Finding Subwoofer

[marcl]

There have been a couple questions about my post regarding the behavior of the LFE channel.   Here are the measurements that I did using the REW Output #4.

I tested with Equalization set to User and no filters loaded.  No equalization - neither Dirac nor PEQ.   I used only the Center Subwoofer output.  The Center Subwoofer output connects to a miniDSP which drives two subwoofers placed front and rear of the room.  No filtering or crossovers in the miniDSP; only a delay on the rear sub to align it to the front sub.

I set out to verify a few things:  Is there a difference in the LFE response with the Crossover Slope set to 12 or 24db/octave? (Yes)  Is there a difference in the LFE response with the processor Mode set to Surround or Direct? (No)    Is there a difference in the LFE response with the Center Sub set to LFE or Mono? (No)

Here's a screen shot of all the results from REW.   I had a question about the crossover frequency, since the Dolby spec says LFE goes up to 120Hz.   That is the spec, but look at the graph ... for practical purposes it rolls off at 100Hz.  Note also that the low-pass at 100Hz is happening in the processor.  I have no reason to believe REW or JRiver Media Center are applying that low-pass.




I noticed something else interesting, the consequences of which (if any) are debatable.  Take a look at the Impulse Response plots for all the test results.




Remember that measurements in Direct Mode are in blue, Surround Mode in pink.   As you can see, in Surround Mode the signal is delayed 3ms.   And for both Modes, the 24 db/octave Crossover Slope is delayed 3ms vs the 12 db/octave Slope.  The Phase plot also showed phase shift varying with frequency that corresponded to the delays.

I don't think the delay of Surround vs Direct is of any consequence since presumable all channels move together.   

[marcl]

Oct 1, 2020 8:13:45 GMT -5 ttocs said:

Am I just a Pawn in your dastardly scheme?

Actually, S2 to H does make sense. I've been trying to figure out the most efficient method of ordering the combinations, but in the end, it not worth trying.

Yeah you know my reasoning.  The Harman/Welti/Toole research says placing a sub at the opposite end of the room will cancel some resonance modes and result in smoother response.  If I recall, you did find this in one of your first tests ... less difference between the highest peak and deepest null.

[ttocs]

Oct 1, 2020 8:18:04 GMT -5 marcl said:

Oct 1, 2020 8:13:45 GMT -5 ttocs said:

Am I just a Pawn in your dastardly scheme?

Actually, S2 to H does make sense. I've been trying to figure out the most efficient method of ordering the combinations, but in the end, it not worth trying.

Yeah you know my reasoning.  The Harman/Welti/Toole research says placing a sub at the opposite end of the room will cancel some resonance modes and result in smoother response.  If I recall, you did find this in one of your first tests ... less difference between the highest peak and deepest null.

S1 was in Positions H and I, but S2 has never moved from Position A. So keeping S1 in Position D and moving S2 to P-H makes for two pairs of opposite walls - S1D & S3C, and S3C & S2H.

[marcl]

Oct 1, 2020 8:38:53 GMT -5 ttocs said:

Oct 1, 2020 8:18:04 GMT -5 marcl said:

Yeah you know my reasoning.  The Harman/Welti/Toole research says placing a sub at the opposite end of the room will cancel some resonance modes and result in smoother response.  If I recall, you did find this in one of your first tests ... less difference between the highest peak and deepest null.

S1 was in Positions H and I, but S2 has never moved from Position A. So keeping S1 in Position D and moving S2 to P-H makes for two pairs of opposite walls - S1D & S3C, and S3C & S2H.

What is the width of the front of the room?  If it's around 16ft, then position A close to center puts S2 at the half wavelength of 70Hz, which might account for that null.  Moving S2 to position H may also require flipping polarity.

[ttocs]

Oct 1, 2020 8:52:28 GMT -5 marcl said:

Oct 1, 2020 8:38:53 GMT -5 ttocs said:

S1 was in Positions H and I, but S2 has never moved from Position A. So keeping S1 in Position D and moving S2 to P-H makes for two pairs of opposite walls - S1D & S3C, and S3C & S2H.

