Ambient levels and noise floor discussion

@Serious what stimulus sound was used for cross calibration between the measurement mic calibrated on the ND-9 and the LCT 550?

Did you take care to cross calibrate with microphones co-located as close as possible with distances carefully maintained from the cross calibration source?

Do not take lightly that cross calibration can easily go wrong and incorrect calibration achieved. I've done this for years in an ISO 17025 compliant laboratory.

 

@Serious Are you only looking only at the most quiet moments of the measurement? Or are you calculating an avarege of the 1 hour?

 

@purr1n You're right, that room was much louder than the other one and I shouldn't have written it like that. What would you use as the figure for the ambient level/noise floor in that case? As @Lasollor mentioned I try to look at the quietest level that is somewhat sustained, so you could argue it's 36-40dBA in that case. Not trying to average it, and I'm especially ignoring peaks. This is why I mentioned on the first page that we're most likely talking about different things. And partly also why I posted these pictures since I'd like to know what you'd say is the ambient level here from the picture. Would you simply take the average? Look at the peaks? They all seem like valid arguments to me and a single figure can be very misleading.
I think what I try to do is to take an average that ignores the peaks. The noise floor of the ambient levels.

@atomicbob Yep, mics as close as possible. Since I'm in a room I didn't use a sine. Instead I used white noise. The mic sensitivity is specified at 1kHz, but going up and down an octave or so shouldn't change much, so in the end I used my phone rather than my speakers at the other end of the room, which results in something like a bandpassed white noise with a 6dB bump betwen 550Hz and 1.3kHz and rolls off sharply beyond 300Hz and 8kHz. I tried to match the dBC levels since that's easier in REW. Phone lying a couple meters away since one of the mics is a cardioid. Using the dBA curve might be better so the mic FR doesn't play as big a role.
How would you do it? I know it's not the proper way to do it, which would probably be a 1kHz tone at a fixed distance (if we can't mount the capsule in a calibrator) in an anechoic chamber, but it seemed repeatable enough.
I think the main problem is the cardioid mic doesn't pick up as much of the reflected energy, right? Should I try to rest both mics a few centimeters above my bed to absorb as much of the rear radiation from the omni mic as possible?

@elmoe This is at 2:30AM so there really isn't anything that makes a noise outdoors.

 

You don't have heating nor AC in your house? No electronics in the room?

 

Regular old oil furnace in the basement with hot water radiators in every room. They don't really make a noise while running, but at night the heater is off anyway. No central AC and definitely not running now.

Obviously my PC was on to record it, but even when holding my ear right next to the CPU cooler I can't hear the fans spinning at 200RPM. No other moving parts for basic tasks that don't require much cooling. Otherwise a couple wall warts that make a noise, but nothing that bothers me. My lamp was on, but it seems to be below my mics noise floor. I kept the CRT off which the LCT 550 seems to be able to pick up and just left on the LCD. LCD PSU noise is below the 10dBA noise floor, but I can still hear it. My cat was in my room and I could just about hear her breathing, but my recording system couldn't.
There's a fridge downstairs that cycles on and off, but it's at 32dBA/1m, no way you can hear it here. There are also two floor clocks downstairs, but with my door closed I don't think I can hear them. Otherwise that's it, I believe. What are the things you can hear at night? I don't think sub 20dBA ambient at night is that expectional.

Actually the biggest problem would've been a bit of a storm that we have at the moment. It calmed at night and I couldn't really hear the wind noise at that point. I bet atomicbob can get a sub 10dBA noise floor in his lab aswell once he changes that buzzing CCFL and measures with a mic with lower noise.

 

Several rules of life.

Generalized:
There is no such thing as a free lunch.
Physics is a bitch.
Acoustics makes physics appear like Little Lord Fauntleroy.

Specific:
All electro-acoustic transducers are omni-directional at low frequencies.
All electro-acoustic transducers are uni-directional at high frequencies.
The difference between them all is rate of change and type of polar pattern transitioning from omni to uni as frequency increases.

Large diaphragm mics have lower self noise, true. But not without a penalty. The larger size translates to faster transition from omni to uni-directional as frequency increases. For many situations they make wonderful recording microphones. Ambient environmental measurements are not a good fit due to increased inaccuracy to integrate all ambient noise as they become uni-directional.

