During this morning's PSL ISS code change, I tested if restarting h1pslpmc model with the FLOW IIR filter replaced with an EPICS part still closes the shutter, and indeed it did.

Looking at full data DAQ trends at the time of model startup shows what is going on. The FLOW ADC data is downsampled from the 64kHz IOP model to the 32kHz h1psliss model. The downsampling filter starts the data at zero and takes several cycles to settle to the digitized value. As an example the plot below shows an LSC channel at the time we restarted h1lsc yesterday. In the PSL PMC case, since any instantaneous FLOW value outside of the acceptible range causes a shutter close trigger, this will always close the shutter on model startup.

One solution we can think of is to add code in the FLOW_ERR C-file which triggers FLOW_OK to become False only if the flow values are out of range for several consecutive cycles. For a decision on whether the issue is of enough concern to warrent this code change we'll defer to the PSL team.