Date: May 14, 2025
Team Members: David Bascom (ddb73), Ajay Sunkara (aks237)
This project seeks to help those who are bad at pointing a telescope themselves enjoy being able to still use there telescope and see distant planets up close. It strives to take away any technical skill or high monitary limit from the anmature astronmers, and let whomever wants have an easy viewing experience of the night sky
- Project Video
- Introduction
- Design & Testing
- Results
- Conclusions
- Future Work
- Budget
- References
- Code Appendix
- Contributions
This project had 3 significant aspects to it:mech anical, electical, and software. The mechanical side involved designing a new mount for a telescope that would be motorized in 2 axes to allow it to point anywhere in the night sky. The main conciderations here were avoiding colisions and ensuring the motors had enough torque. This was acomplished by offsetting the mount for the scope behind the first rotational axis, and implimenting a 4:1 belt drive system.
On the electical side, we needed the ability to drive 2 brushless motors, ensure positional feedback, automatically get GPS location and heading, have a touch screen controller, and power everything. For our motor drivers we used the mjbots moteus c1 driver, which comunicates over CANFD. To suport this off the pi, we also used the mjbots usb-canfd converter. For heading and GPS, we used the Berry-GPS-IMU-4. Our touchscreen was a cheap amazon pi monitor that uses HTMI for signal and USB for power. The pi was powered off a cell phone charger via USB-C, and the motion system is powered using a 6s Lipo battery.
As for the software, we used two librarys: Skyfield and moteus. Upon startup, the mount takes in the current heading, and zeros the first axis to face due north. The second axis is automatically zeroed baced on its position at startup, which it assumes to be level with the ground. When a user selects a target (such as a planet or the moon) from the touch screen GUI, the name of that target, the current GPS location, and the current time are given to Skyfield. Skyfield gives us back degrees in the altitude and azimuth, which we then translate into motor rotations for each axis. We then use moteus to convert those rotations into a CANFD message, and send it over the CAN line
Mechanically, all parts of this project are 3D printed. Linear Rods from an old 3D printer are used to rienforce the prints to make them stronger. Large breakings are also used on both rotational axes to ensure smooth opperation. The design is made in 21 individual 3D prints, and assembled using linear rods and M3 screws. This part of the project went pretty smoothly. Asside from a few print failures, a bit of trouble getting the bearings to sit properly in thier press fit housing, and one of the prints being the wrong size, and requiring a reprint.
Electically, things also went pretty smoothly. Soldering the motors to the boards directly reduced costs for connectors. To power the boards, we made a custom wire harness to split power from the Lipo into two connectors. Similarly, CAN is also split after the converter board. These all required quite a bit of soldering. The pi is simply plugged into the phone charger, and the touch screen into the pi.We had a weird issue where the touch screen would not work with certain USB ports/cables, but this issue was fixed by switching out the USB cable that connected to the screen. The IMU is a pihat, which did require us to remove the piTFT and reconfigure the pi to boot without it. Later, we discovered that the on board GPS antenna did not work, even outdoors, and decided not to use it
Software is definitly where we had the most issues. Querrying data from the IMU was pretty simple to impliment, but it took us quite a while of messing with the software to figure out that we were not the issue, but the hardware was. After figuring this out, we ditched it entirely, and set the telescope up so that it would auto zero in the startup position. Implimenting the GUI went decently smoothly, but took a relatively long time due to the number of targets that we needed to be able to see. This was a large part of the reason why the 7 in touch screen was choosen, as the piTFT was MUCH too small for this. The real chalenge was figuring out what the actual targets are. Skyfield has a number of files you can download that each have differnt targets in them. We wanted to be able to automatically see what the targets in the file were, extract them, turn them into a human readable name, and send that back to the library. We did this so that different files could be loaded and any new targets would still apear on the GUI. To do this, however, we did spend a considerable amount of time figuring out how to extract target names, auto-renaming everything into something that made sense to humans, and automatically, laying out the GUI buttons baced on the number of targets available in a file.
During testing we dicovered several things. The first of which was that a lot of the entries in the skyfield files point to random points in the universe called "baricenters". These are centers of gravitational attraction, and there is nothing there. We ended up having to filter these out of the valid targets that could be displayed on the GUI. Furthermore, we discovered that the the motors we selected we not very precice. While this was not an issue when looking at objects that were large or close by (like the moon), objects that did not fit those criteria were in many cases completely out of the telescope's view. Unfortunatly, this did come down to a physical limitation of the motor, and was not something we could fix
Overall, we were able to meet some of our goals, but missed out on quite a few as well. We were able to make a mount that can automatically point at objects in the night sky without manual aid. However, we were not able to automatically orient the telescope like we initially planned, and we were also not able to see small or distant objects. These issues could have both been fixed given a bit more time and better hardware. However, out GUI was very sucsessful, and was able to automatically extract targets and send them to the motor controler. We were able to pretty reliably look at the moon and the sun (obviously we were not gonna put the telescope on the mount when pointing at the sun). Other targets, we had a lot of trouble viewing reliably, but we were able to track jupiter for a short time at one point
We concluded that while the setup we have is adiquite for viewing the moon and sun, a more precise motion system would be required to see far out targets Brushless motors, or at least the ones we used, do not have the resolution in rotations to be accurate without a larger reduction. We also concluded that a PCB antenna is not strong enough to recieve a reliable GPS signal, and that a real GPS antenna would be needed in order to have the auto-configuration settup actually work. On the plus side, we were able to conclude that using vector math alone for our targeting is enough to get reliably accurate coordinates to see celestical bodies, and that visial or manual feedback is not required for the system to work. While our setup for autocalibration did nit work, were also able to determine that given the correct hardware, this kind of system is possible.
Given more time, we would have made a new version of the mount that changes the motion system to closed loop stepper drivers, and incresed the reduction on the belts. Reducing the size of the belt as well could help to achieve a larger reduction in the same footprint, as they are not under much load. A tripod would have also been a very nice addition, as the mount is very hard to use outside without it. Electically, adding an on/off switch would have been a very nice feature, as plugging in an unplugging the battery to power cycle was both tetious and likely not the safest. A GPS antenna is also a necessitly. Integrating a camera to the settup would have been very cool as well. Initially one of the things we wanted to add if we had time was a sperate tab in the GUI that was a live camera feed, and the ability to take pitures or videos that get stored on the pi.
To avoid budget constraints, most of the parts for this project were repurposed from things that I had laying around, such as an old 3D printer and drone. Preowned parts are marked as so in the table below, and prices/links have been provided to all preowned parts that I was able to find a sorce for.
| Part | Preowned? | Cost per Unit | Units | Total Cost | Link |
|---|---|---|---|---|---|
| Brushless Motors | Yes | ||||
| moteus c1 controller | Yes | 69 | 2 | 138 | Link |
| 608 2RS Ball Bearings | Yes | 4.99 | 1 | 4.99 | Link |
| 6814ZZ Deep Groove Ball Bearings | No | 9.99 | 2 | 19.98 | Link |
| 6s Lipo Battery | No | 21.99 | 1 | 21.99 | Link |
| Portable Charger | Yes | ||||
| Touch Screen | No | 39.99 | 1 | 39.99 | |
| Berry-GPS-IMU | Yes | 79 | 1 | 79 | |
| Linear Rods | Yes | 0 |
See code repository for full listings.
- David: Hardware design (Mechanical and electrical
- Ajay: Software and firmware integration and