This is the best way I have found so far to dump the R-W data of my CD+G disc Rock Paintings: CD+G Sampler, as example. Dumping fully error free the subcode of this disc was the main reason of adquiring my first real Plextor drive, the PX-W4824TU, because I read that Carlos Hernández (IpseDixit) said many times that the proper way of dumping the R-W data of a CD+G disc is a real Plextor drive and the special mode for CD+G which applies error detection and correction, the packed mode. And to clean the P-Q subcodes unaffected by the packed mode the CDGTool program by Truong Hy.
http://redump.org/disc/23529/
http://club.myce.com/showthread.php?p=2128678
The test I ran yesterday, extracting 20 times the subcode of that CD+G to clean after the P-Q subcodes using CDGTool.
PerfectRip and both cdr and ccd rips have to be selected simultaneously.
8x read speed, C2 pointers and the packed mode for subcodes.
Standard offset correction (anyway this disc has no data pushed either to the first pregap or the lead-out) and single sector read commands, not bursts of 2x sectors. With bursts reads, according to the tests I did two years ago, R-W is always somewhat different.
Repeat this 20 times, giving to the drive (and me) some time to rest and cool down.
Results after cleaning the P-Q subcodes of the extracted subcodes using CDGTool:
15/20 identical ones. And these ones match the one I extracted two years ago and archived in the rar archive posted below, when I purchased this drive. Then I remember that I performanced the same method and more or less there were 8-9 matches of 10 reads. And for the rawdump project in the 2012's Summer I redid this obtaining newly more or less constant results and this subcode dump was included in the rawdump of this disc.
The 20 extractions including logs, the uncleaned subcodes and the cleaned subcodes. Only P-Q subcodes were cleaned by software, R-W were cleaned 100% by hardware, the special packed mode for CD+G.
https://www.mediafire.com/?8bdanad3dv6qpyj
===
Said this, I will try the new /cdg option of discimagecreator, to check if can provide the same accurate results than this method described. If actually uses 0xBE read command, packed mode, C2 pointers and single sector read commands should offer the same results, being more confortable (no need to enter offset correction, no annoying naming schemes...).