How to look for room modes in REW measurements

[CleanSound]

So something is really off.

I shut the amp off and ran a sweep, I should get nothing, but look what I got!

What is going on? The room is silent, either there is something wrong with my sound interface or there is some background noise that I am not hearing, but I know I can hear frequency down to this range.

What in the world is going on?

 

[ZolaIII]

So something is really off.

I shut the amp off and ran a sweep, I should get nothing, but look what I got!

View attachment 358277
View attachment 358278

What is going on? The room is silent, either there is something wrong with my sound interface or there is some background noise that I am not hearing, but I know I can hear frequency down to this range.

What in the world is going on?

No such a thing as silent room, judging by that room length is about 3.5 m, take a look at my previous post and use table.

 

[CleanSound]

So something is really off.

I shut the amp off and ran a sweep, I should get nothing, but look what I got!

View attachment 358277
View attachment 358278

What is going on? The room is silent, either there is something wrong with my sound interface or there is some background noise that I am not hearing, but I know I can hear frequency down to this range.

What in the world is going on?

And here is a loop back where I lower the output. Not sure if this is a valid test to test the sound interface input to see if there is any issues.

 

[CleanSound]

No such a thing as silent room, judging by that room length is about 3.5 m, take a look at my previous post and use table.

But that freq at 53hz is 73dB! Yet, I'm not hearing it?

 

[No. 5]

So something is really off.

I shut the amp off and ran a sweep, I should get nothing, but look what I got!

Well, you will get something. The mic capsule and preamp have self noise, as does everything else up the capture chain in addition to ambient background noise. However, your time scale is set to two full seconds, and whatever it is looses almost no energy over that time (it probably disappears because of the IR window), so it's not random noise and unlikely to be something of architectural acoustic nature. Assuming the measurements are level calibrated, I would certainly hope that you can hear something ringing out for almost two seconds at over 70dB!

What in the world is going on?

Good question! I think it's something electrical, but isolating potential sources is what I would do to try and track it down. Swapping cables, trying different inputs, measuring in a different room or just making a short recording with your measurement gear and see if the noise appears on playback. That might be a quicker way to track it down.

 

[er|κzvio1in]

So something is really off.

I shut the amp off and ran a sweep, I should get nothing, but look what I got!

View attachment 358277
View attachment 358278

What is going on? The room is silent, either there is something wrong with my sound interface or there is some background noise that I am not hearing, but I know I can hear frequency down to this range.

What in the world is going on?

Maybe this wouldn't have happened with the cheap UMIk-1 instead of the expensive microphone with separate audio interface, only one way to find out. My gut feeling is that this is an electrical issue: a fridge, a fan of any electronic device that's messing with your input.

 

[dasdoing]

try another cable for the mic. also try another input on your interface.

 

[CleanSound]

Good question! I think it's something electrical, but isolating potential sources is what I would do to try and track it down. Swapping cables, trying different inputs, measuring in a different room or just making a short recording with your measurement gear and see if the noise appears on playback. That might be a quicker way to track it down.

Maybe this wouldn't have happened with the cheap UMIk-1 instead of the expensive microphone with separate audio interface, only one way to find out. My gut feeling is that this is an electrical issue: a fridge, a fan of any electronic device that's messing with your input.

try another cable for the mic. also try another input on your interface.

I figured it out. It was I-D-One-Zero-T error. . .the user did not calibrated SPL correctly and I am way too embarrassed to say what exactly I did wrong. This shall be swept under the rug from this post onwards.

Please comment on anything room mode related.

SPL, Left, Right and Both

Waterfall, Left

Waterfall, Right

Waterfall, Both

Spectrogram, Left

Spectrogram, Right

Spectrogram, both:

 

[Keith_W]

Lol, I knew that that 50Hz spike looked awfully suspicious! Don't worry about 1D107 errors, we have all made them. It's part of the learning process

Just a tip - when studying room modes, set REW to display 20Hz - 200Hz and stretch the y-axis. This makes any anomaly super obvious. And another - please don't post unlabelled graphs. It makes it difficult to tell you what we see. In your example below (left, right, and both), I don't know which is which.

In any case, what I see are two huge broad shelves one at 20Hz - 100Hz, and another from 250Hz - 650Hz.

