Ambient levels and noise floor discussion

I'll do a more extensive write up after the Thanksgiving festivities.

Here is a calibrated measurement of a very quiet space, my acoustic lab.

Calibration 1 KHz 94 dB - Bruel & Kjaer 4231 mic calibrator - Schoeps CMC62 omni mic with self noise of 11 dBA:


Calibration of gain staging to Smaart 94 dBSPL = -16 dBFS


Lab residual noise with one CFL buzzing slightly (I need to fix that - ladder time, it's up high)


Better than NC15, exceptionally quiet at 16 dBA.

One can hear footsteps of a cat stealthily padding across carpet.
Clothes slightly rustling is positively loud compared to the quiet.
Soft voices are perceived as loud in here.

I'll put up calibrated recordings and instructions on how to use them later so everyone can hear this for themselves.
Thanksgiving first.

 

Well, my cats aren't very sneaky so hearing my cat padding across the carpet isn't anything special. CFL lights can vary a lot, but the ones at work can drive me nuts. Then again I sit directly underneath one that buzzes.

Another (I admit possibly useless) datapoint...
Here's a recording of a 10 second snippet in my room. I can hear my CRT buzzing. The GPU fan ramped up once becacuse I had the wrong fan profile applied. My TV is on and buzzes. My dad downstairs is listening to music at a level of approximately 65dBA. There's a plane in the sky that I can hear, now about 3km away from our house. Another plane 10km away. As I was recording this I was also rubbing both of my thumbs against my index and middle fingers just 50cm from my mic. I was not holding my breath trying to be quiet. Ambient level in my room could've been as high as 40dBA. What does my UMIK-1 with its 6mm capsule and cheap ADC hear? I recorded it and normalized it to 0dbfs, so turn your volume down before opening the file:

https://www.dropbox.com/s/ckwycsaybx3ngyt/UMIK-noise.flac?dl=0

Nothing, just noise. I can't hear anything past the UMIK's noise floor. My cheap WM-61A rig fares a bit better. A similarly cheap Neewer NW-800 electret capsule into a Terratec DMX6Fire USB soundcard fares a bit better than that and an SM58 into the same soundcard a bit better yet again. Neither of them allow you to take an SPL meter reading like @atomicbob has done, though.

This is not to say that the UMIK is exceptionally crappy, just that 1/8"/6.35mm mic capsules like in many cheap measurement mic and sound meters are incredibly limited in SNR. The eardrum with its 8-10mm diaphragm seems to fare much better.

 

I'm with @purr1n , there is something amiss with your experiment. Achieving below 25 dBA is really hard, almost anywhere.

Check out NWAA labs where Ron Sauro has a room with -6 or -7 dBA. Three containment walls each 10 ft thick.

 

How high is the Vali 1's noise into an HD800? Maybe that's a good example. Back when I only had the Vali I would always take off my HD800s when not listening to music - otherwise the noise would drive me crazy. I'm not saying my room is as quiet as the noise, but I can definitely hear it over the all the natural ambient noises of the outside world, which are far easier to ignore for me than the constant hiss of the Vali.

@atomicbob What do you mean by achieving 25dBA? Achieving such a low averaged noise floor over a couple minutes? Yeah, that seems hard to me, too. But being able to hear a 25dBA hiss doesn't seem hard to me.

EDIT: According to my measurements the HD800 is 115db/2.83Vrms or 106db/1Vrms. The Vali has a SNR specified as >93dB referenced to 1Vrms, so the noise should be <13dB. Well, another data point, I guess.

 

Why not both? Idle noise has nothing to do with gaming noise. Then again I stopped playing video games some 5 years ago or so. I play maybe 3 hours of GTA:Online and 2 hours of CS:GO a month nowadays.
My 5820k is only stable to 4.3GHz sadly. I get maybe 230fps in CS:GO at 1600x1200, which isn't that great considering that I get a 206Hz refresh rate at that resolution. Well, I could play at 800x600 where my monitor can do 370Hz and I reach 450fps in the fps_benchmark map, but that's too shitty a resolution for me personally.
My GTX980 only goes to 1500MHz boost and 7,6GHz core clock. Something like that. Input lag is far more important to me than fps anyway, although both are interrelated of course. And I can't just upgrade my GPU. The 980TI is the best card I could get for my monitor.

