gedlee said:
Dr Geddes,
To this point, I have several 4'x2', 4" thick, 8# rockwool panels from a previous "bass trap" experiment in a small room. Is there a material that you could recommend that I could use cover the faces of these so that they would reflect the mid & high freq and leave the bass (<150hz) range of absorption? I guess it might be something that is refered to as a "limp mass membrane" -not really sure if that would correct.
I have seen some comments of yours regarding the placing of panels like these against the wall as being less than optimal.
-Apologies if you have covered this elsewhere, I have not found it.
Appreciate any insights you care to share.
-Tony
Sub configurations
I think the best passive configuration (as far as sseat to seat consistency) I have seen is what I'll call the "Floyd Toole" configuration, since he suggested I try it. It is 4 subs, located where the 1/4 room dimension lines would intersect. This puts the subs out in the room, so perhaps not as practical as in the corners, but it is very flat and consistent. I would like to try this with ceiling mounted subs, though dont know when i will get the chance. If it were me, I would probably double them up at the four wall midpoints (more practical than Floyd's idea).
However the four midpoints is nearly as good. The eight in all eight corners looks about the same, but probably would be better in terms of standing versus sitting ear heights (which are not simulated here. See attached. And Earl, I know you will have something to say about the regular seating grid used!
See attached
gedlee said:
I think the best passive configuration (as far as sseat to seat consistency) I have seen is what I'll call the "Floyd Toole" configuration, since he suggested I try it. It is 4 subs, located where the 1/4 room dimension lines would intersect. This puts the subs out in the room, so perhaps not as practical as in the corners, but it is very flat and consistent. I would like to try this with ceiling mounted subs, though dont know when i will get the chance. If it were me, I would probably double them up at the four wall midpoints (more practical than Floyd's idea).
However the four midpoints is nearly as good. The eight in all eight corners looks about the same, but probably would be better in terms of standing versus sitting ear heights (which are not simulated here. See attached. And Earl, I know you will have something to say about the regular seating grid used!
See attached
Attachments
Todd, we still see variations bigger than 40dB (!) which is unacceptable to me. Look what a real world DBA is able to deliver: http://www.avsforum.com/avs-vb/showthread.php?t=837744
Best, Markus
Best, Markus
DC mode
John, you got some 'splainen to do. What do you mean:
"Additionally, since the DBA does not cancel the DC mode the response has a nice linear boost as the frequency drops. "
Do you mean the "0th" mode? For that mode, I believe the room acts as a lumped parameter system, so you have 2 sources out of phase and delayed. Sounds like a comb filter, with complete cancelation at/near DC.
What am I missing...?
John, you got some 'splainen to do. What do you mean:
"Additionally, since the DBA does not cancel the DC mode the response has a nice linear boost as the frequency drops. "
Do you mean the "0th" mode? For that mode, I believe the room acts as a lumped parameter system, so you have 2 sources out of phase and delayed. Sounds like a comb filter, with complete cancelation at/near DC.
What am I missing...?
Todd
I mostly see any situation that requires specific locations as being impractical. I setup sound systems in real rooms, although I have done (many years ago) a tremendous amount of simulations in order to get an idea of what to do. When I put these simulations into practice I found that they worked very well and that obsessing about specific locations, which is almost never possible (I can't do DBA in my room or any room that I have built). Hence I no longer worry about "what is ideal?", etc. etc. Floyd's suggestion would never be acceptable to any of my clients.
What I have found is that you get "very good results" with a few simple and easily adopted rules (I won't won't go into those again), but it requires some measurements and parameter setup. Can the ideal improve upon this (in a real room)? Maybe and maybe not, it will be a long time before that is proven in the general case, if there even is a single answer. I would suggest that any setup in a real room with real walls and furniture is going to need measurements to verify it, so there is no "setup free" approach to be had.
