SWIR instrumentation
Compact USB SWIR spectrometers, 500 to 1700 nm, on InGaAs sensors
Short wave infrared spectroscopy has moved out of the instrument room and onto the bench, the belt and the field cart. The reason is a single component: the InGaAs linear array, which sees a band silicon simply cannot reach. This guide walks the optics, the specifications and the applications, with interactive tools at each step so you can test the trade-offs against your own measurement.
01. The detector question
Why InGaAs for 500 to 1700 nm
Traditional silicon detectors perform well up to roughly 1100 nm, then fall off a cliff. Indium gallium arsenide picks up where silicon stops, covering the short wave infrared with high quantum efficiency, low dark current and a response fast enough to sit inside a control loop.
The properties that make InGaAs the dominant detector technology for compact SWIR spectrometers are worth naming individually. Broad spectral coverage, reaching down toward 500 nm in VIS to SWIR variants and out to 1700 nm or beyond. High quantum efficiency, so few photons are wasted. Low dark current, which keeps the signal to noise ratio high without cooling. And a fast response time, which is what makes real time and inline measurement realistic rather than aspirational.
Detector reach explorer
InteractiveSilicon is the default detector in visible and NIR spectrometers, and it is useful up to about 1100 nm. At that point the band gap runs out: quantum efficiency falls steeply as it approaches 1100 nm and reaches zero beyond it. Everything interesting in the SWIR, the water bands at 1450 nm and the C-H overtones at 1660 to 1750 nm, sits outside its reach.
The standard lattice-matched InGaAs composition covers roughly 900 to 1700 nm with high quantum efficiency and low dark current at room temperature. This is the workhorse for SWIR spectroscopy: it captures the first overtone bands of O-H, C-H and N-H, which is where most quantitative chemistry lives.
Increasing the indium fraction pushes the cut-off out to 2200 nm or 2600 nm, which opens the combination bands. The trade is real: extended compositions have far higher dark current and normally need deep thermoelectric cooling, which adds cost, power and bulk to what was meant to be a USB-powered module.
The d-vision from Sphere Photonics extends InGaAs response down toward 500 nm, so a single 512 pixel array carries visible, NIR and SWIR information at once. In practice that removes the second spectrometer, the second fiber and the job of splicing two calibrations together.
The choice of detector material sets the ceiling on everything else. Extending past 1700 nm is possible, but it costs cooling, power and size, and for most overtone chemistry it buys very little.
02. System architecture
Inside a compact USB SWIR spectrometer
A compact SWIR spectrometer folds a classical optical bench into a module small enough to hold, powered and controlled entirely over USB.
Broadband light enters through a fiber or a slit and is separated into its constituent wavelengths by a diffraction grating. Focusing optics then map each wavelength onto its own pixel of a linear InGaAs array, so the instrument measures intensity against position and converts that into a wavelength axis. Because every pixel integrates simultaneously, the entire spectrum is acquired in one exposure, with no scanning and no moving parts. Compact designs benefit from a simple, rigid optical bench and solid state electronics, which is what gives them their stability against vibration and temperature drift.
Walk the optical bench
Tap a numbered point1. Input optics Fiber or slit
A fiber connector or a fixed slit sets what light gets in and at what angles. The slit width is the first term in the resolution budget: narrow it and lines get sharper but photons get scarce, widen it and the reverse. Every downstream component is designed around the image of this slit.
2. Collimating optics Diverging to parallel
A collimating lens turns the diverging cone from the slit into a parallel beam. The grating equation only behaves cleanly for collimated light, so any error here shows up as a broadened, asymmetric line profile across the whole spectrum.
3. Dispersive element Diffraction grating
The grating is the component that decides the instrument’s personality. Groove density sets the angular spread, and therefore the trade between spectral range and resolution. On the d-vision this is the choice that moves resolution between roughly 1.3 nm and 3.1 nm.
4. Focusing optics Spectrum onto pixels
A curved mirror re-images the dispersed slit onto the detector plane, so each wavelength lands on its own pixel. Getting the whole spectrum in focus at once, across a flat sensor, is most of the optical design work in a compact bench.
5. InGaAs linear array 512 pixels
The 512 pixel InGaAs array converts each wavelength band into charge. Because every pixel integrates at the same time, the full spectrum is acquired in one shot: no scanning stage, no moving parts, and no motion artifacts between one end of the spectrum and the other.
