Eberhard: Thanks for the note on the surround forum. The issue of decorrelation is very important, and perhaps at present unresolved. It has been my experience in everything that I do that decorrelated reverberation sounds more natural than correlated reverberation. It also makes sense in theory. Four or more decorrelated signals played back in a listening room will give a reverberant sound field that is closer to the natural sound field than correlated signals. (At least in most rooms - this issue is addressed later in this note.) It is also my experience in installing Lares systems. We go to considerable effort, particularly in small rooms, to keep the reverberation between different loudspeakers completely incoherent. Any correlated component at any frequency is quite noticeable, as the bass tends to localize internally instead of externally. However, in the Lares experiments the low frequency LEVEL of the reverberation is carefully controlled. We make very sure the overall level of the reverberation does not suffer because of the decorrelation of the drivers. In fact, in LARES we have found it often essential to boost the low frequency reverberant level. This boost overcomes the decrease in low frequency reverberation we find in many (bad) halls. (If the hall were good, we would not be installing a Lares system.) Good halls have a natural INCREASE in the low frequency reverberant level. However, I have not done formal listening tests to prove the observation that decorrelated reverberation is always better. The recent experiments of Theile seem to show that a correlated signal (at least a signal picked up from four omni microphones spaced by 25cm) sounds more natural in a 3/2 playback than the signals from four widely spaced microphones. As you know, four omnis spaced by 25cm will be strongly correlated below 300Hz. I think it is essential to understand the reasons Theile got this result. Four closely spaced microphones differ from four widely spaced microphones both in the reverberation and in the direct sound. Which of the two components is more important in Theile's experiments I do not know. This would depend on the distance between the orchestra and the four microphones, as well as on the size and the reverberation radius of the space. If we believe the reverberation is the more important component, it is essential to match the level of the low frequencies in the playback room. Four decorrelated signals will result in a lower level of bass in the playback room than four correlated signals. Thus when we space the microphones more widely there will be a reduction in the level of the low frequency reverberation in the playback room. In this case the low frequencies could be more natural spatially with widely spaced microphones, but weaker and less audible. The difference in level could be 3dB or more at 100Hz. You and I both know that the level of the reverberation in this frequency range is vitally important to the envelopment in a recording - we frequently equalize the rear channels to achieve the necessary bass level. In my own TWO channel experiments with a 90 degree phase shift at low frequencies, I found that when the shift was added a 2.5dB bass boost was required to maintain an equivalent bass level in several playback rooms. Since we are dealing with FOUR decorrelated signals here, a ~4dB boost might be needed. This is a very large difference in bass level. In theoretical simulations of reverberation audibility (with orchestral music) on a computer, I found that there is typically a 3 to one expansion in audibility for reverberation level - in other words a 1dB increase in reverberant level results in a 3dB increase in reverberation loudness (envelopment). If we have a 4dB difference in reverberant level in these experiments, there could be a 12dB difference in envelopment - enough to dominate all other aspects of the experiment. I think it is essential in Theile's experiment to compensate for the reduction in bass level when the microphones are more widely spaced. When this is done, the results might be different. There are a number of ways the bass level in the room could be measured and controlled - for example a dummy head could be used to record the level in the room, and then the recording could be compared on a computer to determine the average difference in bass - possibly between 50Hz and 200Hz. I am more than willing to perform these experiments. I can also calculate the compensation required based on some assumptions about the playback room - perhaps this would be a good idea. We could then compare the calculation to an actual measurement. To be completely fair - there are situations where complete decorrelation of the reverberation could be too much. In a natural reverberant field, there are two axis that produce medial reverberation - the up/down, and the front/back axis. There is only one axis where the reverberation is decorrelated between the ears, the left/right axis. Thus to a listener in a three dimensional diffuse reverberant field the sound energy is 2/3 correlated, and 1/3 decorrelated. If we reproduce reverberation through two loudspeakers on each side of a listener in an anechoic space - or through headphones - the spatial properties will seem unnaturally wide. The same thing can happen in a room where the loudspeakers are all located in a horizontal plane around the listeners, and the ceiling is heavily damped, with dropped acoustic tile or bass traps. This type of room is sonically only two dimensional, and fully decorrelated reverberation might sound too wide. The question is, how common are such rooms? A second question might be - do you find excess spatiality in the reverberation pleasant or unpleasant? Personally I find it pleasant, but it is quite possible this opinion is not universally shared. It is possible that the amount of decorrelation at low frequencies in a surround recording might depend on the acoustics of the playback room and the taste of the listener. In this case it ought to be possible to make a control that would blend the bass for these situations. Certainly an experiment with a blend control for low frequencies would be easy to do. Once again it is essential to correct for the increase in bass sound pressure at the listener as the blend is increased. One further observation. The "bass enhance" circuit on the Lexicon processors inserts a 90 degree phase shift between the left and right speakers. The circuit has no effect on decorrelated reverberation, but does effect the sound of a strongly correlated signal, such as a kick drum or bass guitar. The perceptual result for all listeners is a reduction in bass level, and a perception that the signal is external to the listener, rather than inside the head. I find this circuit very pleasant. The opinion is widely shared among our customers, the great majority of whom are more than willing to put up with the decrease in bass level to obtain the external localization. However, the opinion is NOT universal. Particularly when people who have been listening to conventional stereo in dead rooms for a long time, they come to expect the in the head localization of the bass as the mark of a good system. This has turned out to be particularly true in automobiles. We have found that although it is highly desirable to have the reverberation decorrelated, the bass guitar and the kick drum need to be correlated to make the sale to highly trained automobile sound engineers. So I typically turn the circuit off when demonstrating the system to engineers, and turn it back on when I listen for myself, or with the general public. It is possible that the preference for LF correlation in reverberation might similarly be a matter of expectation. Hopefully this issue will soon be resolved. Feel free to post this note if you think it would be helpful. dg