How to measure Rooms with High Ambient Noise

[FrantzM]

Hi

A bit on my system.

L,C,R speakers : JBL LSR 308
Surround Speakers : JBL LSR 305
Subwoofers: 2 Parts Express 15 inch Powered Subwoofers, 1 Parts Express 10 inch Powered Subwoofer
Denon AVEX -X3400H used as a Pre/Pro

Rather satisfied but Bass could be better. I have used Audyssey to clean it up but remain unsatisfied ,especially when I hear he bass from my headphones and other system. In some areas in the listening room bass is great .. at the listening position bass could be better.. both in HT and music modes.

Main issues is high level of ambient noise. I would like to try to wake up during the night when the level of amboen niose according to my SPL meter is 40~50 dB but cannot do that yet... During the day the lowest noist level is 60 dB with an average of 65 dB...

What can I do, aside from waking up and try to get my bass while I am falling asleep..? Any ideas ? Doomed and have to wait for a better listening environment... Fugedaboutit? Get back to my headphones and be done... I'd like a solution
IOW can I trick REW to measure and to reject the (high) ambient noise?
I am trying to use all 3 woofers and would like to set up profiles

Profile 1: HT with 5 channels + Subs with windows closed in a room that is 4 X 6 x 3 (meters)
Profile 2: HT with 5 channels + Subs with windows open (complicated room opens up to the outside windows are along 3 walls and room volume changes a bit ... a lot
Profile 3: 2-Channel Music Plus 3 subs Windows closed
Profile 4: 2-Channel Music Plus 3 subs Windows open

Why Windows open or closed? Open No Air Conditioner; Closed, AC is on, more comfort ..

I have REW, miniDSP 2 x 4 and UMik...

Help from the collective will be much appreciated.

 

[KSTR]

Use the longest sweep time available in REW (and set the lowest sample rate possible), and the highest possible level you and/or your neighbors can tolerate (but not so much that the subs run out of steam, of course). This gives the best noise rejection... and if it's still not enough, average a set of measurements.

 

[mitchco]

Hi @FrantzM can you post some measurements or your REW .mdat? I believe UMIK1's calibration file also calibrates the level for 75 dB SPL in REW: https://www.minidsp.com/applications/acoustic-measurements/umik-1-setup-with-rew As @KSTR mentions, increasing the output level will help with the SNR, but folks have reported that beyond 83 dB SPL clips REW's mic input... Anyway, some frequency responses measured at the LP would help us understand your bass issues... PS. Measure however you are using the room for critical listening (i.e. doors closed or whatever).

As a side note, listening at 75 to 85 db SPL is where most recordings are mixed and mastered for final tonal balance, as that level are ears are the most flat, relatively speaking. Listening at lower SPL's, our ears bass response falls off quickly, hence loudness controls. I use one in JRiver so that as the volume goes down (below ~80 dB SPL = critical listening volume) the bass level goes up. Are you using computer audio or AVR?

 

[Soniclife]

increasing the output level will help with the SNR, but folks have reported that beyond 83 dB SPL clips REW's mic input

I'm sure @RayDunzl has posted compression tests much louder than that, so I'm confused if there is truth to this.

 

[Soniclife]

I'm interested in how you might separate the possible error from your speakers, from the impact of that background noise on listening to a theoretical perfect system. In most cars I've had when you listen to music whilst driving the bass seems weak, but pull over and turn the engine off and suddenly there is the missing bass.
Have you measured the noise spectrum? Is it mainly in the low frequencies that it's loud.

 

[dc655321]

and if it's still not enough, average a set of measurements.

Wouldn't that decrease noise as only sqrt(N), for N measurements? i.e. it's a fairly weak means for a practical number of measurements?

Obviously rife with assumptions/problems, including the fact that REW does not include a means to invoke it, but would a "background subtraction" technique work for acoustic measurements? Analogous to taking an image with the lens cap on (background) and subtracting that image from a desired exposure. I could imagine building up a series of background-noise-only spectra over some period, making measurements of the SUT, taking another set of "backgrounds" for comparison, then subtracting the noise spectra from the SUT measurements.