6. Readout and USB 16-bit, USB 2.0 / 3.0
A low-noise preamplifier feeds a 16-bit converter, and an FPGA handles timing, sensor control and transfer. USB carries both the data and the power, which is what lets the whole instrument travel in a coat pocket rather than a rack.
03. Performance characteristics
The d-vision spectrometer as a worked example
The d-vision from Sphere Photonics is a useful reference point for what a compact USB SWIR spectrometer delivers today: a cost effective, broadband VIS to SWIR instrument with resolution options from roughly 1 nm to 3 nm depending on grating choice.
Representative specifications
| Spectral range | 500 to 1700 nm, grating dependent |
|---|---|
| Detector | 512 pixel InGaAs array |
| Resolution | ~1.3 to 3.1 nm, wavelength range dependent |
| Integration time | 10 µs to 1 s |
| Interface | USB 2.0 / USB 3.0 |
| AD conversion | 16-bit |
Two of those numbers deserve more than a table row, because they are the ones you will actually trade against each other when specifying a unit. The first is the relationship between spectral span and resolution on a fixed 512 pixel array. The second is integration time, which sets what you can measure and how fast.
Range versus resolution, on 512 pixels
InteractiveFull VIS to SWIR in one acquisition. Visible color, the NIR overtones and the SWIR bands land on the same 512 pixels and share one calibration.
Typical use: Broad survey work, plastics sorting, first-look material screening
Classic SWIR window. Drops the visible to buy back resolution, and still holds every first overtone band worth measuring.
Typical use: Moisture, fat and protein assays, inline process monitoring
Tightened around the O-H and N-H first overtones. Enough resolution to separate bands that overlap on a broadband grating.
Typical use: Water content, amide and protein chemistry, biofluids
The narrowest, sharpest option. Reserve it for a band you already know you care about, because everything outside the window is simply gone.
Typical use: Single-band quantification, laser and fiber characterization
Dispersion is span divided by 512 pixels; optical resolution assumes a slit matched to the pixel pitch. Sphere Photonics quotes roughly 1.3 nm to 3.1 nm depending on grating and wavelength. Axiom Optics can confirm which grating maps to your window.
Integration time, from 10 microseconds to 1 second
InteractiveConveyor speed
At the fastest setting the shutter is open for ten millionths of a second, so a bottle moving under an inspection head is effectively frozen. You need a bright broadband source, because the photon budget per spectrum is tiny.
General bench work
The comfortable default for transmission and reflectance on a decent halogen source. Fast enough to average hundreds of spectra per second and knock the noise down without anyone waiting.
Weak signals
Long enough to pull faint photoluminescence or a heavily attenuated transmission out of the floor. Dark current is now accumulating with the signal, so a fresh dark reference matters more than it did at 1 ms.
Photon starved
The end of the range, for genuinely dark measurements. Expect to subtract a dark spectrum taken at the same integration time and temperature, because the dark term is no longer negligible next to the signal.
04. What the format buys you
Key advantages of the compact USB form factor
These instruments bridge the gap between high performance laboratory spectroscopy and fast, field ready measurement, which makes them unusually versatile for real time monitoring.
Portability and integration
A small footprint allows handheld use or permanent inline mounting, powered by the same cable that carries the data.
Real time measurement
Array based detection acquires the whole spectrum at once, so there is no scan time and no drift between one end of the spectrum and the other.
Robustness
No moving parts means high mechanical stability, which matters as much on a forklift-adjacent production line as it does in a vehicle.
Broad coverage
Visible, NIR and SWIR information arrive in a single device, on a single wavelength calibration.
05. Applications
Where SWIR spectroscopy is used
SWIR provides non-destructive, real time chemical and physical analysis by targeting the O-H, C-H and N-H overtone absorption bands. Because SWIR light penetrates materials more deeply than visible light, and reaches bonds that mid infrared systems need costly technology to measure, it acts as a bridge between laboratory precision and field ready measurement.
Where a 500 to 1700 nm instrument earns its place
Pick a sectorProcess and industrial Real-time chemical monitoring
- Reaction kinetics followed in the vessel rather than in the lab queue
- Blend uniformity checked continuously instead of by grab sample
- Inline quality gates that stop a bad batch at the machine
Because the measurement is optical and takes microseconds, the spectrometer becomes part of the control loop instead of part of the paperwork.