What is the width of the font of the room?  If it's around 16ft, then position A close to center puts S2 at the half wavelength of 70Hz, which might account for that null.  Moving S2 to position H may also require flipping polarity.

The TV wall is 23'-4" wide. The window wall on the right is 21'-2".

[marcl]

Oct 1, 2020 9:05:40 GMT -5 ttocs said:

Oct 1, 2020 8:52:28 GMT -5 marcl said:

What is the width of the font of the room?  If it's around 16ft, then position A close to center puts S2 at the half wavelength of 70Hz, which might account for that null.  Moving S2 to position H may also require flipping polarity.

The TV wall is 23'-4" wide. The window wall on the right is 21'-2".

So the resonance modes related to the width will be at multiples of 23.9Hz (1116/(23.3x2))... round numbers:  24, 48 ... 72!  So it could be that third mode is showing up out of phase at the MLP?

[ttocs]

Oct 1, 2020 9:19:14 GMT -5 marcl said:

Oct 1, 2020 9:05:40 GMT -5 ttocs said:

The TV wall is 23'-4" wide. The window wall on the right is 21'-2".

So the resonance modes related to the width will be at multiples of 23.9Hz (1116/(23.3x2))... round numbers:  24, 48 ... 72!  So it could be that third mode is showing up out of phase at the MLP?

Sounds reasonable. The face of the front firing passive radiator is 18'-9" from the windows, 5'-10" (Mary Ann's height, girl I had a crush on in college) from the dining wall behind. If we need to calculate from the center of the down firing driver then the dimension changes by about 10" (6" radius plus 4" to front edge of subwoofer).

[marcl]

Oct 1, 2020 9:30:28 GMT -5 ttocs said:

Oct 1, 2020 9:19:14 GMT -5 marcl said:

So the resonance modes related to the width will be at multiples of 23.9Hz (1116/(23.3x2))... round numbers:  24, 48 ... 72!  So it could be that third mode is showing up out of phase at the MLP?

Sounds reasonable. The face of the front firing passive radiator is 18'-9" from the windows, 5'-10" (Mary Ann's height, girl I had a crush on in college) from the dining wall behind. If we need to calculate from the center of the down firing driver then the dimension changes by about 10" (6" radius plus 4" to front edge of subwoofer).

How all the modes in a room add up at the MLP is very complex of course, and especially with your vaulted ceiling.  But it's interesting to correlate the obvious ones as a possible explanation of what we measure.  And easy to test the hypothesis by moving a sub half or quarter the distance to a room boundary and see if the null moves.

After all, below the Schroeder Frequency, the reason we see peaks and nulls is because of the way the resonance modes combine at the MLP.  It's all about the geometry (and construction) of the room, the location of the subs, and the location of the MLP.  Any other factors are of little consequence.  And that's why placing low frequency absorption devices at locations in the room where they are effective in stopping reflections, is the best way to flatten out the peaks and nulls and also tame the time domain at the resonant frequencies.

[ttocs]

Oct 1, 2020 9:38:49 GMT -5 marcl said:

How all the modes in a room add up at the MLP is very complex of course, and especially with your vaulted ceiling.  But it's interesting to correlate the obvious ones as a possible explanation of what we measure.  And easy to test the hypothesis by moving a sub half or quarter the distance to a room boundary and see if the null moves.

S2 is the biggest offender of the big null @67hz when measured alone and is about 9" from the wall.
S3 has a null that bottoms out at 68Hz. When this sub moves away from the windows the response improves, but it blocks the Front Right speaker.
S1 has a narrow dip at 72Hz.

[ttocs]

010012DCH
miniDSP REW
Only delay is employed in these measurements. No PEQ, no Dirac.
S2 moved to Position H. Moving it a few inches at a time, it covered every possible increment between the fridge and the windows including pulled out even with the fridge, and facing 3 directions, I didn't try facing the passive radiator to the wall. Too tired. May try that later.

