Hmm, I think all the indicators are that it is just the natural limitations of the sound in something like an iPhone, with a tiny speaker and limited amplification.
The source audio is fine, very clean, which probably just serves to emphasize the distortion that does exist in modern small devices like phones and also laptop speakers. That distortion is usually not noticed, especially on voice, as it just helps the sound of the voice to "cut through" ambient noise and our brain is good at disregarding it. Even a lot of music doesn't cause us to be aware of such distortion, since modern music is typically already distorted, with compression, limiting, saturation effects etc. A clean single instrument sound with a wide dynamic range such as PianoteQ is demanding on audio reproduction, as you said, not every speaker is able to reproduce it well.
I can spot various distortion effects in the recording of the iPhone sound, two I have already suggested, clipping and slew rate limiting (which is an internal characteristic of amplifiers). Comparison of the two waveforms also indicates the iPhone is using an ALC (Automatic Level Control) system, which is another thing that tends to create distortion. It also looks to me like the iPhone is getting a bad result from the stereo processing it does, although I might be seeing an artifact of only having a mono recording, from one microphone. The visible high frequency distortions though do seem to coincide with the harsh, crackly sound during the quieter sections where you wouldn't expect there to be problems like clipping. If that harsh sound in the quiet portion is absent when listening on headphones or earbuds, that would indicate the tiny internal speakers themselves are creating that distortion.
In the louder portions, the ALC has initially turned the level up higher in the quieter section than in the original audio file, so the ascending motif starts out much louder than in the recording, and is clipping (maybe also just the phone volume turned up too high?), the ALC then backs off, as they do, and you can hear the two chords prior to the descending portion are quieter, although still distorted. In the source being played from, it's the reverse, the ascending run is quieter and the chords are louder.
So it seems to me that the good quality recording is simply being poorly reproduced by the iPhone, the sound being mangled by ALC, processing intended to make voice highly audible, and tiny speakers with a low powered amplifier.
Others may have further suggestions and observations. At this point my only thought would be to try creating a potentially "iPhone friendly" export in mono with a small dynamic range and an overall lower level (the current file peaks at about -0.5 dB, perhaps try about 2 dB lower), and see if it sounds any better on the iPhone.
Last edited by Platypus (24-12-2020 01:47)
