As a back-up to the Piksi GPS, which we hope will be our primary system, we selected the u-blox 7 series devices. These are in a similar price range to the Venus638 devices, but hopefully provide lower latency velocity estimates.
It’s taken quite some time for them to become available, and the first modules we found were the NEO-7N from CSGShop. Although the NEO-7N isn’t the most capable module in the series, it has a 10Hz update rate and appears to perform similarly to the higher-end LEA-7 devices, but with less configurability.
The CSGShop order process was straightforward, shipping time was usual for Europe to Australia, and the items were as advertised. No complaints there.
The boards are of reasonable quality, although nowhere near as nice as the Piksi, and the layout (in particular the RF signal path from antenna connector to the shielded u-blox module) appears to be sane. USB and UART/SPI are provided via two JST SH connectors; there’s a solder bridge on the bottom which presumably selects between them. The board has a built-in 3.3V LDO regulator, which is kind of a pain because our I/O boards provide 3.3V already—we’ll most likely remove the LDO from the boards and bridge the power over once we’re done with USB. The UART/SPI signals are presumably 3.3V (I don’t see any level shifters) so no change needed there.
After setting up a JST SH to USB-A cable and attaching a basic GPS antenna, I connected a module to my computer and ran the u-blox u-center configuration and monitoring interface. By hardware vendor standards it’s a capable piece of software, and configuration is straightforward enough if you’ve read their binary protocol format documentation cover-to-cover.
Indoors, next to a window with half the sky obstructed by 14 floors of reinforced concrete, the NEO-7N achieved a 3D fix from cold start within two minutes—for comparison, the Venus638 took 15–45 minutes in the same location with the same antenna, and the Piksi didn’t lock at all (although the production firmware will better).
Once locked, horizontal position accuracy over a half-hour period was generally better than the Venus638, with location estimates varying within a radius of about 50m, and about 8m standard deviation. Five or six SVs were visible at any given time, so results with the 8–10 usually available outside should be good.
We pay basically no attention to GPS altitude, but it’s as bad with the NEO-7N as with any of the other systems we’ve tried—it might stabilise within 10m if you’re lucky, but variation of up to 50m over half an hour is not uncommon.
Velocity accuracy isn’t really something you can test on your couch, but it can’t be any worse than what we were using before.
After bringing up the CPU boards and having enclosures made, we’ll run a comparison of all three GPS units in a test flight. Hopefully by then the Piksi firmware will have been updated, and looking at the progress of the development team there might even be some RTK capability available.