These shelves are not caused by room modes. More likely improper subwoofer setting (for the 20-100Hz shelf) and speaker design (for the 250Hz-650Hz shelf). It is possible that an unfortunate congregation of room suck outs may produce this appearance.

You should also take note of your room's Schroder frequency. You can calculate it with this formula:

You can find the T30 by going to REW's "RT60" tab and looking at the average T30. Note that this is an approximation. The transition zone starts at Fs and ends at 4Fs. This entire region is where sound obeys wave physics, above 4Fs it behaves more like a beam. Knowing where your Fs is will help tell you which part of the graph to concentrate on.

This might be a minority opinion on ASR, but I think that below 80Hz, stereo measurements of left and right channels individually are less meaningful than both channels playing together. This is because we are less sensitive to direction of low frequencies. There is a tendency to get hung up about a dip appearing at (say) 40Hz in the left channel, but if it disappears when both speakers play together ... IMO there is no problem. This is my opinion only! I can not cite any sources to back this up because I haven't seen anybody mention it, so feel free to ignore.

Another point about waterfall graphs and spectrograms. These can easily become meaningless if the measurement was not taken properly, i.e. calibrated with an SPL meter. For example, if your measurement was taken at 60dB while REW thinks it is 80dB, this will increase the room's noise floor and make your result seem worse than it actually is. Choosing how you display your waterfall can also make it look better or worse than it actually is. Unless I am doing the measurement myself and I know I can trust my measurement, I disregard any waterfalls/spectrograms posted by anybody else. Obviously, there are a few variables at play here which is unknown to us at ASR - the skill of the operator, whether your SPL meter is calibrated or not, or with the correct settings, or even if an SPL meter was used. If I don't trust your measurement, then it is only useful for looking at relative decay at different frequencies. I trust your measured frequency response enough to comment on it, but I am careful about what I read into your waterfalls/spectrograms. I hope that makes sense.

 

[CleanSound]

Another point about waterfall graphs and spectrograms. These can easily become meaningless if the measurement was not taken properly, i.e. calibrated with an SPL meter.

How good of a SPL meter is recommended? Super precise and expensive SPL meter? Is a $30 SPL meter from Amazon will do?

I don't see any documents on the quality of a SPL meter needed in the REW documentation anywhere.

 

[Keith_W]

An SPL meter doesn't need to be fancy. This is what I use. It is reasonably accurate, I did compare it to a pro SPL meter and it seemed to give the same reading. I don't care if it's off by 1 or 2 dB, because playing pink noise makes it fluctuate by +/- 2-3dB anyway.

An even cheaper way is to download an SPL meter app for your phone, and calibrate it with a borrowed SPL meter. Set it to dBC, play pink noise at various volumes, and compare the reading. The app may/may not let you reset the reading, and your phone's mic may not be linear - so you need to compensate for it. I don't think it's worthwhile spending a lot of money on an SPL meter, you only use it briefly and put it away.

 

[No. 5]

How good of a SPL meter is recommended? Super precise and expensive SPL meter? Is a $30 SPL meter from Amazon will do?

I don't see any documents on the quality of a SPL meter needed in the REW documentation anywhere.

If you are doing a measurement where true SPL is relevant (and it's not for many), you need to decide how "true" you want it to be. If you want it to be within 2dB and around $30, I'd say find yourself a used Quest acoustic calibrator or a used Extech, Sper Scientific, or Quest class 2 sound level meter. For under $100, Amazon has an acoustic calibrator. If you want to be true to a fraction of a dB and stability over a thousand years, expect to spend many times that on a pistonphone. And I think it's important to emphasize what @Keith_W said: you're only going to use it briefly. And knowing the true SPL is often not relevant, so plan your purchases accordingly.

I have used an SPL meter in the past for setting levels in REW and it works fine, but I currently use an acoustic calibrator as it is a more accurate method and it's a lot easier to do too. A decent new one can be had for around $100, used ones can be found cheaper, and really good ones obviously are more. As mentioned, Quest made a small handful of pretty good ones that show up cheap on eBay pretty frequently. This one for example, but note that the one in the link is for a 1" mic, so you'll need to buy or make an adapter... or accept that the SPL reading will be a bit off. Which is probably not that big of a deal. You're not in a laboratory, after all.

