This changing sound signature over the firmware changes is very strange to me:
With lossless material (FLAC...) and with the EQ off, the only purpose of the CODEC is to decode the original samples of the music, as-is, and send that data to the WM8740, I see no need to alter these data before sending it to the DAC.
All the DSP processing is internal to the WM8740; Filtering, oversampling, volume control, Sigma-Delta...
I personally made some fast measurements on my DX50 (1.2.7), with a scope and a multimeter ( maximum level, volume = 255, phone out, no load)
I use pure sine signal from WAV files, at maximum level (0dBfs)
Frequency response is ruling flat from 20Hz to 20kHz, maybe -0.1dB at 20Hz and -.2dB at 20kHz
(with a load, low frequency response will be affected, but its not the point here)
Personally, i didn't notice any sonic difference between firmware versions, but i think its very difficult to validate that without owning two DX50
At the same time, I also check the EQ function:
Frequency response is also ruling flat
When you activate the EQ, the level of the signal is reduced of 6dB, (1/2 of the full scale voltage), this is normal behaviour; if you want to boost your signal, you need to attenuate it before to avoiding numeric clipping (saturation)
But... there is a Bug:
In the 1.2.7 firmware, Ibasso push the maximum boost to +12dB, but keep the same pre-attenuation of 6dB !! leading to risk of clipping
And there is!!!:
If I put the 1kHz slider to +12dB, the 1kHz sine signal is hard clipped!!, i see this clipping from +8dB to +12dB of boost, But I think that this clipping must begin at +7dB of boost
This can explain the strange frequency response we see in a previous post with all sliders at +10dB
This clipping seem to affect only the frequency bands which were boosted (its logic)