Bandpass trade - 28 May 2026
LWIR remains the first payload baseline.
The first LWIR, MWIR, and dual-band comparison shows MWIR has much stronger hot-fire contrast, but LWIR-only already clears the first-pass detection gate for the current small-target cases.
Current decision
Keep EmberScope on the uncooled 8-14 um path for the first custom payload.
Model boundary
This is a bandpass detectability check, not a final fire-characterisation payload design.
The calculation reuses the small-hot-target model: Planck band radiance, pixel fill fraction, a low-altitude clear-air transmission proxy, and detector NETD.
The LWIR channel is modelled on the preferred (not yet purchased) Boson+ 640 radiometric detector at 30 mK NETD. MWIR uses a 3-5 um comparison channel at 50 mK NETD. Dual-band SNR is an independent-channel root-sum-square proxy, so it measures detection margin only.
LWIR / MWIR / dual-band results
MWIR improves hot-target contrast, but it does not rescue a failing LWIR case.
| Target class | Target | LWIR SNR | MWIR SNR | Dual-band SNR | MWIR/LWIR | Disposition |
|---|---|---|---|---|---|---|
| Flame proxy | 3.0 cm, 700 K | 6179.7 | 63296.1 | 63597.1 | 10.24 | MWIR adds hot-target contrast |
| Ember proxy | 1.0 cm, 700 K | 686.6 | 7032.9 | 7066.3 | 10.24 | MWIR adds hot-target contrast |
| Smouldering char proxy | 3.0 cm, 430 K | 1220.6 | 2529.9 | 2809.0 | 2.07 | MWIR adds hot-target contrast |
| Weak warm edge | 1.0 cm, 330 K | 18.3 | 14.9 | 23.6 | 0.82 | LWIR sufficient for first detection gate |
Results use the current 120 m, 18 degree HFOV, 640 pixel geometry and a 305 K background. They exclude optical blur, calibration residuals, non-uniformity, smoke, and false-positive thresholds.
Interpretation
The bandpass decision is separated from the detection and classification problems.
LWIR is enough for the first gate
The LWIR-only model clears the first-pass margin for flame, ember, hot char, and the weak warm-edge stress case.
MWIR is still valuable
MWIR carries much stronger colour contrast for very hot subpixel targets and may help later fire-temperature retrieval.
Dual-band is not free
A two-channel payload adds packaging, calibration, alignment, power, procurement, and field-validation work.
False positives remain open
The next decision depends on optical blur, calibration residual, thresholding, smoke, sun-warmed clutter, and field data.
Source use
The trade borrows the mixed-pixel logic, not the satellite operating geometry.
Dozier's subpixel two-temperature formulation motivates comparing target fraction, target temperature, background temperature, and thermal channel response together.
Giglio and Kendall's sensitivity analysis keeps the result cautious: retrieval errors can be driven by background estimates, atmospheric terms, and sensor noise. Giglio and Justice motivate testing wavelength selection before treating LWIR convenience as proof of optimality.
Source records: Dozier 1981, Giglio and Kendall 2001, and Giglio and Justice 2003.