Thanks for all your replies. I have some interesting additional observations since my original post...
I got curious of the benefits of Audio-gd's ACSS connection as well as their DACs in general, so ended up picking up an NFB-1WM (purchased direct from Audio-gd, factory upgraded with TE8802 USB input) and NFB-1ES (purchased here on the Head-Fi forum). For the NFB-1WM, I chose the WM8805 SPDIF input chip over the DIR9001 as I do have some 24/192 material so the 24/96 limit of the DIR9001 would be an issue for me.
So, now I have two different DACs with the WM8804/WM8805 as the SPDIF input, and another using the Sabre ES9018 chip. And as a bonus, a TE8802 USB implementation to compare. My first test when I got the DAC units was to use them to see if I can hear the difference between the Audiophilleo and the Evo...
...and to my great disappointment, neither Audio-gd DAC would lock properly with the Evo at 24/192. Up to 24/96, it's fine, but at 192khz it all falls apart. Tried BNC, RCA, TOSLINK... Tried different cables, I even got a set of BNC 75ohm attenuators to see if that helped. And I tried this with two different Evo units to no avail. Note that all along I am using the TeddyPardo PSU so I am feeding the Evo with very clean power. Unfortunately, all I got from the DACs when playing 24/192 was either silence (Sabre's case) or loud static in addition to music (WM8805's case).
And guess what - the Audiophilleo had no issues whatsoever playing 24/192 material through either of these DACs. Both in direct-connect mode and cable mode.
What's interesting is that when I asked Kingwa about this issue, he said he tests his DACs prior to shipping with a non-Evo hiFace and his own DI unit at 24/192 and has not found any issues. Here is his quote:
The NFB1 WM can working find at low to 0.35VPP coaxial input signal in our test, I own a hi face BNC version and test the gears all can work at 192KHz.
Our DAC design maybe not as some other DAC, we have a TTL input after the coaxial socket, it can change to TTL for the interface working better but now some other DAC have not this TTL converter.
Now I'm not sure what the implications of this TTL conversion stage for the SPDIF input are (can someone shed light on this topic?), but two new inferences I can draw from this experiment:
1. The Evo produces either out-of-spec SPDIF levels or too much jitter for Audio-gd DACs to register properly at 24/192, perhaps even worse than the original hiFace;
2. Despite using very similar SPDIF input chips (WM8804 and WM8805), the Neko D100 is MUCH more immune to varying degrees of SPDIF source quality than the Audio-gd NFB-1WM, so implementation around the chip seems to play a huge role.
Now that the Evo vs Audiophilleo comparison ended rather quickly in the latter's victory, I spent more time comparing the TE8802 asynchronous USB input and the SPDIF input of the NFB-1WM to see how well the USB implementation stood up to the SPDIF implementation fed by a quality source, the Audiophilleo 2. Given that the SPDIF input is a more mature platform and the Audiophilleo is a proven (not just by me!) excellent SPDIF source, I have to admit I was biased towards the SPDIF input going into the test. After all, Audio-gd's TE8802 USB input board had no fancy looking clocks or parts, claims of technical superiority on the website (in fact Audio'gd's flagship DAC still doesn't use this input), or a proven user base who have vouched for the platform... The buyer is led to believe it's provided more as a convenience feature than an upgrade.
Anyway, I have played violin in the past, so picking out the timbre and tone of the violin sound often ends up being used to put a final stamp on my assessment.
Here's my summary, to the best of my abilities. The SPDIF through the WM8805 chip sounded rounder and less immediate, at times with less soundstage compared to the TE8802. A bit more forgiving and natural, perhaps? The TE8802 in comparison sounded more direct, with better soundstage, with everything seemingly more defined with better treble presence at the risk of sounding harsh at times. Initially though I wasn't fully convinced whether it was more actual detail of the recording coming through, or if it was jitter-induced digital artifacts causing perceived increase in treble presence (high-pitched dither like noise?) causing a fake expanded soundstage and directness of the sound. It took me a few days of listening to come to a track where a violin solo performance hit me - the song played through the TE8802 sounded real and immediate, whereas the song played through the SPDIF sounded, well, like any other recording and a bit distant. The cues that led me to this? When the bow hits the string and the string starts vibrating, with the TE8802 I was able to hear not just the note being played by the string, but all the extra vibrations that one can feel actually playing the violin - perhaps, the air around the notes as well as the notes themselves. The realism of the TE8802's presentation convinced me the direct USB --> I2S implementation of Audio-gd's DAC was not adding "fake" detail but in fact digging out more from the original material and no more. In comparison, some details were being smoothed out in the added layer of conversion in the USB --> SPDIF --> I2S process. Maybe the DIR9001 is not as lossy and that is why some still prefer this versus the WM8804/8805 despite the lack of 24/192 capability?
Anyway, long post here... if you made it this far, thank you for reading,hopefully this can stir up some interesting discussions. Once I am done with comparing the WM and ES versions of the DACs I'll post some impressions as well. This will be interesting since the power and analog sections of the two DACs are 100% identical and only the D/A conversion boards are swapped out in the two DACs. So it'll be a more objective comparison of the two chip's sound signatures, which otherwise is difficult since a DAC's sound is so heavily influenced by the components built around it...