggests accurate bass is a vitally important aspect of system performance and easily the most difficult portion of the spectrum to optimize. The reason is interference patterns in residentially sized rooms and resonance of this that and the other. Bass is incredibly physical. If I play a 30 Hz test tone at 85 dBs you can hear a symphony of buzzing and rattling in the house. This is usually completely masked by the music. I attack only the rattles I can hear. Room control can only do so much and subwoofers often do more damage than good, but sometimes you have to dance with the devil especially if you were to chose another ESL, and that is a whole other subject. The problem with your MLs is they cross to a dynamic driver at 375 Hz, right in the most important part of the midrange. Remember 256 Hz is middle C! The electrostatic portion of the Ethos can not handle lower frequencies because it is curved. An enclosed dynamic speaker is dramatically different than an ESL. The difference below 100 Hz can be gracefully dealt with, but above 100 Hz not so much. IMHE no ESL can gracefully handle bass below 100 Hz and this is where you have to dance with the devil.
@toddalin This is a great point. You can not know what SOTA imaging is until you have heard it. You have to know what you are missing and you have to delineate the problem in order to attack it. I have many studio recordings that you can tell the vocalists and instruments are in totally different environments and probably recorded at different times and frequently in different studios. I focus on live recordings or recordings in which the entire band is playing together in the same room. Great imaging helps to delineate individual instruments. A good live recording with a horn section is a great example. A top notch system will allow you to identify each instrument in space. Most systems show you the horn section, but you can not separate the instruments easily, the same applies to vocal sections. The spatial cues are usually there, but acoustic errors can easily overcome them because they are at a much lower volume. IMHO there is no such thing as too much acoustic treatment. An anechoic chamber is better than a poorly treated room.