add informative note from PhilB

src/mame/drivers/qix.c

@@ -30,8 +30,16 @@
           decrypt it, derived by comparison with the not encrypted versions.
           The kram3 set used to be playable with the implementation in the MAME M6809
           CPU core, encrypting only the first byte in 10 xx and 11 xx opcodes.
-          This should get a cleaner implementation, such as a derived CPU class or a
-          callback on when it fetches opcodes. Until then, kram3 is broken on purpose.
+          This should get a cleaner implementation. Until then, kram3 is broken on purpose.
+          
+          According to the QIX and Kram schematics, these games should be using 68A90Es.
+          The 6809E has a 'Last Instruction Cycle' pin that is likely tied in with the encryption:
+          "LIC is HIGH during the last cycle of every instruction and its transition from
+          HIGH to LOW will indicate that the first byte of an opcode will be latched at
+          the end of the present bus cycle"
+          Personally I'd expose the LIC output in a similar manner that was done for the 6502
+          sync output and handle the decryption in the driver.
+
         * I applied the interleave change only to elecyoyo because these games are
           really sensitive to timing, and kram's service mode was disturbed by it (the
           high score and audit pages were broken).