I just checked the specs for the Ragnarok and it's specified at >103dB SNR unweighted at 1Vrms in the low gain mode, which should be about 3dB with the HD800. Well, I tried and I can hear it (in low gain) at night with my monitor turned off when there are no planes overhead, but I never noticed it before, so it doesn't really matter. It's very faint, but this reassures me that my previous 0dBA white noise tests weren't bullshit. 3dB noise unweighted is a lot less than 0dBA.
Noise is easily audible with the HD800 in the high gain mode, though. Mid gain seems closer to low gain, tbh. I'm not sure I can hear the noise in mid gain with my speakers at the listening position, but in high gain it's definitely audible. My speakers are around 15dB less voltage sensitive than the HD800 it seems.

 

"Subscribed!"

As in, just saying, Hi, I'm reading along. Coming from the other thread, I'm following the discussion for interest. I don't have much to contribute in practice: that mosquito test, which I get to do often: they are silent for me unless they actually fly into my ear (they do sometimes). It was good when I could hear them 2-3 inches away, because it meant that a swipe down the side of my head with the electric bat would always get them.

Looking back in life, the times that I can can remember as being particularly silent were morning, after snowfall. Especially true when I lived on a rural lane, but even in London it made a huge difference. Otherwise, the silence you can hear is something I've barely experienced since I became a city boy 47 years ago. In this country, if you go rural, as in forest/jungle, you go loud! I could not believe the background racket when I first experienced that. Like the road taking one slowly closer to a vast machine. Except it wasn't a machine: it was a forest.

When I built my current PC, a decade ago, fans became a hobby. If I had taken my own advice ("Audio does not need much computing power: go for a low-spec cpu and graphics that barely needs cooling") it wouldn't have been necessary. In the end, I too ended up with a few items from Noctua, and was left concluding that, however silent the mechanics, the movement of air into, around and out of, the box, still makes noise. The same is true of Air-conditioning machines. My latest-bought is amazingly quiet compared to the previous generation, but it still has to shift air, and that still causes sound, and hey, I can even hear it!

 

Idunno, coil whine generally seems to be much more of a problem than fan noise to me. You can cool things passively, but you can't make electronics with silent coils seemingly. My GPU has coil whine, my laptop has coil whine. Heck, even my phone has audible coil whine in a benchmark. Even when charging my toothbrush there's coil whine. For the most part trying to ignore it is easier than trying to fight it, though.
I think I'd pay to spend time in an anechoic chamber. Or maybe a sensory deprivation tank could be interesting.

 

Serious, why the hell are you doing this? Your puny human snr is limited by your own breath. Human ears have a very limited continuous snr. They have to shift thresholds. The brain creates an illusion of the world through which we perceive. Our perception is an accurate hallucination exceeding our physical bodies.

Noise floors matter a lot more for tape, mic pres, and adcs than they do for dacs and amps. Start processing rough tracks and stacking them and dubbing them down and it’s easy to hit the noise floor. In-line compression and noise reduction algorithms are things for a reason. Anything around redbook is good enough for da. Anything beyond -100 from a reputable (no chifi!) brand is superb. What ultimately matters is the overall perceived quality and accuracy. Steady stage noise measurements don’t tell you that. For AD, you want it to be as clean as possible with maybe a euphonic vibe. Otherwise “outdated” Lavry, Prism, and Apogee gear back from before they ate the poisoned Apple shouldn’t be good. Flavor of the month hifi and chifi gear simply can’t mess with some boring RME, Lynx rack, MOTU, or upper end Focusrite box for both perceived sound and overall measurements anyway. Hell, the latter two can’t even mess with the former for stability at all.

Here’s some cool pdf I found for the members:
https://www.analog.com/media/en/training-seminars/tutorials/MT-003.pdf

 

This thread did make me fix something. I updated my BIOS a few days ago. The thing that pushed my PC to 33dB was the water pump. I had its voltage set too high. Dropped it back down and I can't hear my computer anymore. Yes, I'm a freak and have a full water loop with ultra low noise fans (Cougar. The Noctua white noise bothered me).

