Hardware Question (multi-channel audio)
I am new to audio hardware, especially hardware that interfaces with a computer.
I have built a 16 channel audio installation that runs off of 2 7.1 surround external sound cards. the four analog outs are connected via a mini-plug to rca cable to the stereo inputs on a car amp (four channel). Each channel runs through an amp to a speaker, so each channel operates completely independent of one another. This works fine and as expected (except on of the jacks on the cheap sound card has failed).
I am currently in the planning stages for building a 144 channel audio installation and am looking at several possibilities. One would be to use multiple boxes such as the MOTO 828mkII, but I can’t tell from looking at the specs if the TRS outlets are mono or digital.
Since all I need is output, another route would be to use a PCI slot box and just run 13 sound blaster 7.1 surround cards from it.
Does anyone know of a good resource for gaining knowledge about how audio equipment interfaces with computers, and what one should look for at each point in the chain (computer, interface or soundcard,amplification, speaker)?
Does anyone want to share how they have built the hardware component of their sound installations?
wow… 144 channels? really?
my first question would be: do you really need 144 discrete channels? none of these 144 speakers will be putting out the same audio?
my second question would be: do you plan on generating all of these 144 discrete channels from one computer (presumably running max/msp)? If so, you’d better have a really fast computer. Remember that no matter how many sounds cards you stuff into one computer, it’s the main CPU that is doing all the work of generating the audio. You’re not "subbing out" any of the work to the hardware on the sound cards. If you’re not sure that your computer can handle it, just try generating 144 signals (of a similar complexity to what you’re planning for your installation) in a max patch and see what your CPU % looks like. Unless the audio being generated for each channel is relatively simple, you will likely not be able to get it done with one computer.
my third question would be: 144 discrete channels of audio means you need 144 speakers and 144 amplifier channels. These types of quantities do not come cheap. I’m sure you know that already.
If you elaborate a little more on what kind of things you’re trying to do with these 144 channels, we could probably help give you some better advice. However, I would estimate that you will need to use at least 2-4 computers to get this done realistically.
I’m not sure what the limit is on how many sound cards you can attach to one computer. It probably depends on the interface (i.e. PCI, Firewire, USB).
right, to first say i like you idea, i have heard of a friend doing something similar, or at least trying, but it is possible. But with an extreme cost added on to, a really big cost.
As for the MOTU side, that is a good plan, as the MK series are great, i have an MK3. I love it. But it is best worked with Mac, this is what the MOTU crew told me.
But this would cost a shit load. As you would need many, plus a digital feed to more channels, maybe using light-pipe, or some other method. Plus the cable would also set you back some.
The TRS connections are mono, but the MOTU does have expansion for many many many more channels on top. But it already has 8-10 physical mono outs. So it is more than capable to have a 7.1 set-up.
But the downside to having many sound cards linked is that you would have to invest in a sync clock, i think it is called. Basically it will keep all of the linked sound cards in sync. But a good one alone is worth about the same amount as a MOTU mk3.
I think what you should do, is be a little bit more realistic with this project. Maybe start off a bit smaller in scale and try on from there, building it up. Or maybe just getting to a really performance with what you may have. But i would suggest a multi-channel sound card like a MOTU.
I love my simple set-up with my MOTU and am even getting a small 5.1 studio built up, using just the MOTU to power it. Soon i will then expand it to 7.1
Hope that helped…
Thanks for the reply
Yeah, really. 144
The plan is to build a 12 x 12 grid of speakers (covering quite a bit of space), then using max/msp, define pseudo random patterns through the grid and move sound along that path. I want to use a grid, rather than say a circle (which could achieve a similar effect in terms of the audio), since the speaker housing will serve as a visual clue to the "location" of the sound for the people in the installation. A single sound will start on one speaker, move to the next, and then the next, while others simultaneously follow there own paths.
I don’t think CPU load will be an issue, since I will only be serving 6-10 sounds across the array at a time.
I have built a 16 channel set up using car amps, speakers and external USB sound cards. but before I start making a budget for a larger system, (which I recognize is ridiculously expensive given the way I want to build it). I need to know a lot more about audio equipment. So I was wondering if anyone had any recommendations for texts or websites that might give an overview of using multiple PCI cards vs. external boxes, the limits of how many sound cards an OS can address, side issues to many channel audio, circuit design to avoid ground loop issues, troubleshooting problems such as that damn high pitched whine that I am getting from one computer but not another using identical hardware, etc.
I have the max patches mostly built. (Best tutorials I have ever seen with a software program). Just need to start learning the hardware side of things.
