Hugh, Dave:

On 27th February I modified h1oaf to add two RFM senders, one for each arm. This tested that sending the additional channel around the RFM loop did not adversely impact the existing data transfer.

Today Hugh added a receiver for the X-arm channel on h1pemex. The receiver reported continuous errors once all systems were started.

After thinking about this the answer became obvious, the h1oaf model is running way too long to be able to send an RFM signal to the end station. The Timing System reports the latency of sending a signal over the single-mode fiber-optics link to the end station is 29uS. Therefore sending models should ideally be running under 30uS total processing time, which is the case with h1omc (14uS).

To quickly test this, I moved the RFM channel H1:SEI-OAF_2_EX_MUX_RFM from h1oaf to h1pemcs (I hand edited H1.ipc to change the source model name), recompiled and restarted h1oaf and h1pemcs, at which point the receive errors at h1pemex were cleared.

The temporary solution (until we install the long-range Dolphin network) is to use h1pemcs to receive, process and distribute the ground motion data. Any signals which may also be needed by h1oaf can either be read from the dolphin network separately, or sent from h1pemcs via shared memory IPC channels.