How far from the lab bench can a microscope go and still hold focus?
A field-deployable open-source microscope is really a three-axis robotic stage. This project measures its operating envelope — the range of temperature and vibration over which it still meets its performance targets — then builds and tests fabrication that pushes that boundary outward.
You can't design for field conditions you don't know. So measure the envelope instead.
Normally a field instrument is designed against a known spec — the temperatures, vibration, and handling it must survive. Here that spec doesn't exist. Rather than guess, the project never claims the microscope "works in the field." It makes a measured statement: it meets its targets up to certain conditions, and fails beyond them.
Performance budget
The pass/fail lines that define "still working" — the stage returning within tolerance, focus error inside the depth of field, a scan that completes. Fixed by the imaging task, not the field.
Environmental stress
The conditions deliberately turned up and down on the bench: temperature, its rate of change, and vibration. Dust, humidity, and unreliable power are named and left out, honestly.
Envelope boundary
For each axis, push the stress until a target is crossed and record where. Measured for the stock microscope, then again for the modified one — the gap between them is the contribution.
Anchoring
The measured envelope is compared against rough real-world ranges from published climate data and field reports — used to sanity-check coverage, never as design targets.
Diagnose the problem, fix it, then make it hold up
Three interim deliverables follow one loop. Each produces matching before-and-after numbers using the same procedures, so every later claim is judged against the baseline that came before it.
Baseline the stock stage
Measure repeatability, backlash, focus-axis accuracy, and how motor steps map to micrometres and pixels — the "before" picture every improvement is measured against.
Fix the Z-axis
Design, build, and fit one intervention to the focus axis — a mechanical fix, a position encoder, or both — chosen to address the biggest weakness the baseline revealed.
Compensate & integrate
Instrument the stage with temperature sensors, model how it drifts with heat, cancel most of that drift in software, and prove it with one full automated scan end to end.
The final report ties all three into a single engineering story — not three reports stapled together — and carries the most weight.
Four deliverables, one repository
Every claim is backed by before-and-after evidence and honest error bars; every fabricated part ships with files clear enough that someone else could rebuild it.
Baseline performance of the stock microscope
Send the XY stage to target positions in random order and measure how closely it returns, including direction-dependent backlash. Step Z to find best focus, comparing image-sharpness metrics (Laplacian variance, Brenner, Tenengrad, normalised variance) to see which is most reliable on this hardware. Record the step→mm→pixel conversions.
Focused fabrication intervention on the Z-axis
The Z axis gets the attention because the tightest target is on it. Pick a mechanical fix (anti-backlash preload spring, counterbalance), a sensor fix (linear encoder for an independent position reading), or both — then re-run the ID-1 Z tests on the changed instrument for matching before/after numbers.
Thermal monitoring, compensation & integration
Mount temperature sensors where thermal expansion most moves the stage, log temperature and drift together over a one-hour scan, fit a correction model, and show the corrected drift is smaller than the uncorrected. Closes with a full automated scan proving the three deliverables work as one system.
Final report — the synthesis
One narrative: the question raised by deploying in unknown conditions, what the baseline found, which fix was chosen and why, the evidence it worked, the limits hit along the way, and what should come next — with a complete archive that lets someone else reproduce or build on the work.
Minimum, Target, Maximum — scope that grows
Each endpoint sets how much work the three interim deliverables involve. Target includes everything in Minimum and adds to it; Maximum includes everything in Target and adds more. The final report is required at all three.
Diagnose & fix once
- 30-position repeatability at one site; two sharpness measures across three sites
- One Z-axis fix (mechanical or sensor) with before/after numbers
- Two temperature sensors, one 60-min session, post-hoc correction, one full scan
Add rigour & ablation
- Repeatability at 2–3 sites with cross-axis check; third metric vs z-stack ground truth
- Ablation of a combined fix; statistical tests with error bars
- Third sensor; multiple sessions; multi-input model; objective swap over 5+ cycles
Independent & live
- Independent Z measurement; four metrics in full; image-based vibration breakdown
- Combined mechanical + sensor fix; accelerometer resonances before/after
- Heater-driven test point; live in-scan compensation; an independent-user handoff test
↻ The final report is required at every endpoint — it ties together whatever interim work was completed.
Setup, then the loop, then synthesis
Week one has already passed at the time of the agreement. Weeks 1–2 are reading and procurement (long-lead encoders and sensors are ordered early); the remaining ten weeks run the three deliverables and the final report.
The final report carries the most weight
The three interim deliverables build the engineering record; the final report is where it all comes together. Each interim deliverable uses the same six-part rubric.
Course weighting
Interim rubric (each deliverable)
- Engineering rationale20%
- Technical execution20%
- Quantitative analysis20%
- Evidence-based outcomes20%
- Documentation & reproducibility10%
- Integration & engineering judgment10%
About $590 in parts
Rough estimates from the project's reference documents. Shipping, taxes, purchasing fees, and optional extras are not included.
| Item | Role | Phase | Est. |
|---|---|---|---|
| OpenFlexure motorised kit | Base instrument: 3-axis flexure stage, Pi, camera, software | CORE | $250 |
| USAF 1951 resolution target | Spatial reference & autofocus benchmark subject | ID-1 | $45 |
| Stage micrometer / fiducial grid | Ground truth for XY repeatability | ID-1 | $60 |
| Histology teaching-slide set | Feature-dense focus & integration samples | ID-1 | $60 |
| Blank slides & coverslips | Camera-calibration baseline | ID-1 | $20 |
| Magnetic linear encoder + strip | Independent Z position measurement | ID-2 | $40 |
| Arduino / breakout + cabling | Sensor interfacing & logging | ID-2 | $25 |
| Springs, brackets, fasteners | Anti-backlash preload & counterbalance | ID-2 | $20 |
| 3D-printer filament (PLA, PETG) | Custom brackets & sensor mounts | ID-2/3 | $30 |
| Temperature sensors (DS18B20 ×3) | Multi-point thermal monitoring | ID-3 | $25 |
| MEMS accelerometer (optional) | Instrumented vibration readout | ID-2 MAX | $15 |
| Total estimate | All categories combined | $590 |
Learning objectives covered: precision metrology · experimental design & uncertainty · coordinate frames & transformations · mechanical design for precision · sensor integration · empirical compensation · engineering documentation.