Reflections are not entirely a bad thing, since they let our minds know we are in a certain environment. Without that knowledge, we can feel a little uncomfortable and claustrophobic. The sound from most recordings is also made for a normal room environment where reflections exist, so in a room with no reflections music seems rather lifeless. The key is to understand how the reflections should arrive at the listeners position in order not to degrade sound quality.
To do this we need some listener testing, and such has been carried our by Olive and Toole. Two speakers were set up in an acoustically dead space. The listener was sat directly in front of one, with the second speaker some distance to the side. The signal sent to the off-side speaker had variable level and delay. This allowed the simulation of a side reflection. The level and delay were systematically altered to ascertain the level at which the reflection became inaudible, and what happens as the level increases. The results can be seen in this chart.
Lets look first at a reflection arriving 10ms after the direct sound. We can see that if it is more than -17dB below the direct sound, it is inaudible. As the level is raised and the reflection becomes slightly audible, the effect is to add spaciousness. As the level gets higher again, now -3dB below the direct sound, the ability of our mind to detect the direction of the first arriving direct sound is affected. The location of the sound seems to smear and shift toward the reflection. As the level raises to become equal to, or louder than the original sound the reflection is heard as a distinct echo.
We can gather from this that some level of delayed reflection is a good thing for a listening room, because it adds spaciousness and a sense of life. However, not to detract from the quality of the sound it must be below a certain level. That level is dependant upon how long the gap is between the direct sound and the reflection.
Let us now pay special attention to the first ~5ms of the graph. In most domestic listening rooms it is very common for reflections to arrive within this period. Looking at this time of arrival, we see a sharp increase in listener sensitivity. A reflection arriving only 5ms after the direct sound can cause image shifting even if it is -6dB below the direct sound. If this reflection is not to detract from the quality of the sound it needs to be about -15dB below the direct sound. Unfortunately the earliest reflections are always highest in level!
There is another point to consider about the time of arrival of these early reflections in the first ~5ms. That is, they colour the tonal balance of the sound more than later arriving reflections. Study this simulation of a reflection arriving 2ms after the direct sound, -6dB down. We see quite wide dips and peaks in the response between 300Hz-3KHz and these will cause the tone of the music or voice to be unnatural.
If we now compare this with the simulation below of a reflection arriving 20ms after the direct sound, we see far narrower peaks and dips. When the peaks and dips are narrow and plenty, our mind averages them out easily to hear a smooth response again. Tonal colouration is thus pushed down away from the critical mid and treble range.
This leads us to the importance of the initial time-delay gap. In order to achieve uncoloured sound, with realistic imaging, reflections within the first ~10ms must be at least -12dB below the direct sound. If we want to add spaciousness, then reflections should arrive after this, and also be at an appropriate level not to be separately identified. Late arriving reflections should naturally be more diffuse than early ones and this also helps the case not to cause tonal colouration as the energy of the reflections is spread out through time.
There are essentially two ways to achieve the goal of a clear initial time delay gap. The obvious and most common method in a domestic listening room is to place absorbers on the room boundaries where those initial first reflections come from. Reflections that take a longer path and arrive later are not generally a problem, and although it is better that these reflections are diffuse, it is my experience that adding diffusers in a typically sized domestic room reduces the decay rate too much. One method of creating an ITD can be seen in the diagram below, and the likely results in the ETC plot.
ETC With Early Reflections
ETC Without Early Reflections
The second method, and more ideal in my opinion is to shape the room such that reflections from the side walls and ceiling don't bounce to the listener but head towards the back wall of the room. The back wall can then house specifically designed diffusers. The result is a room where all existing reflections are well diffused.
ETC For Geometric RFZ Room