Microphones:
checking specs more closely and you will find several things:
Lewitt LCT 550 has an effective self noise of 4 dBA and cardioid polar pattern published for 1 KHz only.
Neumann TLM-103 has an effective self noise of 7 dBA and a well defined cardioid polar pattern
Schoeps CMC6 + MK2 has an effective self noise of 11 dBA and a well defined omni-directional pattern

For ambient measurements the Schoeps will best reflect ambient environmental noise.

How would I do it

If for record in the lab I would follow IEC 61094 and in this particular case specifically IEC 61094-8:2012.
You are not likely to find that document available to you unless you pay $200CHF.
Here is a link that will give some insight:
https://audioxpress.com/article/acoustic-methods-of-microphone-calibration

If less formal and attempting to work with what is available at hand I would do the following:

White Noise vs Pink Noise
WN is equal energy per unit Hz and very useful for electronic circuit analysis with linear scale x-axis frequency plots
PN is equal energy per unit octave, very similar to the way we hear and quite useful for sound system analysis with log scale x-axis frequency plots.

1. Calibrate a reference mic and preamp to 94 dBSPL = -16 dBFS for calibrated recording using a mic calibrator.
2. Record 10 seconds of 1 KHz 94 dBSPL from the mic calibrator through the reference mic.
3. Calibrate the Sound Level Meter software for 94 dB SPL.
4. Locate a flat frequency response, ultra-repeatable powered monitor such as Genelec 8030 on a stand in the center of a large, very quiet room. Maintain at least a 2 meter sphere of clearance in all directions. One meter can be used but with understanding that low frequency response will be increasingly impacted by surface reflections.
5. Co-locate the reference measurement mic and the mic to be calibrated 25 cm in front of the speaker, on-axis with the tweeter.
6. Using Pink Noise stimulus, set level on the speaker to 85 dBC as measured by ref mic and SLM software
7. Now adjust the unknown mic preamp until it reads 85 dBC on the SLM software
8. Record both channels simultaneously for 10S.
9. Using a 2-channel analysis system capable of producing a transfer function such as Smaart, either play the recordings or use live input from the speaker (85 dBC pink noise) and capture the frequency response. This assumes the reference mic calibration curve was already input to the analysis system.

You now have the the frequency response of the unknown microphone and can also determine the sensitivity by comparing gains between the two preamp channels, reference and unknown mics as the ref mic sensitivity should be known.

Not changing preamp gains, the unknown microphone may now be used now with the SLM software to measure ambient environment noise and produce a calibrated recording of such.

All the calibrated recordings document this calibration for independent review, such as a regulatory agency wishing to verify results produced from field calibrated mics.

 

Hmmmm.

So... Loading an app on the phone and using it wouldn't actually be good enough?

 

Isn't 85dB excessive? Also why 25cm, won't that still get you some bass boost with the cardiod mics? What about reflections between the mic body and the speakers? Is that an issue at that distance? Sounds to me like it would produce a less regular FR than measuring at 1 or 2m distance outdoors or in an anechoic chamber.
Also, don't you normally start at 1kHz for calibration? I suppose the method you described is close enough, though. Probably won't get too many errors, except probably when trying to calibrate all the way down to 10 or 5Hz.

DId you keep the same 94dB is -16dBFS level for your ambient measurements? That has to require a pretty good mic preamp and DAC for it not to add much noise.

Well, that is until you hit its noise floor, which I assume you likely have. I guess if we had infinite money to throw at the issue we could get a B&K 4179 or a GRAS 40HH. I really wonder how people measured the noise floor of Microsoft's anechoic chamber. Maybe it's just a theoretical calculation.
Can you post a recording of the ambient level in your lab? I'd be surprised if it's not just noise, too tbh. I'll upload a recording of my room at night with the LCT-550, but it's really just noise. The null of the cardioid was directed towards a corner where not much noise will reflect from, so that won't make much of a difference, either.

TLM-103 does indeed seem more regular than the newer LCT-540S at 8kHz (otherwise the two seem to have similar patterns). Why is that? Is it the enclosure or is it the capsule? I can't find any info about the membrane diameter of the K-103. Could be smaller than normal. Or maybe it's just a geometric thing with how the holes are drilled in the back.

