How to look for room modes in REW measurements

[No. 5]

Consider these two waterfalls. For purpose of illustration, I increased the gain so that the peak is at 130dB. The noise floor is unknown because I did not measure it.

Thank you for explaining your position, I see what you are talking about in regard to providing a reference point. I'm going to risk being called pedantic now, and propose that you did actually measure the noise floor.

Consider this waterfall:

This is a measurement sweep (sort of), level calibrated to 94dB and taken with no speakers connected, and the microphone wrapped up in a plastic bag and burred in a large bucket of sand in a quiet basement on a quiet evening. I'm sure that there is still some ambient noises being picked up (especially at low frequency), but for the most part, it looks to me to be in line with an independent lab's assessment that this mic has a broadband noise floor of 33dB. Calculated with the mic's sensitivity, that's -96.9dBu of broadband noise in every measurement made with this microphone. Which also means that I can't measure ambient noise levels that are less than 40dB with this microphone because of the constant presence of 33dB of noise. Now look at the waterfall from my second example again (I think it was taken with the same mic... I can't remember now):

This time there is ambient room noise in there in addition to the inherent noise of the mic and electronics, but the main thing is that this accumulated noise floor continues on after the signal dies away. Another thing of note: I used a longer window in this waterfall than I did in the previous post because favoring time resolution in this case will give the illusion of a lower noise floor than is actually present.

 

[No. 5]

It's relative to room length. Length to speaker orientation so always use longer axe for orientation/placement. In first example it's just above 2 m, in second about 3.7 m if it whose let's say about 5 m first peak would be about 34 Hz and could be used controled for the boost.

I'm sorry, but I'm not following how that's connected to knowing the absolute SPL of a measurement, could you explain? I mean, a virtual bass array is still cool and all...

 

[ZolaIII]

I'm sorry, but I'm not following how that's connected to knowing the absolute SPL of a measurement, could you explain? I mean, a virtual bass array is still cool and all...

Only corelation where the room mode will be is room length. It will be present on background noise and pretty much any SPL level. Of course it will get worse on very loud level and mix with other refractions so going pretty wiled. You want it clearly present but not going wiled so you do measurements on about 76 dB or - 12 from calibration point which is considered subjective as twice as loud (+12 dB).
Table is to pin them down accurate and not be able to miss them, REW room simulator show you the same thing more graphical.

 

[JohnPM]

To see the noise floor tick the "Capture noise floor" box on the Measure dialog and look at the result on the Distortion graph. REW can tell you the frequencies and decay times of resonances in a measurement.

 

[CleanSound]

To see the noise floor tick the "Capture noise floor" box on the Measure dialog and look at the result on the Distortion graph. REW can tell you the frequencies and decay times of resonances in a measurement.

Thanks John. I will play around with that. And keep up the great work!

 

[Bow_Wazoo]

I'll chime in here with my measurement.
Which EQ settings would you choose up to 200 Hz (red)?

 

[ZolaIII]

I'll chime in here with my measurement.
Which EQ settings would you choose up to 200 Hz (red)?

View attachment 376476

All peeks starring from one at 44 Hz. You can go up to at least 1 KHz pretty accurate regarding PEQ Q factor (50). You need to do and uper mids (overtones for most instruments and vocals) in 2 to 5 KHz range but for that it's better to use FIR convolution (all the way up) and finally you slope it down in highs with broad PEQ filter and add either real equal loudness compensation or self low PEQ filter Q 0.71 @ 100~105 Hz and towards SPL for loudness. That's about it, before even starting try to minimise the hole @ 65 Hz with better placement if possible.

 

[Bow_Wazoo]

I haven't gotten any further. I definitely need to change something about the highs. After half an hour it becomes stressful for my ears.

 

[sigbergaudio]

All peeks starring from one at 44 Hz. You can go up to at least 1 KHz pretty accurate regarding PEQ Q factor (50). You need to do and uper mids (overtones for most instruments and vocals) in 2 to 5 KHz range but for that it's better to use FIR convolution (all the way up) and finally you slope it down in highs with broad PEQ filter and add either real equal loudness compensation or self low PEQ filter Q 0.71 @ 100~105 Hz and towards SPL for loudness. That's about it, before even starting try to minimise the hole @ 65 Hz with better placement if possible.

Already above 150-200hz, the response will vary quite a bit with small changes in position. So personally I would not do narrow band EQ up to 1khz. If you find something consistent over several measurements in a wider listening spot, perhaps look at dampening that.

And why slope down the highs?

 

[ZolaIII]

I haven't gotten any further. I definitely need to change something about the highs. After half an hour it becomes stressful for my ears.

View attachment 376480

Change the colour scheme to light, it's preferences - view and on the bottom right. REW calculates PEQ's on it's own and doesn't do it bad (min phase) as you instruct it and for various hardware and software it supports.

@sigbergaudio it's about +5 dB on the bright side as it is @ 10 KHz and should be gradual per octave about - 1.4 dB depending from where you start. At least that's Harman.
Edit: better approach is VBA only, then VBA + FIR to 20 KHz (with VBA applied before) and afterwards doing PEQ peek up to 1 KHz and up to 200~250 Hz to the waterfall plot resonances and again to get them in line with most of other one's and not to kill them entirely. That's clarity and RT60 but a bit too much for beginners.

 

[Bow_Wazoo]

Rew calculated this for me.

I allowed a bass boost (probably a mistake, because the program tried to compensate the hole at 65hz).

