iPhone SE (2020) Recording Quality

I’m a big fan of the 2020 iPhone SE. It’s $399 and has pretty much anything a cell phone normie like myself could want. BUT: Do its audio recording capabilities stack up?

The specs

The SE has three mics – one at the top by the front-facing camera, and two at the bottom where the speakers are. The two mics at the bottom allow for stereo recording.

As it seems is the case with most iPhones, the SE’s sample rate is 48kHz and bit depth is 16 bit. I confirmed this by trying various sample rates and bit depths in the TwistWave Recorder app. The app said that the device didn’t support recording above 48 kHz, and although TwistWave supported 32-bit processing and export it didn’t indicate the possibility of 32-bit recording. I couldn’t find a specs list or website to confirm this, but it’s certain that 48 kHz and 16 bit are supported by the hardware.

Recording app

I’m using TwistWave Recorder (shout-out to Michael Lee for the rec!). There are options for lossless audio, disabling Audio Gain Control (see the “Enable iOS processing” toggle), and controlling bit depth and sample rate:

Testing the frequency response

In Audacity, I generated 15 seconds of a 100Hz to 18kHz sine sweep and 15 seconds of white noise. I played these from my computer speakers at max volume and recorded them on TwistWave Recorder on my phone, angling the dual mics at the bottom of the phone directly towards and in front of the computer speakers.

In Ocenaudio, I generated FFT analyses of the original and the recorded sine sweep and white noise samples. The FFT analysis shows us the dB FS of the spectra of the samples and the recordings, which we can interpret to evaluate the spectral flatness of the recordings. We’re looking to judge whether the iPhone mics have a flat response (the mic accurately reproduces the original sound, with even sensitivity to different frequency ranges) or a shaped response (the mic is more sensitive to some frequency ranges than others – ranges of higher dB FS indicate higher sensitivity). See this Shure webpage for more detailed info.

Compare the graphs and sound files for the sine sweep…

FFT analysis of the original sweep tone

 

…And for the white noise…

FFT analysis of the original white noise

 

FFT analysis of the recorded white noise

 

 

 

 

 

 

 

Both of the original test audio files have pretty flat graphs, with a significant drop a bit below 20kHz (roughly the upper limit of human hearing). This drop is also reflected in the graphs of the iPhone recordings.

Qualitatively, it’s easy to see that the iPhone has a pretty bumpy frequency response, with a weaker sensitivity at the lowest range (0-100hZ), a peak around 1500hZ, a valley around 5000hZ, a peak around 7500hZ, another peak around 12000-15000hZ, a weaker response from 15000-17000hZ, and then the drop from 17000-20000hZ as mentioned above. This holds true for both the white noise and sine sweep recordings.

This makes a lot of sense when you consider the practical purpose of an iPhone microphone: transmitting the human voice at GREAT quality. The weakness at the lowest and highest ranges of human hearing make sense – those extreme ranges are not important in the vocal spectra and can’t be heard very well anyway.

1500hZ, which I mentioned has a peak on the frequency response, is pretty important for decoding the human voice. Vowels fall in 200-2000hZ and consonants in 2000-8000hZ:

Those higher frequencies around 12000-17000hZ are also pretty well taken care of by the iPhone, which is good for enabling a crisp/clear/punchy sound.

So we can’t say the SE mics have a totally flat response curve, but they’re darned good at capturing the human voice. And it’s a relatively cheap phone, and cheap is cool.