Now I can hear the Dynaudio BM12A from 1m or so (still way quieter than a Kali, for example. Not as quiet as HEDD). Luckily, it's quiet enough that throwing on my air filter or a fan completely masks it and I can stop being annoyed.

 

@Psalmanazar Yeah, I wasn't talking about sound quality in any way. I like vinyl even though it has a noise floor much higher than redbook. I like widebanders with peaks up to 2% distortion at 95 or 100db (then again many soft cone midrange drivers have similar peaks, many of them 3rd order which is much worse than 2nd order). I'm sure my preferences are very different from yours. I hate most small monitors (but I want to build my own that I don't hate - the ones I already built I still hate).
Why do I do it? Honestly I'm not so sure, either. Maybe the answer is that I enjoy it and that I want to find these edge cases of things that I can still hear, while I still can. Part of the reason is definitely that I think the commonly accepted numbers are just an average with all sorts data thrown out. Being able to hear 0dB at 1kHz (or w/e the commonly accepted threshold is) is a bit like 20/20 vision to me: It's what the average human should be able to do, regardless of age (to a point).
Kind of like how I did the 24 bit vs 18/20/22 bit vinyl rips and the sample rate comparisons with @trung225's Weiss SARACON. I now know High Res makes a difference to me, yet 100% of what I listen to is still 16/44.1. Or the test with the ultrasonic tones. Or the threshold test I mentioned.
Note that I don't do these kind of things constantly. These 4 things are pretty much all of these kind of "experiments" I've done over the past 2 or 3 years or so. Yes, time might be better spent studying and listening to music, but it's not like I spend a lot of time doing these kind of obsessive, weird audio experiments.

@Hands I uhh, don't get what you mean. What you described (the tweaking part) sounds like trying to reduce gaming noise to me, similar to the tweaking that overclocking is. My PC is just a copy of @ohhgourami's without the Noctua fans on the GPU. It required 0 tweaking for it to be silent in basic chrome browsing, REW, basic speaker simulations, Netflix and (for the most part) Google Earth if there wasn't coil whine. That's pretty much all I use my PC for anyway. Well, I still adjust the fan curves regardless, so maybe you do have a point. But it's well worth it to me, since it's really just a set it and forget it thing. For basic tasks the GPU fan is off, CPU fan is at 200RPM, PSU fan is off and otherwise there are no moving parts.

Marv: it was never about music. I still listen to stuff like Have a Nice Life - Deathconsciousness, even if the masters are sourced from 128kbit/s MP3s or so. I listen to modern, brickwalled music all the time. Most of the music I listen to isn't recorded well in the slightest.


One thing I should've made clearer from the start is that my room isn't at super low ambient levels for very long periods of time. Outside of a 2-5AM window there's a train and a plane maybe every 5-10 minutes. There are cars driving two streets over pretty frequently in the daytime and every 5 minutes or so at night. There could be wind noise or rain. The heater cycles on and off. Maybe I'm charging either my phone or laptop and the charger makes a noise. But all of those things don't bother me, what's important to me is that in-between these bursts of noise all I can hear is... silence (or a faint hiss if I don't get enough sleep).

 

I too did the fan-tweaking thing. Why? because... why not! Because, as previously alluded to, for a while my hobby was making my PC as physically quiet as I could. And because I have occasional techie urges that must be satisfied, and I have to do it by myself because my wife doesn't like that kind of stuff.

As a sidenote, who remembers the noise that common-or-garden PC power supplies used to make two or three decades ago? Not fan, but transformer noise. And I recall CPU coolers that sounded like trucks.

As to what gets sensed by our ears v what gets passed to our consciousness byour brains, I recall an excellent article describing our hearing in terms of a studio-like signal-processing chain. Given the analogy, I suspect it might have been in Sound On Sound. If I can find a bookmark, I'll post*.