I then set, additionaly, a low shelf +7db. 105hz, Q 0,7
And after your recommendation a high shelf -1.4dB

 

[ZolaIII]

Rew calculated this for me.

View attachment 376485

I allowed a bass boost (probably a mistake, because the program tried to compensate the hole at 65hz).

I then set, additionaly, a low shelf +7db. 105hz, Q 0,7

You set low self filter 105 Hz Q 0.71 (Butterwort) afterwards to the A scale SPL reading from UMIK-1 in REW. It's a bass boost to mimic ISO 226 2003 loudness compensation.
You must adjust REW EQ properties correctly. Set filter tasks to Overall Max Boost 3 (or 2), Individual Max Boost 2 (or 1), Flatness 2 (or 1). I use Target settings HF Fall Slope Start 1000 (Hz) and slope 1.4 dB/octave. I don't bust low bass additionally and use ISO 226 2003 implemented loudness in JRiver.
To reference to SPL how to adjust low self filter:

If you want to try VBA and FIR (first two parts of the video):

 

[Bow_Wazoo]

OK, thanks.
I definitely have to correct all the rew settings again. Because what rew calculated, based on my settings, just sounds wrong and "empty" in the bass range

 

[ZolaIII]

OK, thanks.
I definitely have to correct all the rew settings again. Because what rew calculated, based on my settings, just sounds wrong and "empty" in the bass range

Most partly I told you how already including color scheme that it doesn't hurt your eyes (more than it have too).

Edit: please play a bit with REW so you get to know it better and it won't look hard or new to you after a while.

 

[Keith_W]

Here is a real life example from this evening. I was asked to do some measurements at a friend's place.

The subwoofer is crossed over to the mains at 40Hz. We can observe a dip at 40Hz. Dips can be due to crossover phase cancellation, room modes, or SBIR. Only these 3 things. Question: what is the cause of the dip?

A: move the mic forward 1m and re-measure. if it shifts in frequency, it's a room mode. End of story. If it does not shift, it's either XO phase cancellation or SBIR.

The next diagnostic step is to either invert the polarity of the subwoofer, or push the speakers closer to the wall, whichever is easier. Inverting the subwoofer polarity should cause the dip to disappear. If it doesn't, then by process of exclusion, it's SBIR. Pushing the speakers closer to the wall should cause the cancellation dip to shift upwards. If it doesn't, then it's XO phase cancellation.

In this case, I shifted the mic 1m forward and re-measured. Red is baseline, brown is mic shifted forward. Notice that the dip has shifted downwards in frequency and less in amplitude. Conclusion: it's a room mode.

 

[sigbergaudio]

Change the colour scheme to light, it's preferences - view and on the bottom right. REW calculates PEQ's on it's own and doesn't do it bad (min phase) as you instruct it and for various hardware and software it supports.

@sigbergaudio it's about +5 dB on the bright side as it is @ 10 KHz and should be gradual per octave about - 1.4 dB depending from where you start. At least that's Harman.
Edit: better approach is VBA only, then VBA + FIR to 20 KHz (with VBA applied before) and afterwards doing PEQ peek up to 1 KHz and up to 200~250 Hz to the waterfall plot resonances and again to get them in line with most of other one's and not to kill them entirely. That's clarity and RT60 but a bit too much for beginners.

I think I would just rather dampen the bass much less (to 90dB instead of 85dB), but the end result is the same I guess

 

[ZolaIII]

I think I would just rather dampen the bass much less (to 90dB instead of 85dB), but the end result is the same I guess

Calibration is done to (white noise - 20 dB for true peeks in bass) 86~88 dB SPL stereo and that's the no perceived loudness correction level. That's very, very loud of course and 99% of people won't listen on such levels especially not usually. If you want to go over that then you dial down bass instead of up and only low self filter is enough to 60 dB SPL without of need to increase highs (going even under that) as that is normal speech level (still about 71 dB in bass +20 for peeks so theoretically 91 max to ISO 226 2003) common folk's will also rarely go under that.

 

[sigbergaudio]

Calibration is done to (white noise - 20 dB for true peeks in bass) 86~88 dB SPL stereo and that's the no perceived loudness correction level. That's very, very loud of course and 99% of people won't listen on such levels especially not usually. If you want to go over that then you dial down bass instead of up and only low self filter is enough to 60 dB SPL without of need to increase highs (going even under that) as that is normal speech level (still about 71 dB in bass +20 for peeks so theoretically 91 max to ISO 226 2003) common folk's will also rarely go under that.

I'm not sure we're discussing the same (perhaps we are). This what I'd do if I were to EQ the red graph (simplified). Would probably not lift the dip at 65hz. Would leave 200hz+ alone.

 

[ZolaIII]

I'm not sure we're discussing the same (perhaps we are). This what I'd do if I were to EQ the red graph (simplified). Would probably not lift the dip at 65hz. Would leave 200hz+ alone.

View attachment 376517

Yes to a blue supposed target baseline. Zoom in the scale and you will see there are +6 dB peaks after 200 Hz and continues bumpy. It remains to be seen how much of that are harmonics of lower ones so you deal with them first. Starting from first room fundamental and that's what VBA is all about (calculating and averting to first and and lowering the related following two and three deeps related to it) works on same principle as bass trap only in digital domain (FIR convolution).

 

[Bow_Wazoo]

Yesterday I tried again to correct it by hand.

Unfortunately I have almost no possibility to move the speakers...