As to the seating locations, what you show is unaffected by this. Its when you do a statistical analysis on the data. It will be biased by the symmetrical sample spacing in a symmetrical room. Statistics assume independence of the samples, but your points are not statistically independent when laid out that way in that room. If the room were not symmetrical or the sample set was not symmetrical then this would go away. But when both are symmetrical, it is an issue.
I mostly see any situation that requires specific locations as being impractical. I setup sound systems in real rooms, although I have done (many years ago) a tremendous amount of simulations in order to get an idea of what to do. When I put these simulations into practice I found that they worked very well and that obsessing about specific locations, which is almost never possible (I can't do DBA in my room or any room that I have built). Hence I no longer worry about "what is ideal?", etc. etc. Floyd's suggestion would never be acceptable to any of my clients.
What I have found is that you get "very good results" with a few simple and easily adopted rules (I won't won't go into those again), but it requires some measurements and parameter setup. Can the ideal improve upon this (in a real room)? Maybe and maybe not, it will be a long time before that is proven in the general case, if there even is a single answer. I would suggest that any setup in a real room with real walls and furniture is going to need measurements to verify it, so there is no "setup free" approach to be had.
As to the seating locations, what you show is unaffected by this. Its when you do a statistical analysis on the data. It will be biased by the symmetrical sample spacing in a symmetrical room. Statistics assume independence of the samples, but your points are not statistically independent when laid out that way in that room. If the room were not symmetrical or the sample set was not symmetrical then this would go away. But when both are symmetrical, it is an issue.
Todd, that's still 20dB deviation? Still not good enough for me. That's the frequency response at my listening position from 20Hz to 100Hz (3 subs):
Best, Markus
An externally hosted image should be here but it was not working when we last tested it.
Best, Markus
TRADERXFAN said:
Hi Tony
I would put those panels behind the speakers. They would work better there. I wouldn't try and cover them as I can't think of anything that would work very well. When block damping material like that is placed against a wall it cuts its effectiveness about in half compared with placing it away from the wall several inches. The more inches away, the lower the frequency at which it will work - has to do with wavelength away from a wall which is always a velocity node.
markus76 said:
Put even a single modest 10 dB paramteric filter on it, and the response is pretty darn flat, especially when you compare to most rooms which are considered quite acceptable and have easily 20 dB variations (but without a single big boomy one).
In any case I really dont understand this obsession with visually flat responses. Do you listen with your eyeballs? Do you think you could express a consistent preference for a response that was reasonably flat to one that was ruler flat? I doubt you could in a controlled test.
If you are interested in fidelity, the desireability of any room resposne characteristic has to take into account the conditions the music you listen to was created under. Do you think the response in the control room was anywhere near ruler flat when the music you listen to was mixed?!?
There - now I've done it!
gedlee said:
Earl, I dont think the sub locations for the "canned' configurations are super critical. especially since there will be some damping in the room, so the nodes are not so localized. Yes, of course measurements are better, and you have to factor in practicality into any situation. I think we all agree on that.
cap'n todd said:
I think that reasonably smooth is the goal, but I personally find that a slight rise at the LF end of a few dB is desirable. I think that Marcus room, with the EQ that you suggest, would be more than acceptable to virtually anyone (by listening of course!). And he did this without an extreme locating situation - a few subs, some measurements and some simple parameter adjustments. I think that his example, and there are many more, proves the point of practicality and "good enough".
How will EQing affect the other seats? Wasn't the goal in this case not to optimize a single seat but all of the seats?
Well, that raises a question I have thrown in the discussion earlier but nobody responded. How to relate physical measures to perceptual metrics? I really can't answer the question "How flat is flat enough" or "How much modal ringing is too much". Can you? Until the answer is found flatter is better and less ringing is better.
I agree that the original is what was heard in the mixing/mastering room. And no, I don't think the response in the control room was anywhere near ruler flat. But do both realizations help us in defining what the ideal sound reproduction has to look like?
