thanks for the quick reply, I know that this solution works with drscl but not within my code.
I tried to track it down, but gave up very soon. Seems as if AnalyzeWF() which is called by CalibrateTiming() finds to much zero-crossings when it is called the second time.
Time calibration has been changed completely in meantime. With the new V5 boards, we have a new oscillator on the board where on can calibrate each channel individually. This is necessary to obtain a good timing down to a few ps. With the current code the above problem has vanished. We also learned that the time calibration is very stable (less than a ps) over several months, so no need to repeat the calibration over and over again.