Galaxy A52s: Unavoidable noise suppression, even with Dayton Audio iMM-6C?

[Klaus]

Hello!

I need some advice on whether my Galaxy A52s can in principle be used for noise measurements. I recently had the opportunity to compare the readings of its main microphone to a calibrated device at a city office (gave a single-offset of −11.5 dB), but have doubts about interfering filters. I did (too many) checks with Google Gemini, but at this point I need human advice.

Basically: Both with the Dayton Audio iMM-6C and with the builtin main microphone I see a down-correction when starting a white-noise generator on my laptop. For testing, I set my laptop speakers to maximum volume, and used mynoise.net at maximum volume and measured while suddenly turning it on.

Left: Built-in microphone; Right: Daytom Audio iMM-6C. Both show a 2-3 dB(A) attenuation of the white noise reading. This seems to imply the presence of noise filtering, that would make noise measurements highly unreliable.

For comparison, the issue is more pronounced in OpeNoise. 

Left: Built-in microphone; Right: Dayton Audio iMM-6C. For the builtin microphone a very sudden onset of the noise suppression by more than 10 db(A) is visible, for the Dayton Audio the attenuation is similar to the data from Audiotool.

As far as I understand both apps use the VOICE_RECOGNITION audio stream, so it is confusing why OpeNoise would be so much worse, but only with the builtin microphone. 

I have also tested with "Oboe Tester", that the VOICE_RECOGNITION and UNPROCESSED audio stream exists for the main microphone and for the Dayton Audio. 

My main question here: Can I get useful relative A-weighted sound pressure levels with this setup? Am I right to be concerned about the 2–3 dB(A) "dropoff" behavior? What else should I check? Does the vastly different attenuation behavior of OpeNoise give any useful hint?

Best regards,
Klaus

[Ian Eales]

I've used AudioTool with 3 different Pixel models. All are pretty consistent.

The internal mics in my experience are accurate enough for relative measurements.

--
You received this message because you are subscribed to the Google Groups "AudioTool Discussion Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email to audiotool-discussio...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/audiotool-discussion-group/66e975ef-6f53-4471-af77-58c2beee6423n%40googlegroups.com.

[Teemu Järvinen]

In Samsung, there's something like soundcard in usb port. My Lenovo behaves differently.
I was testing how to make audio in my PC, and use that audio in a software in phone/tablet, using usb port.
Did not work with Samsung phone, but Lenovo tablet worked fine.

--

[Klaus]

This is merged from another thread, that I thought had not been created when writing on mobile.

MY ORIGINAL MESSAGE IN the duplicate thread:

On Sunday, March 29, 2026 at 10:16:21 AM UTC-7 bauer.kla...@gmail.com wrote:

Hello!

I am trying to use Audiotool to measure noise levels in a neighborhood conflict.

I was able to get to a point with Google Gemini. But I have run into issues where I need human feedback.

• I have calibrated the main mic with a professional device at around 55 dB(A), which gave me an offset of -11.5 dB.
• I have a Dayton Audio iMM-6C. However, when installing the calibration file I get unrealistic 45 dB(A) as the noise floor (quiet room, microphone and phone under multiple layers of blanket), so there seems to be some artificial gain added to the Dayton.
• According to OboeTest, my phone supports the "UNPROCESSED" Sound source setting.
• The Dayton occasionally show a peak 10–15 dB(A) higher than the final value at the beginning of a measurement (Chat view), so I am worries that some noise suppression is active. The effect was most visible when using a white noise generator app on my laptop as a reference.
• With the main mic this initial drop (after closing an reopening the app) is *not* present for a loud source but may appear for lower sound pressure levels (< 50 dB(A)).
• When measured after midnight in a room quiet enough to hear heating pipes and the noise floor of my own ears, measuring under a blanket gives noise levels on the scale of <10 dB(A) — far below a realistic noise floor for even professional equipment.

Any ideas what I could do to get useful values?

REPLIES SO FAR:

On Sun, Mar 29, 2026, 13:21 Julian Bunn <jjb...@gmail.com> wrote:

I'd recommend purchasing an SPL calibrator, like this one: https://www.amazon.com/Decibel-Calibrator-Calibration-Measurement-Microphones/dp/B0786YDF62

just so that you know the absolute dB levels you are measuring with the iMM6 make sense. The Dayton cal file is only for adjusting relative levels as a function of frequency, not for adjusting absolute values.

Other people on the list may have more useful advice on your measurements.

