Found the reason of Read/Write mismatch errors. This happens when Busy is being removed from TML0. This signal is applied to the Clear of the D-triggers L0 and L0p. If there is another trigger then it can fire L0p (metastability problem). Resolution: L0 triggers should be cleared before the Busy gets removed. This flaw did not harmed data because the error was properly handled by advancing the write counter. It may cause corruption of data only if two gates were separated by less then CIP interval=12 us.
FASTBUS maximum speed is ~ 500 Hz and it is defined by FB readout time = 2 ms, Transfer to host takes .5 ms and is done in the background.
With FECs off. The workers in that case generate data = kZeroOffset – pedestals. When this data will be received then inversed transformation is done, therefore result will be zero. If one looks at plane array then it will have kZeroOffset – pedestals or 0.
Pedestals are working now correctly.
Currently the pedestals are written to the first record of every sequence but they are taken from the separate file. More correct way is to send them directly from mercury, but this requires some protocol.