PWM signal problem

FvM , here is a screen with timing information - 1ms per division.

http://img48.imageshack.us/img48/2120/pulse2.png

I have checked the osciloscope probe against a low frequency signal generator and it works fine.

I "assume" the attenuation is likely capacitive load sort of high pass filter, but no idea what it could be, because nothing stands in between the pin and probe. Generated 3 khz signal on pin is relativly OK, slightly different code without the prescaler counter.

I also tried to drive a small RC servo and such signal just holds it on spot. model. But slightly smaller size RC servo responds and moves corresponding to the signal.