 

[SIY]

Wouldn't that decrease noise as only sqrt(N), for N measurements? i.e. it's a fairly weak means for a practical number of measurements?

That's still pretty significant. I'll typically do 32 sweeps, which only takes a minute or two total, and that gives about a 15dB noise reduction.

 

[FrantzM]

Hi @FrantzM can you post some measurements or your REW .mdat? I believe UMIK1's calibration file also calibrates the level for 75 dB SPL in REW: https://www.minidsp.com/applications/acoustic-measurements/umik-1-setup-with-rew As @KSTR mentions, increasing the output level will help with the SNR, but folks have reported that beyond 83 dB SPL clips REW's mic input... Anyway, some frequency responses measured at the LP would help us understand your bass issues... PS. Measure however you are using the room for critical listening (i.e. doors closed or whatever).

As a side note, listening at 75 to 85 db SPL is where most recordings are mixed and mastered for final tonal balance, as that level are ears are the most flat, relatively speaking. Listening at lower SPL's, our ears bass response falls off quickly, hence loudness controls. I use one in JRiver so that as the volume goes down (below ~80 dB SPL = critical listening volume) the bass level goes up. Are you using computer audio or AVR?

Use the longest sweep time available in REW (and set the lowest sample rate possible), and the highest possible level you and/or your neighbors can tolerate (but not so much that the subs run out of steam, of course). This gives the best noise rejection... and if it's still not enough, average a set of measurements.

Will do.

 

[dc655321]

That's still pretty significant. I'll typically do 32 sweeps, which only takes a minute or two total, and that gives about a 15dB noise reduction.

Fair enough.
But, to knock down the noise by 30dB, we're talking 1000 measurements. A 60-65dB noise floor in the OP is f'ugly...
I had something in mind like trying to do a set of polar measurements and not stamping my feet in impatience, contributing to the noise floor!

 

[SIY]

Well, worst case, 1000 measurements is about 20 minutes. Yeah, you're right, that's annoying, but doable. But getting things to (effectively) 50-55dB, which only takes a minute, will get you pretty decent EQ results if your test SPL is at 80-85dB (not quiet, but not stupid-loud).

 

[Hipper]

This is for a two channel system so I don't know if it will be of any use for higher numbers.

I have high ambient noise from traffic. I've done something to try and reduce it and I've measured it. Knowing where on the frequency spectrum the noise is allowed me to try and deal with it.

To reduce ambient noise I installed a third line of glazing specifically for sound (with a large air gap) to go behind the existing double glazing.

Green - no glazing. Red - glazing. It's knocked off 5-10db off the range 100Hz-1.2kHz but did little below that. Worthwhile though and noticeably quieter. This noise only measurement was made by using the measuring system on REW with the volume on my DAC turned off so no signal was emitted from the speakers.

Using this measuring method I had already determined that the worst of the traffic noise was around 42Hz. You can see this by looking at a Spectogram of the noise measurement in REW. You will see that the noise starts well before the REW signal. Here are spectrograms of ambient noise on a weekday and at 2200hrs.

Knowing that, I used Realtraps free download of bass frequencies and my ears to determine myself what the volume of the frequencies around 40Hz should be:

https://realtraps.com/test-cd.htm

 

[dc655321]

Well, worst case, 1000 measurements is about 20 minutes. Yeah, you're right, that's annoying, but doable. But getting things to (effectively) 50-55dB, which only takes a minute, will get you pretty decent EQ results if your test SPL is at 80-85dB (not quiet, but not stupid-loud).

You're right, targeting a -15dB noise reduction via averaging is a reasonable compromise.

Any thoughts on the (in-)validity of the "background subtraction" means proposed above?
Or should I take the absence of your comment on that as a hint ?

 

[RayDunzl]

folks have reported that beyond 83 dB SPL clips REW's mic input...

