hi, Drohun.
I'm measuring 750ms (max) latency (410ms min), using default configuration in ASIO4ALL (latency compensation: 32 samples), and RMAA to measure. The signal is clean, not like the USB DAC of my DX90 (very noisy). I can't compare to DX50, because I don't own it.
Stock Zishan uses output coupling caps, and they are not big, so you can experience some sub-bass roll-off if your phones are under 16 ohms (-1.5dB at 30Hz @ 16 ohms, not much anyway). But all the phones you have mentioned are above 50 ohms, so you won't notice any sub-bass roll-off. I think that DX50 also uses similar output coupling caps (DX90 doesn't).
Output impedance of DX50 phones output is 2.5 ohms @ 1KHz, while Zishan is under 1 ohm (0.22 ohms @ 1KHz) like DX90.
So Zishan will work better (less tonal alteration) than DX50 when using balanced armature phones, if you ever plan to use them.
The software is basic, but has got main features and it is very responsive. You can select five different DAC modes, and you can use parametric EQ files (rockbox format).
I use rockbox in my DX90, so the user interface is far more complete. Compared to stock firmware, I prefer the Zishan one (mostly because of its ability to use my rockbox parametric eq files). But iBasso provides firmware updates and it's a big brand, while you have to download Zishan firmware updates from a site in Chinese, and we don't know about its continuity.
The ability of replacing the batteries with "standard" replaceable batteries easily, is a very good advantage of DX50/DX90.
Like you say, you can roll low pass filter op-amp easily in Zishan, to get slight sound differences. I've tried stock LME49720NA and Muses similar to 8820, and enjoy them both.
The higher power of Zishan (3.7Vrms vs 2.4Vrms of DX50, @ >64 ohms), allows you to feed harder to drive phones, if you plan to use them.
So both have their advantages, and that's why I use Zishan as USB DAC and when using harder phones, and DX90 as portable.
