Effect of sofa on room measurements

[Keith_W]

Yesterday I decided to look at the effect of having the sofa at the listening position on measurements. This is where the sofa is located:

And without sofa:

These are the measurements. I applied some smoothing to make the measurements more legible:

The mic was not moved from the position between the two measurements, nor were any settings changed. Both measurements were taken before any room correction was applied and before I had volume matched all the drivers (hence the lumpy and uneven looking response). We can see:

- About 2dB more bass with the sofa in position
- The null at 150Hz is almost completely removed by the sofa, but the null at 220Hz is 10dB deeper (both are low Q nulls though)
- There is comb filtering caused by the sofa between 400Hz - 3kHz almost certainly due to reflections from the sofa (as you can see, it is a leather sofa. I suspect that a fabric covered sofa would not be as reflective).

The usual recommendation is to remove any obstacles between the speakers and the mic when doing room correction. However, I have a somewhat silly question - given that the sofa will always be in that position, shouldn't it be part of the measurement?

 

[DWPress]

You mean sofa's - I see two of them and they are both going to affect the signal somehow. Were they still in the room somewhere when you did the 2nd measurement?

Given that we also like to live in the spaces we listen to music in I agree that if you're gonna listen on a couch in that position - keep it in the measurement. As you say though, if the potential comb filtering bothers things in the final corrections or you feel the need for more absorption then maybe consider getting something fabric covered or maybe just a slipcover.

Otherwise, great looking space! Post the final room correction measurements when you have them.

 

[Chrispy]

I think this is a room only an audio nut would enjoy Definitely man cave stuff

 

[Keith_W]

You mean sofa's - I see two of them and they are both going to affect the signal somehow. Were they still in the room somewhere when you did the 2nd measurement?

Given that we also like to live in the spaces we listen to music in I agree that if you're gonna listen on a couch in that position - keep it in the measurement. As you say though, if the potential comb filtering bothers things in the final corrections or you feel the need for more absorption then maybe consider getting something fabric covered or maybe just a slipcover.

Otherwise, great looking space! Post the final room correction measurements when you have them.

Yeah, there are two sofa's. The usual recommendation is to remove any obstacles between the speakers and the microphone. I took that to mean that I should also remove my listening sofa. The second sofa (on the left) is not an obstacle so I left it in position. The sofa that I removed was pushed into the adjacent dining room, so it should not affect the measurement too much.

It got a bit too late last night to run verification sweeps, but I did do a quick and dirty correction and it sounded absolutely awful. I have a bit more diagnosing to do today to find out why. I too am really curious to see what my rushed hamfisted correction did to the FR to make it sound so bad.

 

[Deleted member 48726]

Yeah, there are two sofa's. The usual recommendation is to remove any obstacles between the speakers and the microphone. I took that to mean that I should also remove my listening sofa. The second sofa (on the left) is not an obstacle so I left it in position. The sofa that I removed was pushed into the adjacent dining room, so it should not affect the measurement too much.

It got a bit too late last night to run verification sweeps, but I did do a quick and dirty correction and it sounded absolutely awful. I have a bit more diagnosing to do today to find out why. I too am really curious to see what my rushed hamfisted correction did to the FR to make it sound so bad.

Well. You want to measure the room as it usual is. Furniture included of course. That means placing the mic at the listening position where there hopefully is not anything in the way between the mic and the speakers.

How would you otherwise get representative measurements..

N.B. I've had the exact same model of sofa. Quality leather. You sit very well and lie down great with the nice soft rounded arm rests. My wife thought they weren't blending in though. -Out they went.

 

[Kvalsvoll]

The usual recommendation is to remove any obstacles between the speakers and the mic when doing room correction. However, I have a somewhat silly question - given that the sofa will always be in that position, shouldn't it be part of the measurement?

Certainly no silly question. This is not possible to get "right" no matter how you do it. If you remove the sofa, floor reflection will be different, if you sit in the sofa, that also changes the frequency response.

Typically, the range from around 400hz - 2khz will be severely affected by reflections from boundaries close to the mic, such as seating, floor. And indeed, that is what you see here. And if you use this measurement to hammer the frequency response, the result will not be correct, it will be flat only for the mic in that position, with all boundaries close to the mic exactly as it was when measured.

If you want a more presentable frequency response graph, try moving the mic slightly up or down, and closer or farther away from speakers. A soft pillow on the sofa can remove some reflection at higher frequencies.

Solution? Have a speaker with decent on and off axis frequency response, and do only wide-bandwidth correction, above say 200hz, if at all necessary.

 

[Philbo King]

Some sofas do creditable service as bass traps. A leather one would also reflect higher frequencies. Test it with the sofa in place. Pile some throw pillows on your listening position to simulate your body (maybe not strictly accurate, but far better than nothing).
I have 3 sofas in my studio primarily as bass traps. Properly positioned they can do a world of good.

 

[ernestcarl]

Frequency dependent windowing used doesn't help much in this case. And just one more reason why I do not like over-relying on single point measurements only.

 

[ozzy9832001]

Yesterday I decided to look at the effect of having the sofa at the listening position on measurements. This is where the sofa is located:



And without sofa:



These are the measurements. I applied some smoothing to make the measurements more legible:



The mic was not moved from the position between the two measurements, nor were any settings changed. Both measurements were taken before any room correction was applied and before I had volume matched all the drivers (hence the lumpy and uneven looking response). We can see:

- About 2dB more bass with the sofa in position
- The null at 150Hz is almost completely removed by the sofa, but the null at 220Hz is 10dB deeper (both are low Q nulls though)
- There is comb filtering caused by the sofa between 400Hz - 3kHz almost certainly due to reflections from the sofa (as you can see, it is a leather sofa. I suspect that a fabric covered sofa would not be as reflective).

The usual recommendation is to remove any obstacles between the speakers and the mic when doing room correction. However, I have a somewhat silly question - given that the sofa will always be in that position, shouldn't it be part of the measurement?

Looking at the pics, there are probably 100 other things that are causing odd reflections and asymmetry's around the room. The sofa is probably the least concerning. If anything, it probably adds some minor lower end absorption.

I added a sofa to my home office for that very reason. I didn't want panels or any other nonsense in front of the windows, so I felt a sofa was a decent compromise. It does help.

 

[JeremyFife]

In general this will depend on why you are making measurements. If you are measuring in order to EQ your room then, yes, you should include the sofas because they'll be there when you listen to the corrected signal.
Of course that also implies that you should be sitting on the sofa too, and you should include a representative average number of friends ...

If you are making more stand-alone measurements (impact of changing a piece of equipment?) then it probably doesn't matter as long as you compare like for like.

Love the room

 

[Zek]

It would be interesting to measure with 1-2-3 listeners on the sofa.

 

[Keith_W]

Otherwise, great looking space! Post the final room correction measurements when you have them.

That's left + right speaker with the target curve superimposed on it. This is an actual verification measurement and not a simulation. I windowed tightly up to 200Hz, after this very loose windowing to avoid correcting the comb filtering from sofa reflections. The target curve has a +3dB bass shelf to 100Hz, then flat to 1kHz, then -4dB at 20kHz.

I don't like the way it sounds. It is too bright for me. I will try adjusting the target curve and i'll try again.

(edit) and BTW I think I can hear some pre-ringing. I have uploaded a video into Youtube:

Before each drum transient, listen carefully for a brief hum (it lasts maybe 1/4 of a second). It is loud enough that I can hear it on this Youtube video. At home, it is even clearer and it precedes every drum note.

This is probably because the step response looks like a dog's breakfast. This is from the same verification measurement as the above FR:

Now I know I am losing street cred by posting this. I have gotten good results before, I promise! With this round of filter iterations I made four experimental filters, this was one of them. I think I know where I screwed up, but I don't think I have seen anybody post a video on ASR before demonstrating what pre-ringing sounds like so I thought I would post it so that all of you can laugh at my expense! The other 3 experimental filters do not have this problem.

 

[Fahzz]

Nice room. Maybe strike the sails on that clipper too? LOL

 

[DWPress]

Now I know I am losing street cred by posting this.

Lol, no worries. The frustration and excitement of setting up speakers with active filters is known to many of us.

Have you tried measuring with MMM method to see how things smooth outside of your listening position? I find it to be a brilliant reality check sometimes when I get too far into the weeds. Not sure about the pre-ringing, not something I've had to deal with in my system before.

 

[dasdoing]

i have been thinking a lot of this problem. the midrange where a sofa will be causing comb filtering should be ignored. but the effect below could be merged to a measurement without the sofa.
on the other hand, one might argue that the seating effect is natural. if you listen to the real event on the same sofa there is no EQ compensating for it.

 

[tmtomh]

View attachment 301269

That's left + right speaker with the target curve superimposed on it. This is an actual verification measurement and not a simulation. I windowed tightly up to 200Hz, after this very loose windowing to avoid correcting the comb filtering from sofa reflections. The target curve has a +3dB bass shelf to 100Hz, then flat to 1kHz, then -4dB at 20kHz.

I don't like the way it sounds. It is too bright for me. I will try adjusting the target curve and i'll try again.

(edit) and BTW I think I can hear some pre-ringing. I have uploaded a video into Youtube:

Before each drum transient, listen carefully for a brief hum (it lasts maybe 1/4 of a second). It is loud enough that I can hear it on this Youtube video. At home, it is even clearer and it precedes every drum note.

This is probably because the step response looks like a dog's breakfast. This is from the same verification measurement as the above FR:

View attachment 301270

Now I know I am losing street cred by posting this. I have gotten good results before, I promise! With this round of filter iterations I made four experimental filters, this was one of them. I think I know where I screwed up, but I don't think I have seen anybody post a video on ASR before demonstrating what pre-ringing sounds like so I thought I would post it so that all of you can laugh at my expense! The other 3 experimental filters do not have this problem.

Thanks for sharing this and the details of your process and experiments/iterations - very interesting, and IMHO very valuable to the community here!

I'm fairly new at this kind of thing - lots of study and trial and error, but only started using room correction a couple of months ago. So apologies if this is a dumb/irrelevant observation or if I'm stupidly misreading your graphs.

When I look at the top graph, with your pre- and post-correction response, plus your target curve, I am struck by one thing: your target curve appears to be situated mostly above your room response. In other words, it looks like your filters are mostly trying to fill dips, whereas not many peaks are being knocked down.

I mention this because my understanding is that (a) it's much harder (sometimes impossible) to boost nulls than to knock down peaks, and (b) trying to boost dips with high-Q (narrow) filters can (as far as I know) potentially cause pre-ringing.

Again, I'm basically a beginner when it comes to this, and if I'm being honest I don't even know exactly how room-correction software decides where to put the target curve relative to the amplitude of the measured response. I'm just going by what I see in that graph and my admittedly limited understanding of this stuff.

 

[dasdoing]

correcting FDW's is a little tricky. if correcting magnitude only the minimum phase version of the meassurements should be calculated before aplying the window. else, frequencies that are delayed fall out of the window.
I showcased this problem here: https://www.hometheatershack.com/th...dependent-windowing.99673/page-6#post-1619293

I think the FDW windowing shouldn't start at zero, but count only when the wave begins at the respective frequencies.

 

[ozzy9832001]

Thanks for sharing this and the details of your process and experiments/iterations - very interesting, and IMHO very valuable to the community here!

I'm fairly new at this kind of thing - lots of study and trial and error, but only started using room correction a couple of months ago. So apologies if this is a dumb question or if I'm stupidly misreading your graphs.

When I look at the top graph, with your pre- and post-correction response, plus your target curve, I am struck by one thing: your target curve appears to be situated mostly above your room response. In other words, it looks like your filters are mostly trying to fill dips, whereas not many peaks are being knocked down.

I mention this because my understanding is that (a) it's much harder (sometimes impossible) to boost nulls than to knock down peaks, and (b) my understanding is that trying to boost dips with high-Q (narrow) filters can potentially cause pre-ringing.

Again, I'm basically a beginner when it comes to this, and if I'm being honest I don't even know exactly how room-correction software decides where to put the target curve relative to the amplitude of the measured response. I'm just going by what I see in that graph and my admittedly limited understanding of this stuff.

A true null is infinite and typically very sharp. Sound is directly cancelled because it's the exact same energy. Most of what's shown on his graphs are probably from reflections. They are too wide. You can boost out of the reflections at the expense of headroom. Depending on the frequency and the width of the dip, it might be worth it, for smaller nulls, small boosts to lower the difference are a compromise.

 

[Deleted member 48726]

View attachment 301269

Are you aware of the kind of EQ boosts you're applying?? Over 15 dB some places!

Any problems or issues you encounter you should absolutely start by lowering the target so there are no places being boosted more than 3 dB and then do a reassessment of the situation.

 

[Keith_W]

Are you aware of the kind of EQ boosts you're applying?? Over 15 dB some places!

Any problems or issues you encounter you should absolutely start by lowering the target so there are no places being boosted more than 3 dB and then do a reassessment of the situation.

View attachment 301312

I think you are misinterpreting that graph. The two graphs are left channel and right channel, not before and after. Acourate does not boost nulls, it only cuts peaks. But thank you for taking the time to think about my post, I appreciate it