Quick notice: my first guess of having to do with 16-bit LUT was wrong. The tile in question uses the 8-bit type for elevation data, and ele2png produces simple 10-bit LUT (the "standard" colormapping) for those tiles.
I think it can't be actual error accumulation either, because some data-points are completely wrong whereas others are simple off-by-one errors. Might be a systematical error that I never noticed due to mostly working with -16 type elevation data.
---------- Post added at 15:52 ---------- Previous post was at 13:00 ----------
Fixed the round-trip bug with type 8 elevation tiles: http://www.snoopie.at/face/beta/ot3.zip .
I think it can't be actual error accumulation either, because some data-points are completely wrong whereas others are simple off-by-one errors. Might be a systematical error that I never noticed due to mostly working with -16 type elevation data.
---------- Post added at 15:52 ---------- Previous post was at 13:00 ----------
Fixed the round-trip bug with type 8 elevation tiles: http://www.snoopie.at/face/beta/ot3.zip .