David Gatti said:
That's what I usually recommend. The LF from mains plus a pair of subs yields four sounds sources below the Schroeder frequency. If it's not enough you can always add another sub or two.
I used to recommend Welti's placements exclusively, four corners or four wall midpoints. After talking with Geddes in 2005 or so, I usually point to his placement method as an option too.
Real rooms tend to be layed out where one or the other can be used. I've seen some rooms where Welti's placement fits perfectly. Others don't have those locations open, but Geddes arrangement works. Sometimes, it's best to kind of use both as a guide and try other options.
I also suggest the room modeling program CARA, making a model of the room and placing subs in it, then looking at the results to see if the enegry appears to group nicely in the intended listening positions, usually in the middle of the room.
There were a lot of discussions related to this a few years back, and I think they provide a lot of good information. Scan through them if you have time.
We discussed at length different placement options, symmetrical placements such as Welti suggests, pseudo-random such as Geddes suggests, other placements and the useage of CARA for modeling. The author of CARA chimed in to decribe the algorithms it uses internally. Lots of good information there as well as some historical perspective.
Now that John and Todd are talking about computer simulations, would you guys care to comment on CARA? Geddes and Thomanek discussed it in the thread below, but I'd like to know what you guys think about it too.
Do you think CARA is a useful tool for exploring room modes or not?
cap'n todd said:
That was my problem with the pseudo-random placement suggestion too. I modeled a bunch of rooms with CARA and found some that worked OK but others that had weird shaped areas of energy with nulls in odd positions. Those placements didn't look useful to me because any normal seating arrangement would put some listeners in bad spots.
At that time, Earl had reservations about CARA but I think Thomanek addressed Geddes' concerns. Also, Earl has since then suggested a method that combines measurements with general placement instruction. That may help ensure uniformity in the listening area.
gedlee said:
POPPYCOCK! When you take the variance of the spatial mean (as you essentially are) that is blind to mean spatial variance. When you take the mean of the spatial variance (as I am), that is blind to mean spectral variance. Though it is true that you are "slicing" the same data in different ways, the result is not the same. If you don't believe me I can easily show you a simple example that will force you to accept that fact. Also, I hope you are not confused by the fact that I do calculate the variance of the spatial average (VSA in my papers). That WOULD be essentially equivalent to what you are looking at, but that is not the primary metric that the optimization is based on.
Todd
We are not talking about the same things and pressures from here prevent me from making my point any clearer. Perhaps I will get the time to spend on this, but I am not sure. We don;t do things the same way, but I do claim that we will end up with the same results if you limit your parameters to the simpler set used in my approach.
We are not talking about the same things and pressures from here prevent me from making my point any clearer. Perhaps I will get the time to spend on this, but I am not sure. We don;t do things the same way, but I do claim that we will end up with the same results if you limit your parameters to the simpler set used in my approach.
cap'n todd said:
It could only be as good as the model, that's for sure. And I can sure see how hard it is to accurately model things like drywall on studs, bookshelves, chairs, television sets, people, cats and dogs.
I'm wondering how accurate it is for studying generic rooms. I'm also wondering how good it is for studying odd-shaped rooms without using more accurate damping models, like what the walls and furniture does.
Did you see the discussion between Earl Geddes and Ulrich Thomanek about the algorithms used in CARA? Earl was at first concerned thet the program used ray tracing, which would naturally only work above the Schroeder frequency. Ulrich said it used modal analysis (a series of eigenvalue functions) and he gave some examples of comparisons between modeled and measured rooms.
Link again here for convenience:
By the way, the thread was prompted by a private discussion between Earl and I where he compared your method with his. I used CARA to compare the two. It looked liek a competent piece of software, and from Earl's final comments in the thread, it looks like he eventually agreed. But like you've said, I think it would be good to compare CARA results from your software and from John's, and of course, ultimately to measurements of rooms modeled.
cap'n todd said:
I've mostly stayed out of this thread. For one thing, I consider myself a casual student of audio where you and Earl are more formal in your studies. For another thing, I respect you both too much to interject much in a way that might get sideways. And that's the rub, in the past, some of the honest disagreements seemed to get a little out of hand. By the way, I now see that John is in the mix too, along with some other guys that have given this a lot of thought.
Anyway, now that I've qualified myself on where I stand, I'll go ahead and speak my mind.
Way back when, when Earl first told me about his random proposal, my first reaction was that would be a great idea. But almost immediately, my thoughts turned, and I began to think of Geddes' proposed arrangement as "rolling the dice". Some random configurations would probably do exactly what Geddes intended, Others might not.
Here's an example discussion on this very topic:
- We've discussed this before, Symmetrical verses random placement
The pockets of uniformity weren't lumped around the center of the room or any other useful position, just broken and scattered jumbles. Sometimes, the pocket, while oddly shaped, was large enough to be useful but often times not.
Was it statistically smoother throughout the room? Maybe. Was it smoother throughout the listening area? It depended what arrangements were used. Some appeared to be better than others.
Now then, if CARA isn't accurate for this kind of study then many my models didn't tell me anything useful. But to be honest, I think the study did tell me something. I think the energy distribution is at least approximated by those models, and that comparisons could be made. Is it as accurate as the models you and John have made? I don't know. Is it close enough to be useful for setting up a room, for finding configurations to avoid and others that might serve as a starting point, I hope so.
I think that by putting measurements into the method, as proposed by Earl and Markus, the Geddes method is probably better than just blindly "going random". Is it as good as a (passive or SFM) Welti configuration, a DBA setup or a placement determined with the help of CARA or some FEM/BEM program? I don't know. I'm sure it's like a day at the races, every man has his day. Some rooms will respond better using one method, other rooms will work better with another. But one thing I'm sure of, and that I think we would all probably agree, is that's multiple subs are better than a single sub where uniformity of indoor bass response is concerned.
BTWWayne Parham said:
For a rectangular room with evenly distributed absorption (and not too much of it), the image source method is equivalent to the modal calculation we are using. I wasn't too sure about this until I found a derivation that proves it. you can find it in this paper (and probably other places):
J. B. Allen and D. A. Berkley. Image method for efficiently simulating small-room acoustics. J. Acoust. Soc. Am., 65:943-950, 1979.
cap'n todd,
I have a question for you about the paper "Subwoofers: Optimum Number and Locations." In the FR plots, what is the reference level (0 dB)? I can't seem to figure it out based on what I read in the paper. Is it the pressure that would result if the sources were in half-space at a distance of 1m (what many simulations spit out)?
I have a question for you about the paper "Subwoofers: Optimum Number and Locations." In the FR plots, what is the reference level (0 dB)? I can't seem to figure it out based on what I read in the paper. Is it the pressure that would result if the sources were in half-space at a distance of 1m (what many simulations spit out)?