Then there was a presentation by digital-age JJ Johnston demonstrating the amount of data which is discarded in our body-brain audio processing. **

Links if I can find them. Or somebody else might have. Here:
(NB: I have not read/watched for some time. I just remember finding them useful/informative at the time)

* Sound on Sound: How The Ear Works

** Might have been part of this presentation?

 

Long, long ago I was able to use an EMC-1 mic to test with REW and achieved about 16 db ambient noise level in my room. It's cement walls on three sides, in the basement as far away from mechanicals as possible. No lamps, no gaps, and at the time I had a shit ton of sound absorption panels up on the wall (6 of them, in an 11 x 13 room). If I just sat in the chair with nothing going on, it was eery. Like, almost hear one's heart beat quiet. Not quite, but close.

I designed the room so I could stick the PC in a space behind a wall in the far corner so it's out of the room entirely. And that was before I re-oriented everything a few years back. I just went back to that setup, with my 34" monitor on a swing arm and a 14' DP cable, I can make it work.

My room isn't that quiet now, but it also sounds better. I think if one can get to 25-30 db, at least anecdotally, it's a good balance between not being overly damped while still being quiet enough for low-level detail to poke its head through the mix and say Hi.

Using a phone app these days (I'm lazy and PC out of the room makes using REW a PITA), I'm at about 28 db with nothing but an air purifier running, about 8' away from my listening position.

The downside, IMO, is that in a very quiet room, all the warts and imperfections on a piece of vinyl are kinda laid bare. I've adjusted to it, but I actually kinda prefer casually listening in a bit louder space, when vinyl is on. But it's a catch-22, because some elements of vinyl do shine when the room is very quiet. Just takes very clean, very well-pressed music. We should be so lucky if it just all came that way

Todd

 

Okay, I am now at the point where I should've been when I first started this discussion: I gathered some data.

First up: I compared my volume level calibration against 3 SPL meters, one of which I now bought. It was within tolerance for all of them and it's 0.4dB or so off from the one I bought. The most expensive one I tried was a Voltcraft SL-451, though which isn't that good. Possibly it'd be more off comparing against real measurement equipment. Overall I decided to keep my calibration from the ND9, especially since atomicbob measured it.

The other thing I did was to try to get a reference for how loud certain areas are. I took two measurements at the technical university in Munich, both of which I consider relatively noisy areas. Keep in mind the scales are different for all three graphs.

The first was an empty lecture hall at noon. More than half of the right wall is made up of ancient windows which don't seal at all. It's next to a very busy street with trucks accelerating frequently, etc. Lots of cars, lots of street noise:


The other measurement I took was in the library. This is a much more noisy area. There are hundreds of people here making all sorts of noises:

I'd say both rooms fall within the ca 35dBA area that Marvey calls a quiet room, but I have to disagree here. I'd think most residential areas are quieter within the houses, just that measurement equipment gets expensive for lower noise than 30dBA. My SPL meter with a 1/2" electret capsule seems to be limited to 28.2dBA and will write "under" for lower levels.
BTW: Those two were recorded from my laptop, after I calibrated the mic against my UMIK. Laptop mic noise is 30dBA, so the empty lecture hall gets a little too close to comfort for me to the noise limit. In reality it might be quieter than 30dBA.


Now compare both against a 3 minute snippet out of my bedroom at night. For now my system noise seems to be limited to 10dBA, but maybe I can still slightly tweak it for less noise. This was recorded with a low noise 1" condenser mic and I doubt most 1/2" capsules (that aren't specialized for high sensitivity) will have as low a noise.

I was lying in my bed when I recorded it, trying to be quiet, but the small spikes are from when I moved a bit. Obviously no train, no plane and no car within those three minutes. The trains are about 20-25dBA, planes maybe a bit louder and cars most likely a bit quieter.
When I get close enough the mic can actually hear me blinking and record my heartbeat, but there's still a lot more noise than I hear IRL in the recordings.

 

I wonder if you bought a lead lined box to fake measurements at home now. There's no way your bedroom at night is 11dB on average.

 

Rig is LCT550 into a Focusrite 2i2 gen 3. Gain maxed out. I don't see why it's not 11dBA or lower - I can't hear anything. Without me in the room it could be significantly lower than that, even. I could try to measure it when the heating cycles on and it's a bit louder. Could be above 10dBA in that case.