Hi John,
I am also interested in this subject as well. As a matter of fact, I've been trying to persuade a local electronics company to participate in this effort since they do plate amps. But if something like this can be done in a small room, it opens lots of possibilities. Having worked in simulation for a few years, I know various sound queues and how important they are.
I am also interested in this subject as well. As a matter of fact, I've been trying to persuade a local electronics company to participate in this effort since they do plate amps. But if something like this can be done in a small room, it opens lots of possibilities. Having worked in simulation for a few years, I know various sound queues and how important they are.
Looking at the modal problem from a musical perspective, I think it's less of a problem if we hear "nooooote" (modal ringing but flat frequency response) instead of "note". But perceiving "note" as "NOTE" (peak in frequency response) or "NOOOOOOOOTE" (peak in frequency response and modal ringing) alters the musical content in a way it becomes audible as a defect. The same is true for dips in the frequency response.
Thinking a little bit more about Earl's approach that uses the mains as additional low frequency sources, I'm not sure if this is something that should be recommended by default. The simple reason is that we can't be sure that we will find coherent low frequency signals on every channel in each and every recording. But this is a requirement for multisub to work properly.
With Earl's Nathan I don't see much of a problem here as the woofer falls of with 12dB/octave from around 150Hz. With other speakers this is something that every AVR would be capable of by applying a high pass filter (but only to the mains as we might need the subs to run higher). Another solution could be to cut all low frequency content from each channel, sum and refeed it to each speaker and the subs. The latter might be the best solution.
Best, Markus
Thinking a little bit more about Earl's approach that uses the mains as additional low frequency sources, I'm not sure if this is something that should be recommended by default. The simple reason is that we can't be sure that we will find coherent low frequency signals on every channel in each and every recording. But this is a requirement for multisub to work properly.
With Earl's Nathan I don't see much of a problem here as the woofer falls of with 12dB/octave from around 150Hz. With other speakers this is something that every AVR would be capable of by applying a high pass filter (but only to the mains as we might need the subs to run higher). Another solution could be to cut all low frequency content from each channel, sum and refeed it to each speaker and the subs. The latter might be the best solution.
Best, Markus
Hi John,john k... said:
So, your plots indicate that the response of the sub+room is non minimum phase. I have done my own investigations and had concluded that the phase is mostly minimum phase below 80 Hz in a small room. I did this by looking at a number of room responses, removing sufficient delay to make the unwrapped phase response as flat as possible, then comparing it to it's minphased version. In most, but not all cases I saw quite small differences. It wasn't a comprehensive study by any means. Your measurement seems to indicate otherwise. I would in general caution against "eyeballing" the impuse response. I just means that how we hear it doesn't necessarily correlate to our visual impression. Toole and Olive have done probably as much subjective investigation of resonances as anyone and concluded that time domain is not the best way to judge resonances. Not to say that in your example you wouldn't hear a difference, just that it's hard to say how big it would be or whether or not it would be preferred.
Have you ever listening to a subwoofer in an anechoic chamber? It sounds kind of pitifull/odd. If you remove the room by equalizing prefectly flat and removing any excess phase, you will have to add significant energy to get the same perceived bass level, and I think it will still sound odd.
john k... said:
While we're discussing the application of DBA (all issues of dry vs. big-room bass, or practical matters aside), I've begun to reconsider imagining DBA in my room - which is not 6-side shoebox. To clarify, my room is 12.4' wide at the listening end; 22' deep; and (due to a 4' lateral jog situated behind the listening area) ~16' wide at the rear wall. Basically a very chubby "L" shape, me sitting forward in the 3-wall 'top' end of the L
Walk with me for a moment here . . . It ought to work (even if different numbers of drivers at each room end) - as long as each array is aligned to its given wall dimensions, and capable of sufficient displacement. A plane wave is a plane wave . . . It simply becomes a matter of having the amplitude, delay, and phase of the rear array adjusted correctly.