Best, Markus
Re: DC mode
What you are missing is that the subs on the back wall are delayed by a time delay = L/c. Therefore there is still a net, sinusoidal variation in room volume as you approach the DC or 0th mode, although this change in volume decreases at a rate of 6dB/octave as F goes to zero. Thus what we see is a 6dB/octave increase in SPL (as F goes to zero) as opposed to zero for a dipole and 12dB for in phase monopoles.
cap'n todd said:
What you are missing is that the subs on the back wall are delayed by a time delay = L/c. Therefore there is still a net, sinusoidal variation in room volume as you approach the DC or 0th mode, although this change in volume decreases at a rate of 6dB/octave as F goes to zero. Thus what we see is a 6dB/octave increase in SPL (as F goes to zero) as opposed to zero for a dipole and 12dB for in phase monopoles.
cap'n todd said:
I am in agreement that seat to seat consistency is a goal that should be at the top of the list. In that regard I have a code that allows me to see the mode by mode influence on the SPL at an arbitrary listing position. Up to 8 sources can be considered. The code tells me if a given mode is excited by any or all the sources; if excited to the contributions for all the sources cancel; and finally, for those mode which are excited but don't cancel, how the contribution will vary form point to point. The point being, that if a mode is excited its contribution to the SPL must vary spatially.
Re: Re: DC mode
Yes, so what you mean is that the "0th" mode is not completely cancelled . Perhaps mincing words, but anyway I understand. If/since the 0th mode acts as a lumped element, you still technically have a comb filter (and 6 dB/oct loss in room gain). Above the pressure zone, it probably it doesnt matter, since the 0th mode is swamped by the other modes anyway. Ok, not cancelled, but not unaffected either.
john k... said:
Yes, so what you mean is that the "0th" mode is not completely cancelled . Perhaps mincing words, but anyway I understand. If/since the 0th mode acts as a lumped element, you still technically have a comb filter (and 6 dB/oct loss in room gain). Above the pressure zone, it probably it doesnt matter, since the 0th mode is swamped by the other modes anyway. Ok, not cancelled, but not unaffected either.
The 0th mode is not a realistic concept. It only occurs in a perfectly sealed room (like in a computer model) - this is the kind of room in which you would suffocate. There are no such room in any home or office and this mode is never seen in practice. What happens is that the room becomes a Helmholtz resonator (the listener is on the inside) with the "leaks" acting as the lumped mass, which moves it up from 0 Hz to some finite value. It will tend to be very low Q since there will be a lot of loss associated with the leakage. Thus the hypothesized +12 dB gain never occurs in reality. There "can be" some gain, we've measured it in cars, but in a home I have never seen it. The response always goes to near zero below the first mode.
I also want to add a couple of of thoughts about DBA versus SFM. In a very general sense, DBA is a speceial case of SFM. That is, if you allow polarity inversion and a longer than usual allowed delay, SFM will tend to find the DBA configuration among the top solutions. I have done this in a real room. Interestingly, SFM found several solutions that gave very similar results to dBA (sometimes slightly better since we are in a real room). As the room gets more non rectangular (shape, furniture, wall construction or whatever) SFM will find better solutions than what you get from DBA. The mian difference is that DBA will presumably optimize over a larger area, potentially the entire room. This is great, but in most rooms we listen from one or a few locations. The minimumphase thingy is another matter. One which i would like to look further into, by looking at some of our data to see what SFM does in that regard. With DBA, using (relatively) long delays, if it doesnt work perfectly, I wonder if the response might be MORE non MP? So far in SFM we have only found it useful to go up to 10 ms max signal delay, whereas in DBA you could easily double that, or more.
SFM has been shown to (more often than not) get more low bass output, whereas DBA will be less efficient. With SFM I can look at a number of top solutions and hand pick the one i want. Thereby I can do some horsetrading between flatness, consistancy and bass output.
Not that I'm biased!
SFM has been shown to (more often than not) get more low bass output, whereas DBA will be less efficient. With SFM I can look at a number of top solutions and hand pick the one i want. Thereby I can do some horsetrading between flatness, consistancy and bass output.
Not that I'm biased!