Yes I used smaller cell size, but this didn't solve the problem. (In the backround green polygon, in the front the raster after using the polygon to raster.)
I have some polygons and I try to create special management zones for these. So firstly I use polygon to raster option to create one raster layer, than Boolean layers and some reclassify, after all I use the raster to polygon options. If the first raster layer smaller than the original vector data, this causes lack of information and topology problem. So if I use smaller pixel size, the topology error will be greater at the end of the process. It would be good if the raster layer is larger, "tooth style" is ok. Every small 0,15 square millimeter patches would have values. When I use the raster to polygon I can clip it, so it is unimportant.