010012DCH-PLAN


S1D


S2H


S3C


S1D-S2Hinverted-S3C


edit: A little bit about the process. When a sub is in a new location, I exhaust every possible spot within the area where it would be reasonable to keep the subwoofer, and sometimes even spots where the sub would be in the way of daily life. Orientation is changed as well.

Once the "best" spot for that sub is determined, then a second sub is combined with the current subject sub. Delay is altered to find better response from the pair.

Then the third subwoofer is added to the group. Again, delay is altered and sometimes requires a back and forth between the two subs' delays to tweak the "mix". When things are far away, it can be tedious, but once the measurement curve starts looking better and better it becomes a welcome challenge to get to the best outcome.

I'm not finished with this even though this result is looking very good. There are some locations I need to try before I'll be satisfied.

[ttocs]

010012ACH
These plots will show some interesting results from just two subs, and three subs, non-inverted and inverted on S2 behind the seating area.

S1D-S2Hnoninverted


S1D-S2Hinverted. Huge difference with only inverting S2 in miniDSP.




010012-S1D-S2Hnoninverted-S3C


010012-S1D-S2Hinverted-S3C. Again, huge difference after inverting S2 when it's behind the seating area. Also note how much smoother everything is above 70Hz with S3 added compared to only S1 and S2 above, and the slight increase in the lowest frequencies.

[ttocs]

00929ACD-S1D-S2A-S3C


010012-S1D-S2Hinverted-S3C


The choice from these is obvious to me. The one with Position H included, because that's the one marcl was voting for  . Seriously, it's the better starting point pre-correction.

[marcl]

Oct 1, 2020 18:15:39 GMT -5 ttocs said:

010012ACH
These plots will show some interesting results from just two subs, and three subs, non-inverted and inverted on S2 behind the seating area.

S1D-S2Hnoninverted


S1D-S2Hinverted. Huge difference with only inverting S2 in miniDSP.




010012-S1D-S2Hnoninverted-S3C


010012-S1D-S2Hinverted-S3C. Again, huge difference after inverting S2 when it's behind the seating area. Also note how much smoother everything is above 70Hz with S3 added compared to only S1 and S2 above, and the slight increase in the lowest frequencies.

That is amazing!   But I'm not surprised 

[ttocs]

marcl I grabbed the wrong plan in today's main post which showed S2H pointing the wrong direction. It is now correct. Facing left in plan.

[marcl]

Oct 1, 2020 18:41:03 GMT -5 ttocs said:

marcl I grabbed the wrong plan in today's main post which showed S2H pointing the wrong direction. It is now correct. Facing left in plan.

I wouldn't think the direction of the sub would matter much.  They are effectively omnidirectional.  Do you see a difference with how they're pointed ... all else being equal?

[ttocs]

Oct 2, 2020 0:30:23 GMT -5 marcl said:

Oct 1, 2020 18:41:03 GMT -5 ttocs said:

marcl I grabbed the wrong plan in today's main post which showed S2H pointing the wrong direction. It is now correct. Facing left in plan.

I wouldn't think the direction of the sub would matter much.  They are effectively omnidirectional.  Do you see a difference with how they're pointed ... all else being equal?

The Sumiko subwoofers have a passive radiator on the front, and the down firing active driver is on the bottom. So there is a direction, and the measurements change based on direction.

edit: Also, when trying to put the back of the sub against the wall, it's the attached cords that prohibit it from getting closer than about 3". But when facing sideways like it is now, it is millimeters from the baseboard. 

[ttocs]

010031-KHL
miniDSP REW
Only delay is employed in these measurements. No PEQ, no Dirac, no crossover.

010031-KHL-PLAN


010031-S1K


010031-S2H


010031-S3L


010031-S1K-S2Hinverted-S3L


After moving S1 around to all the different spots in that area of the room, only Position K worked slightly better than Position H. Bass extension is a tiny bit better and is a little stronger.

