December 2018 - May 2019, for MRover. A spectrometer to read Raman spectra in-situ from a mobile robot.

Problem and Background

University Rover Challenge 2019 was the first year that all sample analysis had to occur on-board the rover, as opposed to on bench-top with retrieved samples. Additionally, the rules specified that analysis of samples that could not be manipulated would be required, necessitating instrumentation that could take measurement directly in-situ, or directly from the sample as it rests in the environment.


In response, the team decided to build a Raman spectrometry instrument due to the high ratio of scientifically salient information to cost and mass usage and ability to take measurements directly on the ground quickly.

The instrument was comprised of the probe and the spectrometer. The probe was responsible for exciting a sample using a laser and return the reflected light to the spectrometer. The spectrometer would then be responsible for reading the distribution of that light across the component wavelengths (i.e. measuring the spectra). I was responsible for designing the spectrometer.

Design Goals

This was the first time the team had ever attempted optical instrument design. As such, there were no quantitative requirements but rather design priorities:

  • Maximize spectral resolution (i.e. how precisely we can tell the wavelength of some photon)
  • Minimize noise due to light leakage and sensor heating
  • Maximize serviceability and tune-ability

Additionally, there was an extremely tight timeline to deliver the product. Development started December 26th, 2018. The first iteration had to be done by February 28th, 2019. The last iteration had to be done my April 28th, 2019.


Iteration 1: 533 nm

Based on fl@c@'s ramanPi design of a Czerny-Turner mirror configuration with a 533 nm excitement wavelength. Validated and optimized in Zemax OpticStudio over the course of 8 days, mechanically designed and drafted over 7 days, manufactured over 18 days, and integrated over 7 days.


Lessons learned from Iteration 1:

  • 533 nm excitement causes a high level of background fluorescence in the sample of interest, drowning out the Raman effect signal
  • Spectral resolution needed to be improved (overlap was unacceptable)
  • Room-temperature-vulcanizing (RTV) rubber is difficult to deal with for the purpose of light sealing
  • Internal radii should be larger than endmill radius to improve surface finish
  • As many important registration and datum features should be machined in the same setup

Iteration 2: 635 nm

A completely original optical layout (still in a Czerny-Turner configuration). Designed and manufactured on a similar timeline to Iteration 1.



This project was a massive undertaking and the results were better than expected from the outset. The work is currently being carried on to be used at the next competition with a more amenable timeline.


  • The spectrometer was well-focused for certain wavelengths
  • Increased CNC machining talent level on team
  • Increased clean integration skills: appropriate handling and storage of delicate optics
  • Light-sealing and reflection minimization was effective


  • Shim-based optical adjustment mechanisms were difficult to use
  • Thermo-electric cooling was not effective: there may have been moisture trapped in the spectrometer that would condense on the sensor during cooling cycles
  • Cooling solution was also over-powered, as it was based on far too conservative thermal resistance estimates
  • Optical coupling introduced difficult-to-debug issues with alignment of optical components
  • Very high noise level

Future work:

  • Remove optical coupling: integrate the probe and the spectrometer and move the sample to the instrument
  • Improve serviceability: optics and adjustment mechanisms should be easily hand-accessible
  • Incorporate shock/impulse tolerance requirements in design
  • Stray-light analysis and minimization