We also used Ethernet on our rocket at PSAS, WiFi for telemetry and
forwarded all data to the internet in real time.
I think I would recommend Ethernet, it's clearly worked for many
launch vehicles. However there are a minefield of gotchya's so you
have to be careful. Assuming you really understand Ethernet and
networks, TCP/IP etc, you can build a robust, high speed, low latency,
extremely high bandwidth communication path between embedded systems
with very cheap COTS parts.
I'll note that our networks were pretty well partitioned, firewall'd, etc.
I've been thinking about how to correctly do cryptographic vehicle
command verification for some time, we might add that in future.
I'm sure the Chinese were *very* interested in live data from our
dinky student-built amateur rocket :p
-Nathan
PSAS
On Mon, Jan 18, 2016 at 11:34 AM, Ben Brockert <wikkit@xxxxxxxxx> wrote:
Xombie and Xoie of the NGLLC ran ethernet between the two computers
onboard. One was hooked up to the GPS and IMU (over serial lines) and
did the GNC magic. It then sent gimbal and thrust level commands in a
packet to the other computer that actually ran all the hardware
(digital out for solenoids, PWM for gimbals and throttles, analog
inputs for transducers, etc). Because everything was on the same
network, the command computer could log the messages from the GNC
computer.
Ethernet then ran from there to a wifi router to the ground, and again
ethernet from the wifi router on the ground to the command computer
and the dGPS base station computer.
Just don't use Netgear for any of those parts, their hardware is shit.
If you want some real craziness, a program that I won't name had a
similar setup but the network wasn't segregated, it was connected to
another really big network that was connected to the internet. So in
theory the vehicle could have live tweeted its own flight, or the
Chinese could have downloaded flight test data in real time.
Ben
On Mon, Jan 18, 2016 at 1:11 PM, Mark C Spiegl <mark.spiegl@xxxxxxxxx> wrote:
Speaking of embedded systems in general:
Im a big fan of connecting smart components using a high speed, fault
resistant serial connection like a LAN. The LAN makes adding new modules
simple, and it makes the wiring more fault resistant than other choices like
a bus.
Creating new communications paths, like Henry described, isn't horrible.
At this point the hard real-time folks run away screaming because the system
is non-deterministic. But I have never worked on a system were a 100M or 1G
LAN connecting components a few inches or a few feet apart didn't smother
the requirement with speed.
Debugging is simplified too. For a few $k you can connect a laptop with its
LAN chip in promiscuous mode and monitor subsystem communications.
Caveat: Ive not built rockets for a living.
--MCS
