Jeff, Jim, Dave
The h1calex, h1caley models were changed to add RFM sender IPC parts. The h1calcs model was changed to add two receivers for these channels. When the new code was ran, random single and sometimes double errors were seen at the receiver for these channels at a rate of about one per minute from both end stations.
Working on the X-Arm, our first test was to turn off the h1iscex model, which has 8 RFM senders, to check if the addition of one channel has exceeded a limit. The CAL error rate was unaffected.
We noted that both h1iscex and h1calex are processing for about the same length of time (6uS and 5uS respectively). We thought that therefore the models are writing to the RFM card in h1iscex at roughly the same time. The h1lasex model, on the other hand, is writing to the RFM card at the 33uS mark.
We delayed the execution of h1calex by modifying the PCAL_MASTER.mdl model and adding a bunch of filter modules. Jim created filters to run on these FMs to slow down the model. Our first try did not delay by much (cpu about 10uS) and the error rate was about the same. Our second try delayed too much, to about 33uS and the error rate shot up to hundreds per second. Our third try put the cpu time to 25uS which is what we wanted. Unfortunately the receive error rate was unaffected.
At this point we were considering sending the channel over the the ISC model via shared memory and letting that model RFM the channel to the corner station (similar to what ODC is doing). Jeff decided at this point to remove the new IPCs for now.
To back the change out, we did the following: