Share Your Measured Noise Floor

[Steve Dallas]

These measurements are interesting to me and useful for providing context.

I let REW's RTA run for one minute in two of my rooms before taking speaker measurements. A/V equipment was on at idle. The house was empty in both cases, and I turned off the heat and other noise making things. The house is two years old and is sealed with spray foam. The attic is double insulated.

This is what I measured in my office, which has a PC running in the closet:

Note the mains spike at 60Hz, but no rectifier harmonic at 120Hz. Also note the spikes at 1K, 2K, 3K, 4K, 6K, 7K, 8K, and 9K. I wonder what those are? Why not 5K and 10K?

Here is what I measured in my living room. The refrigerator may have been running in the background, as the living room and kitchen are open concept.

Here we have higher noise in the lower octaves. This could be from that refrigerator, nearby construction, wind, something else, or a mixture of these. We have a similar 60Hz mains spike, and this time, we have its 120Hz harmonic. We again have the spikes starting at 1K and are again missing them at 5K and 10K.

Here is an overlay of the two rooms:

There is amazing agreement here. Weird.

I poked around in my REW files and have apparantly never measured the noise floor in my media room / home theater. I will do that next time I have the house to myself, but that may be a while, since we are currently experiencing a blizzard.

With 34dB minimum noise in my listening environments, how important is obsessing over certain measurements?

 

[SKBubba]

That's pretty interesting. Unfortunately, in my case the ambient noise floor is lower than my tinnitus.

 

[Ron Texas]

@Steve Dallas the last time I checked, it was comparable to what you measured.

 

[Steve Dallas]

That's pretty interesting. Unfortunately, in my case the ambient noise floor is lower than my tinnitus.

I had COVID in late October. One of its lingering gifts has been tinnitus. I keep hoping it will go away, but it has not improved at all.

 

[Blumlein 88]

What FFT size?

 

[Steve Dallas]

What FFT size?

 

[Blumlein 88]

So what microphone are you using and has it been calibrated for SPL? If it is a Umik 1 for instance over USB then it can set calibration without you having to do it.

 

[Blumlein 88]

Using a Umik 1 here is my result. The fridge is running which is the source of the 60 hz hum I think. Also it is a windy day with lots of heat pumps running in the neighborhood. With fridge off and later at night it would drop below 40 db spl and the region around 3-6 khz would be down around 3-5 db SPL.

 

[AnalogSteph]

Unfortunately REW seems to have no way of easily applying perceptual weighting to results, it can only unapply it if the measurement mic has some built-in.

UMIK-1 self noise has been measured, about 30 dB(A) does in fact sound about right as that is pretty much the best you can hope for with a 1/4" electret capsule. (Note: This assumes analog input gain being maxed out, which it may not be in calibrated mode, so ADC noise may still prove the limit first.) Having a multipattern LDC around for comparison may be handy.

 

[Blumlein 88]

Unfortunately REW seems to have no way of easily applying perceptual weighting to results, it can only unapply it if the measurement mic has some built-in.

UMIK-1 self noise has been measured, about 30 dB(A) does in fact sound about right as that is pretty much the best you can hope for with a 1/4" electret capsule. (Note: This assumes analog input gain being maxed out, which it may not be in calibrated mode, so ADC noise may still prove the limit first.) Having a multipattern LDC around for comparison may be handy.

The results I showed are with UMIK gain at 18 db. I forget what max is and you can change it, but it is more than 18 db.

I have a KSM44a which is 4 dbA self noise. Maybe I need to get it calibrated and give it a try in omni mode.

 

[Steve Dallas]

So what microphone are you using and has it been calibrated for SPL? If it is a Umik 1 for instance over USB then it can set calibration without you having to do it.

It is a UMIK-1 with calibration file. It tracks within 1dB of my SPL meter.

 

[ernestcarl]

That's pretty interesting. Unfortunately, in my case the ambient noise floor is lower than my tinnitus.

Same here. Actually, my HF tinnitus rises above my already loud ambient noise machine 70Hz to maybe 14kHz. I presume many of the frequencies are “phantom” tones existing above my current actual hearing range. It can be maddening at times.

 

[Pio2001]

Here is an overlay of the two rooms:
[...]
There is amazing agreement here. Weird.

What you are measuring is the background noise of your microphone, not the background noise of your room, that is much lower.
That's why it looks the same in both measurements.

