Some strange clipping with Samsung dongle

[danadam]

I'm posting this more like a curiosity rather than a problem to solve.

In the attachment there are 2 files, 44.1 Khz sample rate, with a series of pulses in groups of 4:

• In "bw.sweep" each group has bandwidth 500 Hz wider than the previous one, starting from 500 Hz. All pulses are normalized to -10 dBFS.
• In "gain.sweep" each group is normalized to 1 dB higher than the previous one, starting from -20 dBFS. All pulses have bandwidth 2000 Hz.

The files are safe to listen if you want to. Here are their spectrograms:


And here is a waveform and spectrum of one of the pulses with bandwidth 2000 Hz:


I played the files on Fiio K3 (ESS), Tanchjim Space and Samsung dongle and recorded on ADI-2 Pro FS BE at 192 kHz sample rate. On Fiio and Tanchjim everything was fine. On Samsung dongle it clipped, heavily. Or maybe rather some limiter is engaged? Here are the spectrograms of the recordings converted to 48 kHz:


And here are the recorded waveforms of the first and the second pulse in the 4th group of "bw.sweep" recording:


The 3rd and 4th pulse in that group look the same as the 2nd one. This pattern repeats for the remaining groups, only the amount of clipping increases with increased bandwidth. Here are the peaks for each group:

             Overall     Left      Right
Pk lev dB     -11.74    -11.75    -11.74
Pk lev dB     -11.69    -11.70    -11.69
Pk lev dB     -11.70    -11.70    -11.70
Pk lev dB     -13.17    -13.17    -13.20
Pk lev dB     -14.12    -14.12    -14.18
Pk lev dB     -15.10    -15.10    -15.14
Pk lev dB     -15.10    -15.10    -15.14
Pk lev dB     -15.12    -15.12    -16.09
Pk lev dB     -16.02    -16.02    -16.07
Pk lev dB     -16.00    -16.00    -16.04

There's nothing obviously wrong with the dongle when listening to normal music. Well, at least not to me. ¯\_(ツ)_/¯

 

[AnalogSteph]

If it takes such a special test signal for this issue to appear, I would suspect it is related to something like analog rail switching (i.e. a Class G or even Class H output stage). This may be too slow to follow the signal accurately. Makes you wonder how it would do on (original mastering) '80s pop music...

 

[danadam]

Makes you wonder how it would do on (original mastering) '80s pop music...

Yes, it can happen

I played "Flim & the BB’s / TriCycle / 01. Tricycle" and recorded at 88 kHz sampling rate. For the most part it is OK but in one place it triggers the issue. Here's PK metric:


And waveforms comparison, the file at the top, the recording at the bottom. First at the beginning, where there's no problem:


And where the issue happens: