As requested, here is a picture of the XO on its dedicated board.
The board has been installed in a small free area on the main board, not too far from the large 6800uF caps, which are to the left, out of picture. The large white wire in the picture runs to where the XO originally sat on the PCB.
This modification should make it easier for me to swop out the XO. I'm not sure if the 14 pin board will help here.
I've had the unit a week now, and after sounding pretty raw on day 1, it is slowly starting to improve. Partly due to the the op amp rolling no doubt, and partly perhaps due to the fact that the DAC is still burning-in.
One of the nice things in the LKS design is the XO is very close to the DAC.
LKS very carefully terminated the XO input to the DAC to avoid the signal bouncing back and forth between the DAC input and the XO output.
In the LKS it is complicated as the signal in split 2 ways.
Below is the XO termination in the LKS.
Adding a coax cable between the XO and the DAC input changes the relationship, the coax cable capacitance adds a delay that is not constant across the frequency band and this distorts the shape of the square wave coming out of the XO, this affects the performance of the Digital PLL in the DAC.
The extra wire also changes the impedance due to the capacitance of the cable, the CCHD-575 is specified to drive only up to 15pF of load and that is not a lot.
The impedance is now mismatched and you will get reflection, this worsens the jitter performance of the XO.
This might explain why you are hearing a very harsh output out of the DAC.
Here is an example of remote clock line done correctly in the Gustard X20u, the rectangle in green is the line driver to handle the extra load from the coax cable capacitance.
The EE Minimax Supreme also places the clock very close to the DAC
The stock LT1763 regulator noise is specified at 20uV over 10Hz-100kHz. This gives a noise density of 63nV/√Hz
https://cds.linear.com/docs/en/datasheet/1763fh.pdf
Check that Effective Audiomod regulator you are using is better than this number otherwise return to using the stock regulator.
Early numbers from DIYAudio suggest the noise is 1mV over 250kHz, this gives a noise density of 2000nV/√Hz
http://www.diyaudio.com/forums/digital-source/84495-effective-audio-clock-singapore-4.html#post983797