BTW:

I calculated the mean, median, mode and standard deviation values for the LAF values of the files that I generated the graphs of. Pick your poison. What I tried to describe above is closest to the mode value and it's what I'd consider the most valuable, but you decide for yourself. I'd actually skew slightly closer towards lower values, even.

Overall not that far off from the 35dBA, tbh. Closer than I expected. Library is 4dB louder than street noise
I really just gave the 35dBA figure as a rough ballpark figure. In that sense ±3dB isn't that far off I'd say.

 

@atomicbob So according to this: http://www.montana.edu/rmaher/ee417_fl08/MIC HANDOUT.pdf
A cardioid has 4.8dB less output in a diffuse sound field and in the treble those mics transition towards hypercardioid which is 6dB less efficient for random incidence sounds. So you could resonably assume that my 11dBA could in a worst case scenario be something like 16dBA, right? But given that the null was directed towards a room corner where not much sound would reflect from, it's likely even less. And since I only got noise it's pointless anyway. I wonder why I'm getting so much more noise than specified for all my mics. Rechecking the volume calibration for the 8th time won't change that.

 

No, it isn't excessive. In fact it is a relatively standard level for speaker testing, microphone field calibration, sound reinforcement adjustment, etc. When performing a single ended measurement with pink noise the high level helps to maintain signal to noise over local environment noises such as HVAC that may not be controllable. Ever wonder why the THX Dolby specification is at 85 dB SPL?

Really big hint here. You aren't supposed to be using cardioid microphones for acoustic measurements. Again, the distance is a compromise, in this case for small rooms (ie real rooms in homes) exploiting the inverse proportional law such that S/N is high for the single ended measurement and room reflection contributions are minimized.

Not if using proper measurement microphones and aiming them on-axis as I mentioned. If you have an anechoic chamber of sufficient size or an OATS (Outside Area Test Site) with favorable weather and low noise contamination then larger distances will enhance calibration performance. Or you could use a true measurement microphone with externally polarized diaphragms and send them to Scantek for calibration like I do for my "for the record" legally compliant measurements.

Typically small diaphragm electrets are pretty flat in frequency response below 1 KHz. But the differential technique I describe can work down very low and has been demonstrated at AES meetings with a number of interested measurement experts in attendance.

Another really big hint. Look at the equivalent input noise specifications for a True Systems P-Solo mic preamp. Note that is is -132 dB EIN. Now look at the noise floor for an ADI-2 Pro ADC and note we have a dynamic range of 119 dB unweighted. Limiting factor will be the microphone self noise. And unless we are measuring an anechoic chamber or a recording studio with a wonderfully engineered HVAC, either the Schoeps CMC6 + MK2 at 11 dBA or ACO Pacific 7046 on a Josephson C617 body at 12 dBA will be below the room noise level.

Bad assumption. And I live in the PNW so I have colleagues with anechoic chambers. I have been in several Microsoft anechoic chambers (they have way more than one), four anechoic chambers at Boeing Aerospace, NWAA Labs (Ron Sauro) and have access to a number of recording studios in the neighborhood such as Opus4, Robert Lang, StudioX (which has moved), etc.

 

I know, but I find them a useful tool for testing specific things and they help eliminate the room more than typical omnis. I often do things in an unconventional way though and I'm not a fan of using an omni mic in a room, too. HRTF past 2kHz isn't omni at all, in a way it's closer to cardioid. I've read about other people using cardioids for in-room speaker measurements, too. Averaging both past 2kHz could be a good idea sometimes.

I got the LCT-550 for a pretty good price. Still not entirely sure if I'm going to keep it, but it's allowed me to take distortion measurements that I can't get with any mic in that price bracket (like the ones here). I might keep it for that reason alone.

How did they get that spec? I thought the theoretical maximum would be -130dB for the self noise of the test resistor at room temperature.
According to specs the UR22C and 2i2 should make no difference in self noise with the LCT550, but they seem to. Gonna recheck in a bit when it's quiet enough again.

Well, then where does the noise come from? As a person with a lower self noise than your Schoeps mics, what was it? Your buzzing lights? I find it hard to believe that I can get my bedroom to lower ambient noise levels than you can get your lab. Were you inside the room to arrive at the 16dBA figure?

