Samsung Galaxy S23 Ultra Calibration File

[Larry]

Hi all,

I have created (unscientifically!) a calibration file for the bottom internal mic of the Samsung Galaxy S23 Ultra (UP1A.231005.007.S918WVLS4CXE9), v6.1. I have no way of confirming whether it's really accurate or not, but the methodology and source files will be published here for those who want to play with them.

I had a Dayton iMM-6 calibrated mic from a while ago, which is plugged into a Sabrent USB audio card. I loaded the mic and a sound card loopback calibration into REW (Room EQ Wizard 5.31.3). I calibrated the SPL on this mic by first setting up the levels, then recording a rather loud high-speed computer server cooling fan whose manufacturer specifies 40.5 dB(A) at 1 m away. I put the mic 0.5m away in a pretty quiet room, adding 6dB to compensate for the halved distance (this was done to improve the signal to noise level). 

I then placed a pillow (to absorb reflections to some degree, audibly, this works well for recording voice but not sure if this was a good idea) on the desk, lifting it to a 45 degree angle facing upwards (for convenience), and positioned both the phone and mic facing the same way close to each other. The bottom mic on this phone is to the left of the USBC port, when facing the bottom with screen facing up. 

A test sine sweep with acoustic timing reference was played through the ceiling mounted 2.1 in-ceiling speakers in my room, which themselves have some DSP correction applied to make the room response better (but not perfect, because I calibrated them with an uncalibrated microphone before). The measurement has 1/3 octave smoothing applied, to get rid of the mess caused by the room reflections.

In order to capture the response of the phone's mic, I downloaded and installed an app called Voice Recorder v5.12.3, from Simple Mobile Tools, because it allows the selection of the same mic and the same "voice recognition" profile that AudioTools uses. 

I made a recording of the test sweep, (leveraging the fact that it has an acoustic timing reference) at the highest supported audio quality level. Then, I imported this file to the Mac computer, and used Rogue Amoeba's Audio Hijack (playback of audio file into device) and Loopback (virtual audio cable) to send the file to REW. I made sure to disable the sound card and mic calibration files for the Loopback interface, then started a measurement while playing back the audio file.

Then, I used the Actions menu > Alignment tool to visually align the two traces in volume as much as possible by watching the shape. I created these two measurements, then used Actions > export as txt to export at 20-20000Hz, 12 PPO. 

I used ChatGPT (actually Claude 3 Sonnet) to write a Python program to find the difference and convert to the Dayton Audio format. If you would like to use it, you will need to change the file names to the actual names of the files you are using. 

After importing the calibration to AudioTool and loading it (and confirming the direction of correction is reasonable), I played some test tones and compared the SPL to REW, where it was generally in the same ballpark. Of course, I think there is at least +-2-3dB of error. 

Hopefully this is helpful to someone, with the disclaimer that the testing was performed rather unscientifically and not in a controlled environment. I think that it is good enough for what I wanted to do, and it removes an obvious bump in the response of the mic by default.