 

[Rednaxela]

If you are doing a measurement where true SPL is relevant (and it's not for many), […]

OP is doing a measurement to find their room modes. How relevant is true SPL in that case?

(Sincere question.)

 

[Keith_W]

OP is doing a measurement to find their room modes. How relevant is true SPL in that case?

(Sincere question.)

True SPL allows you to tell REW how far the measurement is above the noise floor. This is relevant for generating waterfalls/spectrograms but irrelevant for anything else. You then use the waterfalls/spectrograms to examine decay at various frequencies. Even then, it is only useful for looking at absolute decay (i.e. I can compare his system with mine), but not relative decay (e.g. comparing the decay between his left and right speakers).

This means I can't look at his waterfall and tell him that his 50Hz peak takes 500ms to decay to the noise floor, if I don't know what the noise floor is for example. However, I CAN see that his 50Hz peak does not decay at all. This is what I meant when I said "interpret with caution". I can't look at a waterfall/spectro and tell you if it is good or bad if it has not been calibrated with an SPL meter. But I can tell you that certain frequencies are better/worse within the same measurement set.

In any case, I think that looking at the raw FR plot without the waterfall is sufficient to demonstrate room modes in most cases. I don't think routine use of an SPL meter is necessary when taking measurements. Although I own an SPL meter, I am lazy. I just play the sweep loud enough so that it is well above the noise floor, but not so loud that the speakers distort or that my ears hurt. This corresponds to about 80-85dB. The only time I pull out the SPL meter is when REW nags me into doing it or if I really want to look at the spectro/waterfall. In fact, the only reason I own an SPL meter is because I didn't understand very much at the time and REW kept nagging me.

So to answer your question: IMO not very relevant. I don't want to make the OP think he needs to go out and buy an SPL meter. All I am saying is: interpret spectro/waterfall graphs with caution if it's not calibrated with an SPL meter.

 

[No. 5]

OP is doing a measurement to find their room modes. How relevant is true SPL in that case?

(Sincere question.)

That would be an example where knowing the true SPL isn't necessary. Consider that an offending mode will present itself in two ways: as narrow band of frequencies that are higher than the "average" level, or as prolonged ringing around a narrow band of frequencies. To identify a mode in the first instance all you need to see is what is louder than everything else, and for that you only need to know the relative level rather than the "true" level. In the second instance, you need good temporal detail (to see what has a long tail), but there again, it is relative SPL is what reveals the mode. What is higher in level than the surrounding at any given time index.

Let's look at a couple examples.
The most offensive bass problems are around 83Hz and 120Hz and they present themselves as being fairly narrow band and higher in level than the "average" around them. 265Hz is an issue too, but perhaps not quite what we think of as a room mode, per se...

Here's another one where the main issue is at 47Hz. It's clearly above the rest of the spectrum, and I can say that reducing its level definitely improves the sound.

Lets look at the waterfall of the first example, but remember that there's an additional component: FFT analysis trades time resolution for frequency resolution. Here is with high frequency resolution.

And here with high time resolution.

As you can see, identifying the frequency of the "tails" becomes harder. But somewhere around 70Hz, there appears to be some ringing that wasn't apparent in the measurement with high frequency resolution. Nevertheless, it's still the feature that's higher in level than the surrounding that shows the ringing.

 

[No. 5]

True SPL allows you to tell REW how far the measurement is above the noise floor.

I think I understand what you are saying, but looking at it another way, isn't the difference between the system/room noise floor and the measurement level always relative? I mean, if you know the measurement level is truly 75dB and it decays into noise at 35dB, how is that different from seeing a measurement of an unknown level decay into noise after dropping by 40dB?

 

[CleanSound]

An SPL meter doesn't need to be fancy. This is what I use. It is reasonably accurate, I did compare it to a pro SPL meter and it seemed to give the same reading. I don't care if it's off by 1 or 2 dB, because playing pink noise makes it fluctuate by +/- 2-3dB anyway.

An even cheaper way is to download an SPL meter app for your phone, and calibrate it with a borrowed SPL meter. Set it to dBC, play pink noise at various volumes, and compare the reading. The app may/may not let you reset the reading, and your phone's mic may not be linear - so you need to compensate for it. I don't think it's worthwhile spending a lot of money on an SPL meter, you only use it briefly and put it away.