 

Honestly, I'm not sure if anyone cares, but I said I'd upload a recording of my room at night. Well, here it is:

I put the recording in a dropbox-folder, since it's too big to be uploaded directly: https://www.dropbox.com/sh/ir38sp4xdwrwkz3/AACD8iLMvL-ACnV_boqCPP5na?dl=0

At the 7 second mark I kinda bumped my door from the outside. I kept the parts where I opened and closed the door in the recording. After listening more closely I don't think I reached my system's SNR limit yet and instead the noise is just noise from vehicles. So 10dBA is probably the best-case ambient noise level for my room.

Marv and atomicbob probably think I'm incompetent and that my volume calibration is off by 20dB or so. Well, all I can say is I doubt it is. Maybe at some point it turns out that it is.

 

Ok, here are recordings from my acoustic lab. Please note these are NIST traceable calibrated recordings and follow IEC standards such as 61672. I will attach the cal level recording which is 94 dBSPL = -16 dBFS for anyone who wishes to independently verify using your favorite analysis method, though I strongly recommend vslm found here:
https://sourceforge.net/projects/vslm/

Here is my acoustic lab residual noise analysis using VSLM calibrated with 94dBSPL = -16dBFS


Here is a comparison of @Serious residual room recording and my residual room recording:

@Serious room residual recording in Red, @atomicbob lab residual recording in Green.

Attached are the calibration (94 dBSPL = -16 dBFS) and room residual ambient recording flac files.

 

Thank you, that's actually very intersting! Mine were calibrated to 80dB* being 0dBFS. Not enough SNR otherwise with my equipment. I raised your recording by 30dB and yours is louder indeed, but the SNR of your gear is insane. You seem to get similar amounts of noise as I do, despite the mic being rated for higher noise. Is that buzzing the lamp? Sounds like an old fridge from here.
Here's a 1/48th smoothed FFT with what should be the correct offset.

*Maybe not exactly 80dB, but between 79/81 if my meter/UMIK calibration is correct.

I am assuming that you recorded yours with the Schoeps CMC 62. Either way I think it'd be fair to drop the treble on my FFT according to my mics FR. But then again it's a cardioid so maybe it's just not comparable at all.

EDIT: I'm an idiot. Raising the gain on the mic preamp also raises its noise, to the point where the unloaded input had a noise equivalent to 7 or 8dBA (with this mic's sensitivity). No wonder I was limited by noise! I also seem to be getting some electrical interference that I'll see if I can fix. Or maybe the mic is picking up my mainboard's coil whine (which I can hear, but it's very quiet). But I already managed to get the noise floor to 7dBA now, during the daytime. The oil heating is running, too. At this point I think I might actually need a concrete enclosure to measure if my mic's up to spec .

 

As expected setting the gain to the highest setting does give the best SNR, unlike what I just wrote. I get around 4.8dBA for just the 2i2 and 6.6dBA for the UR22C with my mic sensitivity. A higher sensitivity mic would get less noise, ofc. I attached two 10 second files of the noise calibrated so that 110dB is 0dBFS since apparently that's what you're supposed to do. Since this was recorded with the Focusrite Scarlett 2i2 Gen 3 gain set to the highest setting 0dBFS was actualy 79.5dB and I simply digitally lowered it by 30.5dB. I advise you to either digitally raise it by 30-60dB or use an amp with a shitload of gain so you can really hear it. To me it sounds like towards high frequencies the mic preamp noise is higher than the mic noise. Adding uncorrelated sources of noise of 4.8dBA (pre) and 3.0dbA (mic) gives noise of 7.0dBA, so at least I can assume that my mic self noise is as low as advertised. The lowest noise I got so far was around 7.5dBA or so.

I wonder how much of the noise in atomicbob's recording is from the preamp noise, but with that buzzing it doesn't really matter. At least in my case my "bad assumption" of 128dB EIN and 111dBA SNR limiting the noise floor was right, even when setting the gain so that 80dB is 0dBFS. My mic has a specified sensitivity of -29dBV while the Schoeps is specified at -36dBV, so that doesn't make the job any easier for the gear.