The unfortunate consequence of putting a subwoofer under a dining table is consistency. When I moved S1 under the dining table in the position that measured best, there were a couple storage bins on a couple chairs, plus one chair was pushed out of the way, and a guitar was in the corner. Then yesterday, before I began taking another set of measurements I moved a few items to make room for moving S1 around to check the other positions on the plan. But before I moved S1 I wanted to take a measurement to verify that it was the same as the day before, but it wasn't. It turns out that the items I moved out of the way had an influence on the results and I wasn't able to replicate the plots from Thursday. It kinda doesn't matter because I ended up moving the sub to a better spot a little beyond the end of the table. The initial frequency response with all subs operating had better low end, but the dip just below 100Hz got much deeper and a little wider, and was now impossible to improve until S3 moved a fair amount.

Overall, I think things improved quite a bit. The big peak is about the same, but everything below and above raised up 2-3dB for the most part with less lumpiness. I guess the basic idea here is that this current setup has more energy even though I reduced the output of one of the subs because that central peak had grown a lot.

[ttocs]

Sept 30, 2020 17:27:45 GMT -5 DYohn said:

Have you tried this position?

I tried it with the latest group of measurements. It would've been nice if it worked, but it's got some problems. The biggest issue is that the low bass is lost. Here's the plot. The orange is S1 in Position K. The purple is Position E, where you placed S1 on the plan, and this is about as good as it would measure even after rotating the sub. I moved it a few inches at a time in all directions in the dining area. 

[marcl]

Oct 3, 2020 4:27:26 GMT -5 ttocs said:

010031-KHL
miniDSP REW
Only delay is employed in these measurements. No PEQ, no Dirac, no crossover.


010031-S1K-S2Hinverted-S3L

It's remarkable what you've been able to achieve with position and polarity ... and without room treatment too!

Most significant is the lack of deep nulls.  Even the narrow one above 100Hz is gone.  Dirac has so little to do you'll hardly get your money's worth 

Interesting about the effect of objects in the room.  Years ago I emptied a couple of desk drawers that were filled with paper and removed a cabinet from the room.  My transition frequency dropped about 10Hz as did the location of two nulls between 100 and 200Hz.  This is because acoustically it appeared that the room's volume increased a bit and some reflections changed.

[ttocs]

Oct 3, 2020 8:26:06 GMT -5 marcl said:

Oct 3, 2020 4:27:26 GMT -5 ttocs said:

010031-KHL
miniDSP REW
Only delay is employed in these measurements. No PEQ, no Dirac, no crossover.


010031-S1K-S2Hinverted-S3L

It's remarkable what you've been able to achieve with position and polarity ... and without room treatment too!

Most significant is the lack of deep nulls.  Even the narrow one above 100Hz is gone.  Dirac has so little to do you'll hardly get your money's worth 

Interesting about the effect of objects in the room.  Years ago I emptied a couple of desk drawers that were filled with paper and removed a cabinet from the room.  My transition frequency dropped about 10Hz as did the location of two nulls between 100 and 200Hz.  This is because acoustically it appeared that the room's volume increased a bit and some reflections changed.

It's been a very good learning experience for me. But there's something that is bothering me, the delay. I don't yet know what Dirac will do with the amount of delay I've had to employ to get these results. I hope it doesn't throw off the lip sync.

Here are the settings in miniDSP:

Notice also the Gain varies a lot. I had the subs gain matched before venturing on this odyssey, but found that using gain adjustments in miniDSP helped where nothing else would. I will take the time to lower the gain equally so nothing is above 0.0dB on the miniDSP. In a previous test I found that if the miniDSP's gain was increased higher than 0.0dB it didn't work as well as raising the sub's gain, the subwoofer's gain did a better job, but I'm not expecting that to be the case every time, so I may just have to adjust all those settings in miniDSP lower by the same amount - or raise the individual subwoofers' gain, but I won't know which is a better path until I run Dirac again because I don't want the subs to be too loud for Dirac to handle without reducing the subwoofer slider in Vol Cal.

Here is a comparison of plots from before S3 was moved from Position C to Position L, and after. Notice that pretty big dip in the 90Hz range.
010022-S1K-S2H-S3C  S3 is at Position C.


010031-S1K-S2H-S3L  S3 is at Position L.


010012-S1D-S2H-S3C  Here's the one from a couple days ago for comparison.