After whole weekend of making v1 on a prototype board and whole today of debugging, first try was unsuccessfull. I only had about 15min to test it, but it looks like at 1,5V rise time is about 10us and fall about 1us. I was measuring fall time and it's too fast to measure with atmega at 8MHz. Aditionally my comparator was biased too much and didn't even fire. More testing with better setting will follow next week. Until this time don't use my scheme as final reference.
Also - don't make such complicated projects on a bare prototyping board. At least use board with connected pads.
In such projects oscilloscope is indispensable for validating projects, in this instance 10mhz of bandwidth was absolute minimum.