2021-03-27, 06:31 AM
(2021-03-26, 06:54 PM)seandepagnier Wrote: This is a real concern. Has anyone else noticed this?
What if you delete the signalk connection in opencpn, and instead convert signalk to nmea in the signalk server and use that in opencpn? Same lag?
What if you connect the gps to gpsd (or opencpn directly connects to gps) ?
There are really a lot of ways to route data these days, would be good if you can try several and report the results.
It would be good to determine where this lag is. It's quite possibly just a small logic change to fix.
The other issue is wifi not deterministic and does not have guarenteed latency. This is why I am skeptical of wireless wind sensors for autopilots. It needs a dedicated wifi link if wifi can work, or other RF link where both sides are on realtime operating systems.
I agree about the wifi. I do use it aboard, but I have cat-5 wired all the primary devices together. Gigabit ethernet is so noticeably faster and Pi's already have bandwidth issues with SD cards, so why not remove the wifi bottleneck when there's a better solution?Cat 5 cables are cheap. My handheld devices are the ones that are wifi.
I believe I solved the lagging issue. First of all, I had two sources coming into SignalK. As a result, I had many "Source Priorities" applied to determine which source provided which path, etc. As soon as I removed all those filters and the extraneous source, leaving just the can0 adapter, SignalK processed many more deltas/second while receiving less actual data (I had some stuff off to reduce network flow). The canbus adapter is seriously fast.
The other device I removed from SignalK was a Yacht Devices YDNU USB adapter, which I moved over and hung on the OpenCPN Pi to give OpenCPN a direct feed. OpenCPN does channel data back through to SignalK as well as receive a couple of missing NMEA0183 sentences from SignalK (albeit at a lower OCPN priority). I had to be very careful to not create any loops. So far so good. Route data is sent from OpenCPN back through signalk. I've never found OpenCPN's SignalK source to be very worthwhile. TCP was always faster and more reliable in my experience to date, but I may give it another try just to make sure.
So everything now seems fine. I'm headed out tomorrow to shake down my dual Pi set up for the first time and give WilhelmSK a real shot at "remote control". I'll definitely notice if the latency issue continues. I'll let you know if there's any issues.
Steve