IF I'm able to maintain room symmetry within the front and sides of my listening zone (say, from my chair's 8 o'clock to 4 o'clock range), and assuming arrays properly setup at each "end", might it still be possible to pull-off the effect?
The room is a sealed volume. The sudden room width change at the jog shouldn't matter, given that the (properly setup) system = zero net pressure change in the room, right?
Just thinking out loud here . . . Help me through it, Docs!
--Mark
Todd,
I am not advocating listening in an anechoic chamber though I have head sound reproduction in a few. (At one time I was involved in active noise suppression for submarines). Low frequency reproduction and sound reproduction above the Schroeder frequency are very different. Above the Schroeder frequency we are able to discern the difference between the direct and reflected sound in a room and the reverberant field adds “space” to the recording (artificial as it may be, but seemingly necessary for the illusion). Below the Schroeder frequency all we hear is the room. I agree that the energy would have to be increased if the room were free of modal excitation at low frequency. After all, in an enclosed space the SPL is related to 1/V the rate at which the sound decays is dependent on dissipation and transmission through the wall, a simple case of conservation of energy.
As for minimum phase, that again seems to be very dependent on where you are in the room, which modes contribute what to the SPL at the listening point, and how close the listening position is to a source. The only way to be sure the response is MP is to eliminate the modal contributions.
Maybe I'm looking at this in a different way than might be considered conventional but if I just look at it from the same point of view, well why would I bother? But I listen to what everyone says and then look into it and draw my own conclusions and attempt to move forward. I don't mind if I hit a dead end. At least I know I explored all the options.
But, as I said earlier, if you allow modal excitement to enter the picture then the quest for even low frequency response is impossible because the SPL at any listening point depends of the spatial variation of the modes as the listener moves around the room. You simple can not have even low frequency response over a wide area in an acoustically small room if modes are excited.
By the way, the impulse response was a simulation. FWIW, here is the same type of simulation for a dipole with dipole axis aligned with the room axis so that axial and tangential modes normal to the dipoles axis are not excited. Again, green = what the MP response would look like, blue the in room response eq to the same amplitude as the MP response. The comment All Modes means that all modes were included in the simulations, but not necessarily excited due to positioning.
and here is the result for the same dipole position but rotated 45 degrees:
I am not advocating listening in an anechoic chamber though I have head sound reproduction in a few. (At one time I was involved in active noise suppression for submarines). Low frequency reproduction and sound reproduction above the Schroeder frequency are very different. Above the Schroeder frequency we are able to discern the difference between the direct and reflected sound in a room and the reverberant field adds “space” to the recording (artificial as it may be, but seemingly necessary for the illusion). Below the Schroeder frequency all we hear is the room. I agree that the energy would have to be increased if the room were free of modal excitation at low frequency. After all, in an enclosed space the SPL is related to 1/V the rate at which the sound decays is dependent on dissipation and transmission through the wall, a simple case of conservation of energy.
As for minimum phase, that again seems to be very dependent on where you are in the room, which modes contribute what to the SPL at the listening point, and how close the listening position is to a source. The only way to be sure the response is MP is to eliminate the modal contributions.
Maybe I'm looking at this in a different way than might be considered conventional but if I just look at it from the same point of view, well why would I bother? But I listen to what everyone says and then look into it and draw my own conclusions and attempt to move forward. I don't mind if I hit a dead end. At least I know I explored all the options.
But, as I said earlier, if you allow modal excitement to enter the picture then the quest for even low frequency response is impossible because the SPL at any listening point depends of the spatial variation of the modes as the listener moves around the room. You simple can not have even low frequency response over a wide area in an acoustically small room if modes are excited.
By the way, the impulse response was a simulation. FWIW, here is the same type of simulation for a dipole with dipole axis aligned with the room axis so that axial and tangential modes normal to the dipoles axis are not excited. Again, green = what the MP response would look like, blue the in room response eq to the same amplitude as the MP response. The comment All Modes means that all modes were included in the simulations, but not necessarily excited due to positioning.