If you are doing a measurement where true SPL is relevant (and it's not for many), you need to decide how "true" you want it to be. If you want it to be within 2dB and around $30, I'd say find yourself a used Quest acoustic calibrator or a used Extech, Sper Scientific, or Quest class 2 sound level meter. For under $100, Amazon has an acoustic calibrator. If you want to be true to a fraction of a dB and stability over a thousand years, expect to spend many times that on a pistonphone. And I think it's important to emphasize what @Keith_W said: you're only going to use it briefly. And knowing the true SPL is often not relevant, so plan your purchases accordingly.

I have used an SPL meter in the past for setting levels in REW and it works fine, but I currently use an acoustic calibrator as it is a more accurate method and it's a lot easier to do too. A decent new one can be had for around $100, used ones can be found cheaper, and really good ones obviously are more. As mentioned, Quest made a small handful of pretty good ones that show up cheap on eBay pretty frequently. This one for example, but note that the one in the link is for a 1" mic, so you'll need to buy or make an adapter... or accept that the SPL reading will be a bit off. Which is probably not that big of a deal. You're not in a laboratory, after all.

I just ordered a class 2 with calibration certificate SPL meter.

 

[Rednaxela]

Thanks @Keith_W and @No. 5, most enlightening.

 

[Keith_W]

I think I understand what you are saying, but looking at it another way, isn't the difference between the system/room noise floor and the measurement level always relative? I mean, if you know the measurement level is truly 75dB and it decays into noise at 35dB, how is that different from seeing a measurement of an unknown level decay into noise after dropping by 40dB?

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.

It looks absolutely horrible. You don't know where the noise floor is, so you don't know what has been caused by the speakers or background noise.

Now we take the same measurement, and for the purpose of this discussion, we assume the noise floor is at 80dB. Yes I know it seems very loud, but I am simulating a situation where REW has no idea what volume it is measuring at here.

It looks much more reasonable.

That's all that knowing the noise floor does. It tells you where to set the cut-off.

 

[ZolaIII]

That would be an example where knowing the true SPL isn't necessary. Consider that an offending mode will present itself in two ways: as narrow band of frequencies that are higher than the "average" level, or as prolonged ringing around a narrow band of frequencies. To identify a mode in the first instance all you need to see is what is louder than everything else, and for that you only need to know the relative level rather than the "true" level. In the second instance, you need good temporal detail (to see what has a long tail), but there again, it is relative SPL is what reveals the mode. What is higher in level than the surrounding at any given time index.

Let's look at a couple examples.
The most offensive bass problems are around 83Hz and 120Hz and they present themselves as being fairly narrow band and higher in level than the "average" around them. 265Hz is an issue too, but perhaps not quite what we think of as a room mode, per se...
View attachment 358610
Here's another one where the main issue is at 47Hz. It's clearly above the rest of the spectrum, and I can say that reducing its level definitely improves the sound.
View attachment 358611
Lets look at the waterfall of the first example, but remember that there's an additional component: FFT analysis trades time resolution for frequency resolution. Here is with high frequency resolution.
View attachment 358613
And here with high time resolution.
View attachment 358614
As you can see, identifying the frequency of the "tails" becomes harder. But somewhere around 70Hz, there appears to be some ringing that wasn't apparent in the measurement with high frequency resolution. Nevertheless, it's still the feature that's higher in level than the surrounding that shows the ringing

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.
Take a look at table I posted earlier.
I forgot to mention how to use it. You fill the green column. First length and it tells you where to expect them so that it's hard to replace them for something other. Then from measurements where they are and so on.
The guide and credits go to Obsessive Compulsive Audiophile Youtube chenel and hire is the guide for VBA, impulse response to target and phase correction for FIR filter creation.

It's a bit messy and you will probably need some time to digest it but very helpful when you do. Regarding VBA it's relatively successful, makes them lass pronounced but FIR is generally not very efficient for those on a good side it doesn't alter the impulse response. If you want to use only one or two filters you do those and export them to FIR (wave which you import as file to convolver), use table only for room mode peek identification and averaged first peak calculation and then PEQ it out by hand. Have fun and take your time.