I added longer versions of the files to the dropbox folder I linked above.
I'll let go now. I think me and atomicbob have demonstrated that you can achieve very low ambient levels without an anechoic chamber. And others have mentioned that even 35dBA (which I still think is a fairly high figure) can be optimistic under certain circumstances (although exactly how, I'm not sure). In the end it is what it is.

Now that I'm more sure of my volume level calibration I think I'll repeat the threshold tests. I think that people partaking in these studies don't really do it in their best interest to do very well and as such I think it's entirely possible that the less than -10dB I got from ca 500Hz to 5kHz are accurate.

 

I think the issue troubling Serious is that we tend to wish a room remain below a specific sound level and spectrum when no noises are being made inside while excluding those from outside. Serious is finding that at certain times of the night (or possibly day depending on geographical location) it IS possible to achieve a very low SPL and spectrum when:

1. All internal noise sources are silenced
2. No external noises occur during the time of reading
3. humans are removed from the room or remain absolutely still and breath hold

This is cherry picking the noise data. My room typically remains below 25 dBA except when there are flyovers or a noisy vehicle near the premise. But as demonstrated above I could cherry pick a lower reading down to 16 or 17 dBA. Now change to dBC and incorporate the low frequency contamination, often from vehicle noise at a distance and the SPL will rise considerably by between 10 and 20 dB. Real world with all the constraints removed are as @purr1n originally stated and multiple supporting studies may be found on the topic, especially by cities performing urban or suburban planning.

What really irritates me is use of a single number, once again, without the context of all the other competing goals to intimate lack of a particular, inappropriate tool for the job microphone (LCT 550), was limiting my (or anyone else's) ability to measure a room residual, never mind that I've been doing this for decades and either own or have access to the entire gamut of ambient measurement tools. Getting hung up on a specific number is a fool's errand as has been proven over and over and over and over and .... geez it gets tiresome. I could obtain a B&K 4179 or an ACO 7020 (and maybe I will), however residential site survey measurements are adequately served by ACO 7046 or B&K 4165 mic capsules. Except for those cherry picked moments in a rural setting devoid of external noise, MOST residential spaces are 30+ dBA, and 40+ dBC.

Homework assignment for those curious: site survey ambient sound level measurements

 

Already did that homework assignment last year for my acoustics class. We were doing measurements at a dog kennel, as the play room was loud (had a reverb time of about 1.6 seconds).

Background noise measurements taken at 10 different locations with a B&K 2250 throughout the room yielded a value of 66dbZ, meeting the NC-50 curve. However, these results don't really mean anything, because it was a dog kennel, so they just took the dogs outside, and was still pretty loud with barking on the inside.

Background noise measurements are hard to do, as its hard to isolate the external noises sometimes. Good thing is we didn't really care about background noise for this project, focused on reverberation time.

 

Ah, thanks - improved my cheap-SPL-meter mic calibration, I was low (edit: high).

Computer/measurement space, with mic (Dayton IMM-6) sitting in a flat-plate coupler in its usual measurement location:



This is a bedroom in a townhouse - one and a half floors up from the basement and backing onto a lane. The peak is Mrsdegraded slamming the gate to the courtyard (dammit, hoped I'd finish before she came back).

The location is also almost directly above the building's electrical room. Even A-weighted, Note 60 Hz and multiples thereof, particularly 240:





Looking forward to the calibrator to do a further test and check Dayton's cal file for this mic.

 

@Serious You’re gone now but I'm going to post what I was going to post anyway. You're full of shit about this and none of this bullshit matters. Thousands of people will buy noisy live recordings, rehearsals, and demos of bands they like. Yes they can tell what's going on. Hardcore punk and black metal were real things that happened so you're just whining. Black Flag's Damaged and Immortal's Battles in the North were legendary. The bands made great records and became very, very popular.

The recordings are good enough. The productions fit the mood of the music and you can tell what's going on if you pay attention. The distortion and noise is louder than the music on half of the albums. The records were recorded and mixed in good studios by experienced engineers sometimes with gear much noisier than a pair of JBL 305s in rooms with drunk people making a ruckus in the background. The technical people might not be known names to the average music fan but were professionals. Hissing monitors, noisy power amp and computer fans, beeswarming guitar amps, and single coil pickups are real deal stuff noisier than any domestic goods short of a furnace or microwave that are used by real people to make real records for real money in better rooms than your bedroom everyday.