Ambisonic (Sound with height)


    Jul 17 2013 | 9:02 am
    I would like to try setting a surround system as Ambisonic with only 4 speakers Fostex NF-01A (not enough amount of speakers due to my financial problem). I know it's difficult to achieve sound reproduction in both horizontal and vertical directions around a listener. But, I wish somebody helps me tiny reference to me. What I want to know:
    * Room size would be 10m2 about 3m heigh 1, How to setup viable with-height 4 speakers layout 2, How to program surround sound with height panner patch (mention me how different from the Quadrophonic)
    Any reference URL, Movies, Patches and so on that would be helpful to me.
    Thank you very much!

    • Jul 17 2013 | 9:49 am
      Hi, if you are using 3d ambisonics you can have height, but with 4 speakeers you can only have 1st order ambisonics... Have you ever considered using conversion to binaural ? I am currently developing a patch for this purpose. It wouldn't cost you a penny, you'd just need good headphones, and no sweetspot problem. Otherwise I know a guy who developed an encoder and a decoder for Ambisonics and that is free of use, but I guess there are also libraries for this in Max. Regards Coralie
    • Jul 17 2013 | 10:49 am
    • Jul 17 2013 | 11:41 am
      > Coralie Thank you very much for a comment. With 4 speakers that should be not enough to provide a surround preproduction, but I cannot choose a binaural system. This project can be for a sound installation. So it should be open for everyone with no any tool. I do appreciate your mentioning anyway!
      Cheers
      > Woyteg Thank you for sharing fantastic info! You are amazing. I will check them out, see what happens with lovely will!
      Regards
    • Jul 18 2013 | 6:26 am
      4 speakers should be enough to cover half-a-sphere, but you would have to write the ambisonic panner for this setup all by yourself. while it is easy to do the direction/azimuth (because it does not differ much from using 4 or 20 speakers), i had no clue how to calculate the weighting between the "ceiling" speaker and left/right/rear for the elevation parameter.
    • Jul 18 2013 | 10:07 am
      >ROMAN THILENIUS Thank you for sharing the info! I will try my best to do for it. I am still wondering how 4 speakers should be located as half-a-sphere. Any picture about it?
      Cheers
    • Jul 18 2013 | 10:38 am
      Are you sure your layout is the optimal one ? I mean, i have little experience, but i would rather do like this (see attached image - notice the graphic quality, all made in max.) because it seems important to have a front stereo balanced image - it has more importance than the rest in spatialization - but maybe not ? actually your setup seems maybe less balanced but maybe better. I'd be glad to hear the result or if you can compare those possibilities !...
    • Jul 18 2013 | 11:24 am
      @roman "i had no clue how to calculate the weighting between the “ceiling” speaker and left/right/rear for the elevation parameter." In the atteched screenshot one can see a sound source(or speaker) 1: top front left and 2: rear right bottom. What was the problem with weighting the ceiling? cheers!
    • Jul 19 2013 | 4:12 am
      Bad news, I am afraid. You cannot set up a viable full-sphere Ambisonic layout with only four speakers, not even first-order. From the Wikipedia article you referenced:
      "A minimum of four channels of information are required for distribution and storage of a full-sphere soundfield, and three for a horizontal soundfield. (This is fewer than other surround systems). Full-sphere replay requires a minimum of six loudspeakers (a minimum of four for horizontal), the signal for each speaker position being derived using appropriate circuitry or software."
      The tetrahedral arrangement you reference will only work for frequencies below about 700 Hz. Frequencies higher than this will be pulled into the nearest speaker.
      The "bible" for full-sphere Ambisonics is the paper: M.A. Gerzon, "Practical Periphony", Preprint 1571 of the 65th Audio Engineering Society Convention, London (1980 Feb.)
      The paper states on page five:
      "The tetrahedral speaker layout shown earlier in fig. 2 does not satisfy the diametric decoder theorum, and the Makita and energy-vector localisations do not coincide. In fact, computations of the energy vector localisation show that sounds at high frequencies are very much drawn towards the four loudspeakers of the tetrahedral layout, ... This problem of sounds being pulled towards the speakers was in fact noticed in early experiments in tetrahedral recording, and is the reason why other speaker layouts must be used."
