Photogrammetry Based Ranging

This project is one of the more aggressive undertakings I have set out to pursue. Long term, I intend to set up a remote-controlled astronomical observatory on my property.

In the short term, however, I am going to construct two remote-controlled smaller but nevertheless precision telescopes with integrated CCD cameras in order to do range and heading measurements on high-flying aircraft in the air space above where I live. Of course such an activity must be noninvasive (no active sensors like radar or lasers!). Perfecting this instrumentation during daylight hours will serve as a prelude to making LEO and MEO earth satellite measurements at night time for my other orbit calculation interests.

You must look very carefully at the photo to the left in order to see the jet aircraft in the picture. A close-up of this jet is shown in the picture below.

As already mentioned, the measurement instrumentation will consist of two precision telescopes which will be trained on the aircraft overhead. Precision triangulation and image cross-correlation at the pixel level will be used to construct a mathematical model for the plane’s position and velocity vector.

Simultaneously directing two telescopes to track a moving aircraft over head is a challenging task when faced with a finite budget! Calibrating the two telescopes (which will be separated by several hundred yards) could be fairly daunting- if it were not for my interest in astronomy. Calibration of the two optical systems will be done at night by making use of the star fields each telescope will see as a function of its respective azimuth and elevation settings. So long as the telescope mounts provide good pointing repeatability, this approach should solve what would otherwise be a very difficult problem.

Part I write-up

The Part II write-up goes into more detail about the 3-phase direct-drive motor and first machining aspects of the telescope mount.

Part III describes the milling aspects of many of the mechanical aspects of the precision Az-El telescope mount.

I was originally planning to use German equatorial telescope mounts for the project like that shown in the left photograph, but after further consideration have adopted a standard Azimuth-Elevation mount approach. The telescopes (like the one shown to the right) will necessarily have to be fitted with CCD cameras, and the telescope mounts will have precision servo drives involved. The servo drives will come as a natural by-product of my CNC machine work. Wireless control/telemetry will be required from each remote telescope, but I already have Arduino-based WiFi links up and operational with my local LAN. Additional considerations will be required to time-tag and transport the digital images from each telescope of course, so that precision triangulation can be performed, but this part of the project should not be too difficult based upon earlier work in the field.

One of the more intriguing elements of this project is that it entails multiple disciplines. Servo drives will be required for each of the Az-El telescope mounts where my CNC machines will likely come in very handy. The same thing can be said of the CCD camera mounting hardware- where having my own CNC machines should help side-step otherwise potentially pricey hardware. Telescope optics are of course involved. Some elements of control systems will be needed to accurately steer the telescopes autonomously (as will later be used for my satellite tracking project). And there are of course other software elements (Arduino as well as desktop C/C++, and C#) which will be required as well.

Part IV was focused primarily upon optical encoders, gear/pulley arrangements, and additional milling activities. Use of the Broadcom AEAT9000 encoder proved to be challenging from an alignment perspective. Better news is already anticipated, however, for Part V.

Part V concluded the design and fabrication of the AEAT-9000 optical encoder mount. It took three design iterations to obtain the desired results, but it was well worth the effort. On the encoder software support side, I made the transition from the Arduino Mega2560 over to the Texas Instruments TMS320F28379D DSP. The capabilities of this single-chip device are simply remarkable!

Part VIA discusses the remainder of the hardware design less a few items like mounting the power supply, DSP, etc. A finder telescope was designed and fabricated along with a short-term tripod to permit using the telescope outside my workshop. In addition, a RaspberryPi plus HQ video camera was married to the main telescope and the optics needed for this step completed. A new harmonic drive assembly was adopted for the elevation axis.

Part VIAB completes the remainder of the hardware design, including all of the electronics involved. As a related side project, a discussion about my own anodizing setup can be found here.

Part VII is likely the last update I will provide on this project for now. The path forward is briefly described in the update. A short video showing the elevation axis in motion is provided here. This axis relies upon a combination belt-drive 5:1 reduction followed by an 80:1 harmonic drive reduction. The small motor is a 3-phase variety, all controlled by the Texas Instruments TMS320F28379D processor which is hosted in the open electronics box just to the right in the video. The short video is behind this link.

Rather dated details are described in Part 1 of my GoTo telescope mount write-up. Additional details can be found in Part 2 . I’ve modified the direction for this project significantly as briefly introduced below.

Re-Design

My re-design effort is intended to be lower cost than the original approach while also providing both an equatorial as well as az-el solution. Since I plan to use the night-time starfield to calibrate and assess the performance of the mount, it only made sense to include an equatorial capability as well. The latest write-up on this re-design effort can be found here.

This design uses a combination of a 23:1 friction drive reduction ratio and a 30-tooth worm gear rather than a full direct-drive + harmonic drive in order to primarily reduce cost. The original design used 18-bit optical encoders on the two axes whereas this design uses 12-bit magnetic encoders on the worm gear axes which translates to about 16.57 bits at the mount axes. Much finer resolution is actually achieved once in track-mode by virtue of using genuine 3-phase servo motors on each axis.

The most significant change with the re-design is completely moving away from a TI TMS320F28379D digital signal processor as a slave to a PC-type processor. As I looked at the complete end-game, continuing with the Texas Instruments device just made no sense. The chip design was on the order of 10 years old, quite slow at 200 MHz maximum, and there was no route to ultimately fold in real-time video capturing short of adding yet another processor. So in December, 2025, I looked at shifting over to a Raspberry Pi 5 (RP5) (for general number crunching, image recognition, graphical user interface, etc.) working with a Teensy 4.1 slave microprocessor which is perfectly tailored for doing all of the real-time motor control etc. The programming was so much easier (!!!) than working with the TMS320 device, and the computational horsepower was incredible. Here is a picture of the main circuit board as of May, 2026. A complete write-up is forthcoming, but until then, the connectors support a GPS receiver, rotational encoders for both axes, motor-controller outputs for both axes, optional extra SPI port, and a port for a real-time manual joystick for pointing the scope. The RP5 has complete support for two independent video streams (!) which will serve one wide-angle view (for acquisition) and one narrow-angle view from the main telescope.