I have a computer dedicated to Audio : a cheaper PC with lots of harddisk storage and a good soundcard (RME Digi96). This soundcard has an ADAT I/O (8 channels 24/96) and a SPDIF output (2 channels 16/44). My DAC has a SPDIF input. Why not connect the two ?
So that is what I did, using a run of coax cable found in the leftover parts bin and two no-name RCA sockets (gold-plated, eh).
The DAC's Clock output is unused. This means it performs asynchronous reclocking on the incoming signals. This means the difference between the DAC clock and the Soundcard clock will generate a beat frequency, and at every period of this beat frequency, one sample will be a bit longer than the others. Quartz clock precision being quite high, this only happens every few seconds and should not be noticeable. Some have tried asynchronous reclocking with success. I know it looks a bit fuzzy, but as the Brits say, the proof is in the pudding (or something).
So now the computer sends its bits to my DAC through SPDIF.
And it sounds an order of magnitude better than the previous digital connection between CD723 and DAC ! Every aspect of the sound improved vastly, especially air, stereo image and details.
How can this be ? I suspect the CD723 has a compromised digital output. This output has a digital volume control (from the remote) ! Argh ! I'm pretty sure now that even at 100% volume the signal does not run through untouched and that something evil is done to it, like scaling it without dithering. Bleh.
Here is the sad truth :
The CD723 spdif output is not bit accurate to what is on the CD.
These manufacturers have no shame. You may have noticed that the remote control has Volume buttons which also work on the SPDIF output, acting like a digital volume control, ie. a multiplier. However, at maximum volume, the multiplication coefficient is not 1 (as expected), but 0.9975, which leads to truncations in integer arithmetic and loss of information and audible detail carefully dithered into the LSB by conscious mastering studios.
To prove that I recorded the CD723's digital output with my soundcard and compared it with a rip of the same CD. I know my rip is bit accurate (I compared it with the burnt wav files) and that my soundcard is bit accurate (RME Audio is a Pro brand and they advertise their product as bit accurate). I then wrote a little program to compare the two files (find the offset and look at the differences). I had to make it look at the scaling factor, too, which is how I came to this 0.9975, which was actually measured using my program. It also reported 98.997% of samples being different between what the CD723 read and what was actually on the CD.
Incidentally, I have repeated the same experiment with a Harman-Kardon DVD1 used as a CD transport. Here, my program finds a scaling of 1.0 (which is correct) and no bit errors (all samples are bitwise identical between the SPDIF recorded stream, and the wav file which was burnt on the CD). This proves the testing protocol is valid, and that the CD723 indeed has a problem...
Therefore, as we already knew, the CD723 is unusable for tweaking because its digital filter is crippled (it does not dither its output), then now we also know the SPDIF output is broken. I'm afraid we can do nothing useful with this player.
Argh.
Now I have to listen to more music, and try synchronous reclocking with the computer (this is a hack, I just have to feed it a SPDIF signal synchronized with the DAC clock and the soundcard will automatically slave its clock with it). This way I can switch between synchronous & asynchronous reclocking just by plugging/unplugging a connector. More on this later.
I also have to try a BJT in the output stage as Rudolf Broertjes suggested.
Coming soon ! Stay tuned !
Coming soon ! Stay tuned !