 

[Blumlein 88]

Here is a quiet measurement with a quieter microphone.

I turned off the heat, and fridge. The rising trend above 20 khz is hiss from some active speakers that were on in the room. This is my video system and the surrounds and center are actives. I was suprised how low the level is in the 3-6 khz range.

This is a microphone I cross calibrated with the Umik 1. Generally the curves are within 2 db of each other the whole way to 20 khz. I calibrated the level as well. This microphone has 17 dbA of self noise, and I only needed 25 db of gain to match the UMIK.

 

[Matthew J Poes]

It is a UMIK-1 with calibration file. It tracks within 1dB of my SPL meter.

I know this is old, but I stumbled upon it.

A few notes: First, your measurements aren't really valid useful measurements of your rooms noise. umik1 is too noisy for that. Most rooms have a noise floor that is close to or quieter than the umik1 is, so the measurement isn't useful. Like most measurements, you want about 10dB of margin between your mic and the room.

Last note, you aren't really measuring it right, I understand the desire to look at the spectrogram and see a very fine grained measurement, but the problem with this approach is that its highly non-intuitive and gives misleading results devoid of any relationship to perception. What you are depicting is the way that noise has been divided up into tiny little FFT buckets. Those buckets get ever tinnier the finer the FFT. The total sound energy is finite so the finer the bucket the less sound energy in each bucket and the lower the level. What does that tell you? My brain can't think in terms of FFT buckets.

Instead, the industry standard approach to noise floor measurement is to use 1 octave RTA because it combines the sound energy into larger buckets that better map to how we perceive that noise in the room. Some look at it in 1/3 octave to better see specific problems sources, but it isn't the standard.

And I personally had John add the industry standard weighting schemes into REW for room noise floor measurements. If you go to the RTA and set it up for 1 octave RTA measurements, it will automatically show you what the NC, PNC, NR, etc. are for the room. You can google these, they are all industry standard ways of perceptually weighting the measurement to match a theory of noise floor perception. Like anything, they are all imperfect, some more heavily weight LF's to account for the fact that, while we aren't as sensitive to LF's we tend to be more bothered by them. NR and NC are the two most common standards used for listening rooms and they match closely to A weighting.

The new umik-2 is quite a bit quieter than the umik-1 and can measure down to 20dBa which is pretty close to NC20 or NR20. When I get a chance I will do the sandbag test and place the microphone into a box surrounded by sandbags to test what its true noise floor is. In any case, for anything short of a soundproofed room, I would guess that mic would be able to measure room noise fine. I've used it successfully in my temporary apartment and in a number of other standard spaces, as the measurement I got was identical to that of my quieter reference measurement mic.

Some will be tempted to use really quiet large diaphragm recording mics. There is a problem here and I suggest against it. Most if not all of these have a non-flat response and a cardioid radiation pattern. The accurately measure noise floor you need to measure the spatial average of a large area of the room and capture the noise evenly and randomly. An omnidirectional pattern is critical to that, as is the flat response. Responses are mostly correctable, but radiation patterns are not. I've tried measuring with low noise cardioid mics and the results don't jive with standard measurement mics, they tend to be too low.

One of the ways to know if the mic is a limiting factor, besides the fact that you may get basically the same result in multiple obviously different rooms, is if it has a smiley face shape. That is because capsules have a rising HF response. The mic preamp has a rising LF response. This combines to make a smiley face. If you see that, you are at the limit of the mic (But keep in mind nearly all rooms will naturally have a rising LF noise floor anyway, its rare for the room to be flat).

You are probably going to latch onto the spatial average thing and wonder how to achieve that. The right answer is you probably can't without introducing noise and so the better approach is just to measure it in a few locations and mentally average them. You could try saving the results and doing an average. I'm not sure how accurate that would be. To do the basic temporal average part, set averaging to infinite in the settings, FFT length doesn't matter too much, 256k or 512k is fine. Longer is needed for wider bandwidth and lower noise floors, but you aren't there. Then let it average for about 1-2 minutes. I do this remotely but you can just run a long cable and leave the room. Don't measure with yourself in the room, it will pick up your breathing, key types, etc.

The way I do it is with multiple matched mics or simply doing multiple measurements. I have also used a special device that spins the mic around, but that device is expensive. I had it given to me to use on that job. They are mostly used in labs.