Thank you for the reply. I disassembled it. The position detector is indeed optical. I don't see that easily repaired without some replacement parts or at all (at least easily and economically at home, unless the problem is dust on the sensors and/or reflective surfaces - there are two sensors and three reflective sticks [two in a group] to unambiguously distinguishes the four dialer positions, it's a clever an indeed a robust design). But I guess the problem is the USB ribbon cable. Before the disassembling I heard a small clicking-like noise and saw the notification lights rapidly light up for a short moment as I moved the dialer away from the reflective/display position and the USB connection got lost halfway through to the self-calibration position. Now I that I see this ribbon cable in place and I heard what kind of noise it makes when it moves, I guess it was that. And not there is no USB connection anymore, no matter how I move that cable. This should be easily to fix (by finding a similar cable and replacing it). >>ccss/ccmx Ah, now I remember. That's where I got confused. The difference was significantly smaller with a ccss file than with a ccmx. The ccmx essentially aligned the two sensors but the ccss didn't really bring the i1d3 readings too close to the CM readings (I even saw some cases when it moved further away). - However, we talk about dE differenced of about 1-2 or so, thus nothing extreme. This made me assume that either the ccss based correction is too "experimental" and can't be relayed on (the i1d3 was a new thing back then) or the CM isn't really that accurate (probably less accurate than the CM). Since this ccss implementation was a new (not a widely and/or long tested) thing, I chose to go with the first assumption and thus continued to use matrices (created with the CM). I think this CM malfunction was a good reminder that I should reevaluate this.