An externally hosted image should be here but it was not working when we last tested it.
and here is the result for the same dipole position but rotated 45 degrees:
An externally hosted image should be here but it was not working when we last tested it.
Todd, I'm not kidding and I've read Floyd's book. I would love to see your data.
Best, Markus
P.S. Ethan Winer did some real world tests: http://www.realtraps.com/eq-traps.htm
Best, Markus
P.S. Ethan Winer did some real world tests: http://www.realtraps.com/eq-traps.htm
markus76 said:
Hi Markus,
For starters, page 248, fig 13.24 (not a waterfall, but the data is there). It is perhaps a bit harder to see in a waterfall, but I can find it tomorrow. So, do you think that the room has so much excess phase resposne that bringing a sharp peak (easily as narrow as 2 Hz) down in level using a properly matched filter would not reduce the ringin significantly at that frequency? Or did I misunderstand you? To say that there is still some ringing afterwards does not mean you didn't remove most of it.
And BTW, just because there is excess phase, does not mean it is audible. ..and even if so, that says nothing about wehter it is detrimental or not. Whats a little excess phase between friends?
cap'n todd said:
I did that work last May and I all that remains of it is what I used for my web page. The discussion on room modes starts here where I look at the behavior of monopoles, dipoles and cardioids in a room. In that part I show the Green's function approach that we have been discussing and verify that for a rectangular room the result is in good agreement with a FEM approach. Then in the in the next part. At the bottom of the this section I discuss the minimum phase/non-minimum phase aspects of the response. Additionally, I show that as the listening position moves closer to the source the response does become minimum phase. In the figures at the bottom of the page it may not be clear but the phase from the room simulation is in green and the thin, dark blue line is the minimum phase based on the simulated amplitude response obtained form the Hilbert-Bode transformation.
Lastly a rather novel woofer setup that I have been experimenting with recently is described. It is only intended for single position listening but it potentially make eq very simple, if required at all.
I've been disconnected for a few days.
I wanted to get back to something that John posted. Yes, I agree that Morse (more likely Ingard given that statment is new from Morses original text) does make the statement about the seperation of the free field Green's function from X(r), which is claimed as being the solution to the homogeneous wave equation.
Here is why I don't see how that can be true. The free field Greens function g(r) has a finite velocity magnitude everywhere. Thus if placed inside of any boundary it will have a finite velocity magnitude normal to that boundary (not necessarily everywhere, but necessarily somewhere). For X(r) plus g(r) to have a zero normal velocity on that boundary (the boundary condition for a rigid walled room), X(r) would have to have the inverse of g(r) on the boundary and hence X(r) would have to contain sources on the boundary. If it has sources then it is not the solution of the "homogeneous wave equation".
This is a minor point and has no bearing on the discussion, and hence not really worthy of elaboration, but I did want to explain why I do not readily accept that X(r) is the solution to the homogenrous wave eqauation. There certainly does exist some X(r) that works as described above, and it can be found from Green's theorem.
Since the rooms sound field above the Schroeder frequency is never flat or smooth (its statistics are well known), I've never argued that the modal region needs to be any smoother than that. Only that it needs to be smoother than it will be with a single source. So the fact that it can't be perfectly smooth because of modal effects is not important. It just needs to be "smooth enough".
I wanted to get back to something that John posted. Yes, I agree that Morse (more likely Ingard given that statment is new from Morses original text) does make the statement about the seperation of the free field Green's function from X(r), which is claimed as being the solution to the homogeneous wave equation.