      Regards, Martin
    • Jul 19 2013 | 4:36 am
      >VICHUG Thanks for sharing the image! I will definitely refer from it to test sound moving check.
      >MARTIN LEESE Thanks for sharing the info. THat is too bad to me about this fact in using 4 speakers which is not enough for vertical direction.
      So, please tell me everyone, what recommends nice speakers for Ambisonics? I thinks I have to buy new ones, and looking for them as cheapest as possible including speaker stands.
      Any recommend?
      What I currently have 4 speakers is Fostex NF 01-A though,,,,,,,,,,,,
      M
    • Jul 19 2013 | 5:32 am
      re:wikipedia: i believe this only true when you also have a perfect room and listening situation. if 3 speakers are enough for 360° it sounds strange that you should need more than 3 for performing 180°. but i think i have a wrong picture in my head when i try to pan from the result of the vertical panning to the horizontal panning. :)
      @woyteg: you are right, in theory left-ceiling and left-right is the same thing. but as it seems it is not that easy in practice. if a 4-speaker setup works worse than using binaural filter to do pseudo-elevation on a stereo setup, it probably does not make much sense.
    • Jul 19 2013 | 6:17 am
      Roman wrote: "re:wikipedia: i believe this only true when you also have a perfect room and listening situation. if 3 speakers are enough for 360° it sounds strange that you should need more than 3 for performing 180°. but i think i have a wrong picture in my head when i try to pan from the result of the vertical panning to the horizontal panning. :)"
      I suspect you have not spotted that Ambisonics separates distribution channels from speaker feeds. The speaker feeds are derived in your living room using a decoder and, in general, you need more speakers than transmission channels. You might want to re-read the extract I quoted above. (Also, I don't understand your terminology: 360° is horizontal-only surround sound, while full-sphere surround sound cannot be expressed using planar degrees.)
      Regards, Martin
    • Jul 19 2013 | 8:02 am
      well what is was thinking is more "if you only have 4-5 speakers, the speakers will and must not differ from distribution channels."
      for example i would not see any reason why you should use 4 distributions for horizontal when you only have 3 speakers. and if i had to use 10 speakers i would ONLY arrange them in a circle or square ... and then use a formula for the distribution/panning which matches that setup.
      2.) well in a sphere positions of 0-360° exist on every axis you want, but if you only have half a sphere an imaginary x-axis is only half of that. you can not at all perform a movement from hard left to hard right unless you add a source-listener-distance parameter.
    • Jul 19 2013 | 6:28 pm
      Roman wrote: "well what is was thinking is more 'if you only have 4-5 speakers, the speakers will and must not differ from distribution channels.'"
      That is not how Ambisonics works. With Ambisonics, the distribution channels are called B-Format. For first-order full-sphere these are W, X, Y, Z, and correspond to the outputs from an onmi and three figure-of-eight mics. For horizontal-only, the Z-channel is simply omitted. For higher orders, extra channels are added. For horizontal-only Ambisonics, the minima are three transmission channels and four speakers.
      Also, in Ambisonics any panning is performed in B-Format; you simply place your sound source in the soundfield at the azimuth and elevation you want. (Panning through the listener is also possible, but trickier.) Production of the speaker feeds is handled by the decoder; at the mixing stage you neither know nor care where the speakers are located.
      Obviously, if you use something other than Ambisonics then everything is different. However, the original poster wrote that they want to use Ambsionics.
      Regards, Martin
    • Jul 19 2013 | 10:34 pm
      ok, so you are saying this (picture) works with 4, 8, or 100 speakers, but not with 3?
      funny thing is, you guessed right, i am actually using something else in practice. mostly i am using a panner which only works for speakers set up in a circle.
      eventually i am using something else because i dont really know how to use the ambisonics stuff. or maybe my encoder/decoder are just wrong. :)
      (however, when it doesnt work with his 3 speakers maybe it wouldnt be the worst idea to use something else, too?)
    • Jul 20 2013 | 4:51 am
      Roman wrote: "ok, so you are saying this (picture) works with 4, 8, or 100 speakers, but not with 3?"
      I am saying that your picture has no W-channel, so cannot be Ambisonics. I do not know whether it will work or not. (I do not understand it.)
      Regards, Martin