Sounds like quite a task. I know you mentioned having the grid of speakers presenting a visual cue but you could do something similar with 4 sets (or even just 2) of 12 speakers and some clever panning. You could have them placed as 2 rows and colums that would be the left, right, top and bottom of your grid and you would still be able to position sounds in any position you wanted.
You also need to realize that the ear, in terms of spatial cognition, has only a certain amount of angular resolution: whether or not a listener is going to know which speaker a sound is coming from depends on how far away the listener is from the speaker array and how close the speakers are to each other. Otherwise you’ll be creating all this over engineered specifications for something that could be achieved much more simply, with only a few speakers and a good panning algorithm. Also if you have simultaneous trajectories, sounds can mask each other (from a spatial perspective) and will create ambiguities, in terms of what their perceived source is. Also, if a speaker is a little closer to the listener than others, it will mask the others, due to interaural time delay and ruin any perception of sound source position in space– in other words the grid needs to be part of a surface of a sphere, wiht the listener at the centre of that sphere. This implies that only one listener will be able to appreciate the spatialization of your work. Called the sweet spot. also, the ear is essentially deaf to spatial information below certain frequencies, so if you are doing bass, theres little point in trying to spatialize it.
Even if you’re only activating a few speakers at a time, the cpu may be spending most of its time calculating lots of zeros, unless you have a clever switching system worked out.
I just read what I’ve typed– I do sound rather negative, but I’m realy just trying to warn you of some of the pitfalls– you might already know of all this anyway.
A suggestion: why not have a bunch of dummy speakers in your array– if every second speaker is not connected, a good panning algorithm will create the illusion that a dummy speaker between two active speakers is sounding, then you’d at least halve the number of channels you’d need, and no one would be any the wiser. Or am I completely out of line here?
Thanks for the input. I will look into the panning solution combined with dummy speakers. it would certainly eliminate figuring out how to serve that many channels.
In terms of the original idea, listeners would be inside the array and there would be a minimum of 4-6 feet between speakers. A voice would originate on one speaker in the array and then fade to the next in its path, slowly moving across the array. This would provide a location for that sound for any listener anywhere in the array. Simultaneously, different voices would move in different paths around the array. Initially, i had been thinking that the simplest (not the cheapest) solution would be to have each location wired independently, since with panning, location of the sound is dependent on the location of the listener, but I will certainly look into the panning options in max.
would it be possible to use MSP with some sort of wireless audio receiver? If each speaker was attached to a network, I’d imagine you could route audio to a specific speaker based on its IP address… or would a network like that be too busy? Is that even within Max’s ability?
Check out this thread- may provide an alternative approach:
At a minimum, you’re looking at $15k-$20k for equipment for a setup like this… You’ll probably need two equipment racks that are chock full of amplifiers. And you’ll need a hell of a lot of cable. 144 lines of speaker cable could easily be 6" in diameter.
Your plan can be done, but it will be expensive, and it will be extremely difficult to do well if you have little or no experience. If you actually have the budget for something like this, I would suggest finding or hiring someone to help you out with the hardware design and installation.
P.S. You may want to look into Cobranet for the audio output from your computer. You can send up to 32 channels (in each direction, for a total of 64) over one cat-5 cable with cobranet. So, you’d only need 5 cobranet cards instead of 18 Motu interfaces. The downside is that you’d need hardware on the other end of the cat-5 to convert it back into analog audio. Unless, of course, you find an amplifier with cobranet inputs (which do exist).
To avoid ground loops, do the following:
1. Make your own audio cables. For all balanced connections, connect the ground wire at one side only. Make sure it’s at the same side for every connection (i.e. input side or output side). Don’t connect the ground wire at the other end.
2. Make sure that all the power to your amplifiers is coming from the same electrical panel, and make sure it’s all coming off of the same phase of the power signal. Minimize the other things that are powered off of this panel (i.e. lighting, HVAC, etc.)
I love the project idea! When there is a will there is a way.
I like the idea to use addressable wireless speakers, or just addressable speakers. Also you could always link laptops together playing the same program and syncronize them with a timing/trigger signal. You could have 12 laptops control the system, 1 per row, each with 2 7.1 surround sound cards.
Good luck! please let us know if you got this working I am very curious!
Thats a great post. We are building this planetarium for a green school and we want to do the same thing. It a big dome building ( 6m radius) and we want to make the sound very realistic .So 72 channels x 2 speakers per channel =144 speakers. They will be build in the walls. We have collected most of the speakers and now we are wondering about the sound cards or the MOTOs… Could you please update your results. It will be very helpful. Cables is not a problem for as, so as the fast computer :). But i couldnt find much info about the dummy speakers??? what are they and how do they work?.
We hope to get a reply from PCM :), we did try to email you, but no luck there so far. :)
Thanks for your help friends :) and have a great holidays.
Forums > MaxMSP