1. Gary Eickmeier says:

    Geoff – I say again, they are operating on the wrong stereo theory if they are trying to eliminate side wall reflections. Stereo is not caused by the direct sound alone. Rather, it should be modeled after a typical live sound field. With live sound a small amount of direct sound is followed by a rich diet of reflections, the side wall reflections giving us the spaciousness and the front wall reflections giving the depth of imaging beyond the speakers themselves and making them disappear as obvious sources of the sound. A correct design would have a radiation pattern that is basically Omni but with about 6 dB more output in the reflecting directions. You could at least include that as one of your patterns.

  2. Hi Gary,

    The BeoLab 5 had three primary goals with respect to horizontal directivity. The first was to have as constant a directivity as possible to ensure that the magnitude response of the direct sound from the loudspeaker was as similar as possible across a wide range of angles on- and off-axis in front of the loudspeaker. The second was to have as wide a “beam” as possible across a wide frequency range to ensure that the direct sound was not suffering from beaming if you were sitting off-axis (say, on the end of the sofa). The third was a limit on that beam width (if my memory is correct – although it often isn’t) of no higher than a target magnitude at 90° off-axis relative to the on-axis response. This third limit was derived from the results of the Archimedes project which, in part, determined the minimum magnitude at which a wall reflection would affect the perceived spatial impression of a sound source. If you would like more details on the results of the Archimedes project, please let me know and I’ll post a bibliography of the documentation from that research.

    In the end, it turned out (although it’s not surprising) that the wide directivity of the BeoLab 5 had the added benefit of making the loudspeaker’s power response smoother when equalised to have a flat frequency response. Although the on-axis response of a loudspeaker is interesting, for most of our products, we are at least as interested – or possibly even more interested in some cases – in the power response and the frequency-dependent directivity. This is primarily because a typical use for a loudspeaker for “normal” people is to listen to it when you’re NOT sitting in the “sweet spot”. Therefore, the on-axis response for most people most of the time is of little importance.

    The BeoLab 90’s “Wide” mode was designed to be an improved version of BeoLab 5’s horizontal directivity for all the same reasons as the BeoLab 5 – in addition to one more reason. There are customers who have systems with BeoLab 90’s in the front and BeoLab 5’s as their surround loudspeakers. In this case, it makes sense to have loudspeakers with (fairly) matched directivity patterns to provide a more coherent sound field when the signals in the front and surround channels of the recording are correlated.

    Of course, all of this discussion is only related to directivity. There are many other factors that are worth considering as well, including (but not exclusive to) temporal response, overall spectral balance of the power response, and smoothness of the power response curve, thermal compression compensation to ensure that the loudspeaker’s balance and directivity is maintained regardless of changes in voice coil temperatures, and all the various versions of distortion…


  3. Gary Eickmeier says:

    Thanks for that Geoff. From my experience the freq response of the direct sound is of far less importance than the power response, the same as with the live sound situation. If you have a radiation pattern that puts more sound into the reflecting direction and make the power response as desired (room curve, sometimes called hall curve), and position the speakers so that the image model spaces the actual and virtual sources equally, you have better imaging, better freq response for everyone, and less of a sweet spot problem – a totally happy result from any design standpoint. Such an approach would not likely occur to most audio engineers who are trained to think the direct sound is what stereo is all about.

    Yes I would like some references to the Archimedes project results if convenient. I tried to talk to Soren Bech about it during that time, but he never got back to me. I wish I could have participated in it.

  4. Hi Gary,

    I’ll put together a bibliographic list for you next week when I get back to work (today is a holiday…).

    One thing to keep in mind regarding the Archimedes research is that they were simulating single sources – not coherent groups of sources (such as a pair of stereo loudspeakers being used to generate a phantom centre image). So, the results of that research have to be interpreted correctly without jumping to conclusions regarding stereo (including multichannel) systems.


  5. Gary Eickmeier says:

    OK thanks. It’s just that they had an incredible chance to simulate any image model they could dream up in that space at the push of a button. I had mine for them to try, but alas.

  6. Patrick Bateman says:

    Wow, there’s a lot of great info here. I’ve built a ton of those SAW lenses, and was mostly guessing on what the design rationale was. I believe this page has the most info yet, on what motivated the design.

    If you’re curious about my humble attempts to make those lenses, here’s a few threads I wrote at diyaudio:



    I also took a stab at the beam steering from the Beolab 90: