Effect of sofa on room measurements

[ozzy9832001]

You mean you want me to leave the microphone in place, use the same filters, and perform the same sweep a few times with the only variable being sofa present or not? Yeah, I can do that.



I have read links about MMM, but it has never been clear to me what they are trying to accomplish. It would seem to me that it would average out some frequency bands more than others, namely short wavelengths, and I suspect that how much averaging depends on how far you move the microphone. I guess I could resolve all these questions by performing the experiment myself and seeing what the results are.

MMM shows how everything actually sounds. A sweep is a very short, single point measurement. A lot of peaks and valleys won't fully fill in until sound actually fills the room. We also rarely keep our head in the same position. So adjusting for that is crucial.

I've used it a lot if a particular song seems to have some weird behavior and is triggering a resonance or something that sounds odd.

This is why we use pink periodic noise.

 

[tmtomh]

MMM sounds very interesting and makes a lot of sense - thanks!

Am I correct in understanding that it's most useful for seeing what the in-room response is like for mid and high frequencies, since their shorter wavelengths produce greater variation based on small changes in microphone/head position?

 

[DWPress]

MMM seems to give me an accurate representation of the full frequency range. Pink periodic noise is full spectrum. I use it to doublecheck filters spit out of Dirac, REW and even to crudely level match subs and other drivers. Here are the settings I use in REW:

 

[ernestcarl]

MMM sounds very interesting and makes a lot of sense - thanks!

Am I correct in understanding that it's most useful for seeing what the in-room response is like for mid and high frequencies, since their shorter wavelengths produce greater variation based on small changes in microphone/head position?

MMM's usefulness is not at all limited to mid and high frequencies. However, indeed, it is very effective at minimizing comb filtering artifacts and extremely variable peaks and dips by sampling a representative "spatial average", say, a half meter to one meter rectangular volume of space.

Alternatively, it also can be used to design precise magnitude FR correction EQ in the mid-HF range that's normally highly questionable using a single-point only measurement.

Yet, just to be clear, it does not replace more detailed (ideally anechoic) off- and on-axis speaker reference curve measurements.

According to @jlo the microphone should ideally "move in all directions, even height, but avoid to be too near of a reflecting surface, ie a bit higher than the back of the chair."

*Hmmmn... for some reason I'm seeing a "viagra" link inside his MMM section description. Someone might have hacked and modified the page(?!) LOL

 

[ozzy9832001]

MMM sounds very interesting and makes a lot of sense - thanks!

Am I correct in understanding that it's most useful for seeing what the in-room response is like for mid and high frequencies, since their shorter wavelengths produce greater variation based on small changes in microphone/head position?

I would argue it's actually the opposite. You'll possibly notice, depending on your room, how all the sounds interact with one another. It wasn't until MMM that I noticed a massive spike at 215hz. It's not present during a sweep, even a long one, only during MMM either with pink noise or listening to music.

 

[dasdoing]

I have some problems with simple averaging like MMM (besides the problem of the noise not beeing removed). if we have a sharp peak that moves a little up and down in the FR depending on the position the averaging will bring this peak down. I think it makes more sense to use "peak" on RTA, though we than have aditional problems with noise (causing peaks)

 

[dasdoing]

average of the peaks

I'd like to propose an alternative method of averaging, which may be particularly suitable for averaging measurements in multi-seat scenarios. This approach is based on the assumption that peaks are prominent, while dips are obscured. Let's consider an example with two "frequency responses" from...

www.avnirvana.com

 

[Keith_W]

Well, I finally nailed it tonight.

Individual driver correction with time alignment. Full range correction (20Hz - 20kHz) with a target curve as depicted above - a straight downward tilt, 0dB at 20Hz, -4dB at 20kHz. I played with different tilts ... -2dB was too bright, -6dB had a tad too much bass, so -4kHz was the Goldilocks zone. All measurements taken with the sofa in place and a stack of cushions to simulate a listener. That step response is not perfect, but it is a lot closer to perfect than I have ever managed. Ever.

Only after I completed making all the filters that I realized I should have taken measurements with and without the simulated listener to "show and tell" on ASR. By then I was too impatient to listen to music, and OMG veils were lifted, wife commented from kitchen, etc. I honestly think I can stop here and not worry about making any more refinements, but there are probably more improvements to be made so when I get bored again i'll get back to refining. Now I am going to get back to listening to music

 

[ozzy9832001]

Well, I finally nailed it tonight.

View attachment 301852
View attachment 301853
Individual driver correction with time alignment. Full range correction (20Hz - 20kHz) with a target curve as depicted above - a straight downward tilt, 0dB at 20Hz, -4dB at 20kHz. I played with different tilts ... -2dB was too bright, -6dB had a tad too much bass, so -4kHz was the Goldilocks zone. All measurements taken with the sofa in place and a stack of cushions to simulate a listener. That step response is not perfect, but it is a lot closer to perfect than I have ever managed. Ever.

Only after I completed making all the filters that I realized I should have taken measurements with and without the simulated listener to "show and tell" on ASR. By then I was too impatient to listen to music, and OMG veils were lifted, wife commented from kitchen, etc. I honestly think I can stop here and not worry about making any more refinements, but there are probably more improvements to be made so when I get bored again i'll get back to refining. Now I am going to get back to listening to music

What smoothing is applied on the graph?

 

[Keith_W]

What smoothing is applied on the graph?

FDW of 10/10.

 

[ernestcarl]

I have some problems with simple averaging like MMM (besides the problem of the noise not beeing removed). if we have a sharp peak that moves a little up and down in the FR depending on the position the averaging will bring this peak down. I think it makes more sense to use "peak" on RTA, though we than have aditional problems with noise (causing peaks)

I don’t think using the peak RTA curve for EQ is a good idea. I sometimes record and save it as another interesting “reference” esp. in the past, but my experience is one should not use it for the final correction filter. It’s more useful as an indicator for certain room/speaker resonance and reflection-decay issues — though, often not exactly EQ correctable.

average of the peaks

I'd like to propose an alternative method of averaging, which may be particularly suitable for averaging measurements in multi-seat scenarios. This approach is based on the assumption that peaks are prominent, while dips are obscured. Let's consider an example with two "frequency responses" from...

www.avnirvana.com

Also, John M. actually makes sense in his statement that one should avoid fully flattening all peaks. For multiple seats or an entire couch, I look at things individually or overlay the curves on top of each other — particularly giving more weigh to the central listening position and lesser to the periphery. This almost certainly means irregularities (and peaks) will always continue to exist. What matters is the correction still sounds acceptably very good everywhere as we move about the room space.

 

[Deleted member 48726]

any surface close to the mic that is reflective will create comb filtering, including you head.
not sure if a cover on the sofa will be absorbive enough, but sounds like the best compromise

But what if you're ripped and only wear your trident and fustanella when critical listening?