Here is why I don't see how that can be true. The free field Greens function g(r) has a finite velocity magnitude everywhere. Thus if placed inside of any boundary it will have a finite velocity magnitude normal to that boundary (not necessarily everywhere, but necessarily somewhere). For X(r) plus g(r) to have a zero normal velocity on that boundary (the boundary condition for a rigid walled room), X(r) would have to have the inverse of g(r) on the boundary and hence X(r) would have to contain sources on the boundary. If it has sources then it is not the solution of the "homogeneous wave equation".
This is a minor point and has no bearing on the discussion, and hence not really worthy of elaboration, but I did want to explain why I do not readily accept that X(r) is the solution to the homogenrous wave eqauation. There certainly does exist some X(r) that works as described above, and it can be found from Green's theorem.
Since the rooms sound field above the Schroeder frequency is never flat or smooth (its statistics are well known), I've never argued that the modal region needs to be any smoother than that. Only that it needs to be smoother than it will be with a single source. So the fact that it can't be perfectly smooth because of modal effects is not important. It just needs to be "smooth enough".
Why is it harder to see in a waterfall? A high resolution CSD should reveal ringing clearly. Modal ringing can be seen as "ridges".
When you attenuate a modal peak then you attenuate its "ridge" accordingly. That's a good thing but doesn't necessarily mean that the decay rate of the ringing was reduced. If the system is minimum phase and Q and frequency of the EQ matches the peak then the "ridge" should be gone completely. I haven't had the chance to see a waterfall of a real listening room that shows this.
I guess that's the core problem, to relate physical measures to perceptual metrics. I'm not a physicist and I do not care if a room behaves minimum phase at low frequencies or not. The fundamental questions (to me) are more like: How much deviation in low frequency response would you allow? How long is a mode allowed to ring before it becomes audible? What decay time is best?
Since a few, many or all of you may doubt the results I posted for impulse response based on a simulated in room woofer response I though I would make a measurement of a woofer in my room and use that response as the basis. The woofer is a 12" woofer of a Martin Logan Monolith speaker, slightly modified to have a response defined by a Q = 0.5, 2nd order 20 Hz HP cascaded with a Q = 0.5, 2nd order 120 Hz LP. The first figure shows the measured amplitude response at my listening position overlayed with the target. It's not too bad. The second figure shows the raw phase response with woofer to mic delay removed, roughly 4 M worth, in green. The blue phase response is the phase after equalizing the response to the target amplitude, shown above, using minimum phase equalization. The purple phase response is the minimum phase of the target. So we can see that in this case the phase, even after removal of excess delay, is not minimum phase. The last figure is the simulated impulse response for, Red, the unequalized response, Blue, the response equalized to the target amplitude, and purple, the the impulse of the minimum phase target. The impulse was 1 msec wide.
If you don't believe the impulse responses I can post a comparison of the measure impulse for the driver and the simulation of that impulse. (Hint: they are identical).
Anyway, I think this illustrates the fact that simple smoothing the amplitude for a woofer isn't enough. While it improves the initial attack, the long time response is even worst than the unequalized response, both of which ring on forever compared to the target.
If you don't believe the impulse responses I can post a comparison of the measure impulse for the driver and the simulation of that impulse. (Hint: they are identical).
Anyway, I think this illustrates the fact that simple smoothing the amplitude for a woofer isn't enough. While it improves the initial attack, the long time response is even worst than the unequalized response, both of which ring on forever compared to the target.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
High resolution in time or frequency? No, I just meant visually, because of hte projection angle. Like if you look at someones nose from 45 degrees its harder to see how long it is compared to from the side.markus76 said:
markus76 said:
If you use anything close to a matched filter, you will reduce the slope of the resonance mag response, and thus the Q of the resonance. THis reduces the decay time, not the level (at least in the minimum phase sense, which I contend is more true than not).
gedlee said:
I guess it depends on how we define homogenious. I would suggest that the equation for X(r) is homogeneous. The solution required is that to the homogeneous boundary value problem with boundary conditions such that the velocity of of the bounding surfaces is that required by the wall admittance, zero for rigid walls.