On the Preferences -> soundcard is an input volume setting.

Here, with my somewhat quiet room. Registering about -40dBfs peaks, with the input volume setting maxed. The SPL Meter (unweighted) registers 55.5Lzeq (long unweighted average).

Turning down the input volume (gain) increases the headroom:

The SPL (and RTA and other calculated levels) are automatically adjusted on the next measurement:

I use the 0.531 setting. If that doesn't yield enough headroom, it can be decreased even more, possibly all the way to 0 (which does not infer "no input").

On the Microphone settings in Win10, a setting of 0 = -31dB for the UMIK-1

The 0.531 setting in REW says +0.2dB in Windows.

The 1.000 setting in REW = +24dB in Windows.

 

[mitchco]

Thanks @RayDunzl for pointing that out. I don't use a Umik-1, but an acquaintance ran into this and I did not about the control input volume in the soundcard prefs. I use a separate mic pre and ADC for my measures plus ASIO, so that UI control does not show up in my sound card prefs...

 

[SIY]

You're right, targeting a -15dB noise reduction via averaging is a reasonable compromise.

Any thoughts on the (in-)validity of the "background subtraction" means proposed above?
Or should I take the absence of your comment on that as a hint ?

I think that it will depend on if the noise is systematic/periodic or random. Systematic, mmmmmaybe. Random, you're fighting the Second Law, and in that fight, the Second Law always wins.

Of course, the log aspect is actually helpful here- if the noise is 15 dB down for the signal and uncorrelated to it, it adds as power (sqrt of sum of squares) so doesn't perturb things much.

 

[RayDunzl]

Thanks @RayDunzl for pointing that out. I don't use a Umik-1, but an acquaintance ran into this and I did not about the control input volume in the soundcard prefs. I use a separate mic pre and ADC for my measures plus ASIO, so that UI control does not show up in my sound card prefs...

When you perform a manual calibration in REW, you're told the headroom available over the calibration signal level at the time of the manual calibration, if I remember correctly.

 

[RayDunzl]

Main issues is high level of ambient noise. I would like to try to wake up during the night when the level of amboen niose according to my SPL meter is 40~50 dB but cannot do that yet... During the day the lowest noist level is 60 dB with an average of 65 dB...

I would think that the moderately high noise levels would not seriously affect the Frequency Response measurement, as long as the speakers are outputting more than the noise.



Adding two sources - a tone at 80 and random noise at 60dB, should yield a measured value only 80.043dB total.

http://www.sengpielaudio.com/calculator-leveladding.htm



A problem comes if measuring distortion:

80dB tone and 60dB noise = -20db = 10% distortion, even if there really is none...

http://www.sengpielaudio.com/calculator-thd.htm


Here's an old set of sweeps, you can see the noise intrude at the lower left, and not substantially affect the higher levels:

And the related distortion measures. Low SPL sweeps show highest distortion, because of noise intrusion, louder and louder reduces that problem, until at the very highest tone levels, where the measurement begins to rise, as the actual speaker distortions rise out of the noise floor. I didn't go loud enough to really show the rise at the bottom, but you can see some bunching of the traces as the turn begins at the loudest levels (bottom of the chart)

 

[digitalfrost]

That's still pretty significant. I'll typically do 32 sweeps, which only takes a minute or two total, and that gives about a 15dB noise reduction.

For those of us that require an acoustic timing reference, this is not an option, is it? If you ignore amplifier latency, and you only measure stereo, you could use a soundcard loopback for 1 channel, but once you go multi-channel, either semi-active or 5.1 that goes out the window.

 

[SIY]

For those of us that require an acoustic timing reference, this is not an option, is it?

I don't know about REW, but with every piece of software I have for speaker measurement, there's no issues with setting the number of repetitions.

 

[digitalfrost]

I don't know about REW, but with every piece of software I have for speaker measurement, there's no issues with setting the number of repetitions.

It's greyed out in REW, when you select acoustic timing reference.