Julian

On Sunday, March 29, 2026 at 10:28:04 AM UTC-7 Non Timebo Mala wrote:

All this noise measuring you're doing is going to be useless in court because you're not going to have reference levels.  In other words think about it like this decibels mean nothing to the judge.  Now if you go out in the street and you measure traffic driving by and you get a level then you will be able to say to the judge the traffic level by my phone is so many decibels.  This chart here that I got from the internet says that this many decibels is what traffic is.  Let's say it's 70 decibels.  Well then I went over to my neighbors fence and I measured it and it was 90 decibels.  You see this chart here says that traffic is actually 77 decibels and my phone only measured 70 so that means that this measurement at 90 is really 97 and if we look at the chart we will see that 97 is too damn close to an airport.  And that is how you're going to have to make your argument because the judges ears are not calibrated in decimals they're calibrated in airport.  Think about it before you walk into that courtroom with a bunch of numbers that don't mean the damn thing to the judge.

https://keys.mailvelope.com/pks/lookup?op=get&search=0xF77691B7C3A881FE

Ian Eales schrieb am Sonntag, 29. März 2026 um 20:07:36 UTC+2:

+1 on the calibrator. If you get one for a 1/2 inch mic, you can make an adapter with a bit of plastic hose from the hardware store to fit the imm6

It would help to know what phone / OS as they likely operate differently.

I found that regardless of which mic is active, plugging in the imm6 is active.

I recommend creating a nul cal file and always loading that at app start up. If you have a cal file loaded and change mics it remains in process.

To create a nul cal file just start to create a cal file and save it as nul.cal with all zeroes. Then Finish calibration. 

MY REPLIES:

Klaus schrieb am Sonntag, 29. März 2026 um 22:06:25 UTC+2:

For reference, I don't need it for court, but for a discussion with the neighbors. Under the Austrian legal framework, in front of court only a court-ordered expertise really counts, and with that we'd be talking about EUR 4,000–5,000 after tax. The measurement would primarly be intended for further discussion to estimate whether the neighbors' complaint has merit and should be taken seriously, or whether I have to say "you expect unrealistic things, but I can't stop you from sueing". 

Klaus schrieb am Sonntag, 29. März 2026 um 22:06:30 UTC+2:

Regarding the calibrator: I try to avoid costs here, or my wife will have my neck for it. I had a chance to compare my phone's main microphone with the calibrated device of a city official (gave a single-offset of −11.5 dB(A)), but I have since accumulated some concerns about the presence of digital filters that may make a calibration pointless. 

[Klaus]

For reference: I tried to compare output levels of my Surface Pro 7 Plus to the reading levels in Audiotool when using white noise. The curve is roughly linear when plotting the Leq(A) value from Audiotool over the volume setting of the laptop double-logarithmic. 

Not sure how useful it is, as I can't say for sure how the laptop translates volume levels into actual output loudness. But the Leq(A) level being linear with the logarithm of the volume level except near the background noise level seems reasonable. An increase of the volume level by 10 increases the measured Leq(A) by 3.64 db(A). 

[Klaus]

The big trap is whether Audiotool receives sufficiently unprocessed data. In hindsight the 2 db(A) drop for a sudden onset of constant loud white noise isn't that concerning I guess. The behavior in the other app (OpeNoise) however, is very concerning when using the internal microphone.

For the Dayton I was originally misled to think that the peak when starting the measurement is a huge noise suppression filter, but later I noticed, that it appears regardless of noise, and that actually toggling white noise abruptly has only the 2 dB(A) drop-off that is less problematic. Not clear why it appears at all though.

My interpretation of the data: Having calibrated the device against a professional device might with a pink-noise reference, the phone could have attenuated the noise level by 2–3 dB(A), such that measurement of other sounds might give a value that is too high by 2–3 dB(A). But since the noise I want to mesure has the characteristic of background noise (running water, sound from water pipes, sound from waste-waster pipes), the calibration should be reasonably accurate.

At night under a blanket I measure a noise floor of around 25 dB(A) which seems a bit too good to be true, but possible (expected range: 25–30 dB(A), but it isn't a high-end device). The Galaxy A52s has much better speakers for instance, than my wife's Xiaomi device, so the microphone might also be a good one. In a very quiet night on the level of "heating pipes in the wall are the loudest noise", I get values as low as 5–10 dB(A), but at that point it has to be noise suppression, as even professional audio equipment and noise meters shouldn't have such a low noise floor. The class-1 device of the city administration has a noise floor of 17 db(A) for instance and 25 db(A) is the lowest value found in applicable norms for construction in Austria.