Originally Posted by schwaggs
I connected everything up last night and took some measurements. I ran separate outputs from the 2X4 into each sub (2400, Left, Right).
The tower speakers are along the screen wall and flank my 120" screen which is centered on the screen wall. Each speaker is about 38 inches from each side wall. I have ZERO flexibility in this setup, there are doorways between the speakers and side walls. The 2400 is against the front wall in the exact center of the wall. MLP is 2 seats centered in the room (each MLP is off center by the width of a chair, if that makes sense).
I turned on 1/48 smoothing to make the above 80 response look less confusing. I can post non-smoothed graph if preferred. It really has little to no effect on the lower parts of the graph.
Subs without EQ
Blue - Towers
Yellow - 2400
Purple - combined
One HUGE takeaways from this for me: The 32hz spike I have been fighting for over a year is caused by the sub in my tower speakers. I was running the subs in the towers through the LR speaker signal. Breaking it out as Mark suggested will allow me to FINALLY tame that beast!
It appears that the 40 - 60 hz dip is room induced (both subs have a similar dip).
All 3 subs with EQ on all 3.
Green - 2400 alone
Missing - Towers alone (forgot to measure)
Purple - Combined
I tried putting a high pass filter on the tower subs but whenever I did, it would cause a dip in the combined response. I was using the "LR 48db/octave" at 20hz setting in the crossover on the 2x4. Is it that the little bit of output the towers have below 20 makes a difference in combined response? Should I put the filter at 15 or 10hz?
What can I do about the crazyness above 60 hz? Is that helped by adjusting the delay between the subs? If so, what should I try? Would room treatments help in that area (I currently have none)? (my room is carpet over concrete, standard drywall, 10ft ceilings, 17' 5" wide, 16' 8" deep)
I have to admit, turning the tower subs on and off while listening does not give all that much of a difference. You can make out the change but it is not night and day. Turning the 2400 on and off does yield a bigger difference, mostly in frequency range, as expected. What I am gaining is more overhead (you can see the 3db gain in the area the tower subs have output) and probably (untested) more even bass response through the room which is totally worth it to me. Thanks for the suggestion Mark!
Very good job on the homework, and very good info. There is a good deal more you can do with the results, and it is very useful to know about the ~35Hz peak.
A few things to understand before you start tinkering. When the 2 curves are at the same level, the two curves can combine to up to +6dB vs the individual. This is where the big potential gain lies in the 40-60Hz range. The key will be figuring out the correct delay and high pass filter settings. Right now you are not getting maximum summation quite where you need it as you can see just above 50Hz where the curves cross, the summed response is almost the same. This means the signals are about 90 deg out from each other. If you left things as is and added 4-5ms of delay to the Polks I expect you would see a dramatic change that interaction, but the delay needed will change once you add some EQ and a high pass.
The complicating factor is that when you add a high pass filter, it also shifts the phase/delay around vs frequency. The steeper the filter, the more delay. There's no need for a 48dB/oct filter on the Polk tower subs, so I would first add at PEQ filter around 35Hz only to the Polk's. Adjust this without the high pass in place, and use a gain of -5 to -8dB to bring the peak toward the 80dB line on the graph you posted. Adjust the width/Q of the filter to not impact too much above 40Hz, but still wide enough so it still looks like a smooth mound (a notch in the middle of the peak means it's too narrow or too deep a cut). Once that peak is tamed some I would then add a 24dB/oct set around 35-40Hz, where the roll off will help further pull down the peak in the Polks. You really only care about the contribution of the Polk's above 40Hz, and this will help make sure they don't work so hard keeping up with the system. Once you have the PEQ and high pass set, now redo the overlay you posted of the 2400, the Polk's, and the combined response.
Once you have the responses to combine, I believe the multi-subwoofer optimizer would allow you to load in the measurements and play with delay and level settings in the computer if you didn't want to do it with the real responses. You can calculate the approximate amount of delay needed to shift a frequency by 1/4 or 1/2 period. If the 2 curves add to the same value, they are 1/4 wavelength, or 90 deg off. If they cause a deep cancellation, they are 1/2 wavelength or 180 deg off. You can calculate in milliseconds the value for either by taking 1/(4*F), or 1/(2*F) and multiplying by 1000. This gives you a starting point to make a big change, and then you can make +/- adjustments from that starting point.
Referencing the curves you posted above, I expect the right settings should allow you to reduce the dip in the 50Hz range such that it doesn't go below 75dB. This will allow you to reduce the boost you have on the 2400 in the 40-60Hz range, and will likely make things a little more even through the room.