Food and agriculture O-H, C-H and N-H overtones
- Moisture, fat, protein and carbohydrate estimated without a reagent
- Sorting and grading on a belt at production speed
- Foreign object and contamination screening
Water dominates the 1450 nm region, fats and oils sit in the C-H bands near 1200 nm and 1720 nm, and protein shows through N-H near 1500 nm. One spectrum carries all three.
Materials and recycling Deeper penetration than visible light
- Polymer identification for sorting streams, including black-on-black cases
- Composition and additive checks on incoming feedstock
- Cure state and crosslink monitoring in composites
SWIR light goes past the surface, so pigment and printing matter far less than they do to a color camera. See the polymer demonstration further down this page.
Pharma and biomedical Non-destructive and inline
- Raw material identity checks through the bag or the vial
- Blend endpoint and drying endpoint determination
- Biofluid and tissue phantom work in the near infrared window
Nothing is consumed and nothing is prepared, which means the same unit can be measured again later, and the measurement can sit inside a validated process.
Optics and photonics Direct infrared coverage
- Laser diode and fiber laser spectral characterization
- Fiber and component transmission testing
- Photoluminescence from infrared emitters and quantum dots
Silicon spectrometers simply cannot see telecom wavelengths. An InGaAs array covers 1310 nm and 1550 nm directly, with no upconversion or exotic detector.
Environmental Sensitive to organic bonds
- Hydrocarbon detection on water and on soil
- Water content and purity screening in the field
- Vegetation and moisture indices from a handheld head
Field measurement is where USB power earns its keep: a laptop, a fiber and a lamp is the entire kit.
It is worth understanding what actually sits inside the 500 to 1700 nm window before choosing an instrument, because the answer determines whether the measurement is possible at all.
Overtone band map
What the window actually containsOvertones are weak echoes of the fundamental vibrations that live in the mid infrared. That weakness is the feature, not the bug: it lets SWIR light pass millimeters into a sample and come back out, so you measure bulk composition rather than a surface film, without the expensive optics and cooled detectors a mid infrared system demands.
O-H bands
Water, alcohols and hydroxyl groups. The 1400 to 1470 nm feature is the single most useful band in the whole window: it is strong, it is broad and it scales with moisture content in almost any matrix.
C-H bands
Every organic material. Aromatic and aliphatic C-H separate cleanly near 1660 nm and 1720 nm, which is exactly what makes polystyrene distinguishable from polypropylene on a sorting line.
N-H bands
Proteins, amides, amines and urea. The 1490 to 1560 nm band is the basis of protein estimation in grain and dairy, and of amide monitoring in polymer and pharmaceutical chemistry.
06. Case study
Sorting plastics by their C-H overtones
Polymer identification is the clearest demonstration of why the 1600 to 1700 nm region matters, and why it has to be measured rather than inferred.
PETE, polystyrene and polypropylene look almost identical to a visible camera and very similar to each other across most of the SWIR. What separates them is the aromatic C-H first overtone near 1660 to 1690 nm: strong in polystyrene, moderate in PET, absent in polypropylene, which has no aromatic ring. Use the viewer below to isolate each material and then zoom into the region where the decision is actually made.
Polymer spectra viewer
Measured on a d-visionThree commodity polymers, measured on one d-vision spectrometer through the same optical path. Across most of the window they are nearly indistinguishable, all sitting near 80 percent transmittance. Everything that separates them happens in the last hundred nanometers, which is precisely the region a silicon detector cannot see at all.
PETE, polyethylene terephthalate
The aromatic ring in PET puts a clear absorption near 1660 nm, so transmittance falls from roughly 0.80 to about 0.54 and then climbs back toward 0.74 by 1700 nm. That recovery on the long side is the tell that separates PET from polystyrene, whose absorption is deeper and does not recover inside the window.
PS, polystyrene
Polystyrene carries a pendant benzene ring on every repeat unit, so its aromatic C-H overtone is the strongest of the three: transmittance drops to roughly 0.36 near 1680 nm. There is a second, shallower dip near 1640 nm, and the pair of features together makes polystyrene the easiest of the three to call with confidence.
PP, polypropylene
Polypropylene has no aromatic ring at all, only aliphatic C-H. It stays high and flat where the other two collapse, and its own absorption sits above 1700 nm, past the detector cut-off. Look instead at the shallow C-H second overtone dip near 1195 nm, visible in the full view.
The decision region
Zoomed into 1580 to 1700 nm, the sorting logic becomes a simple depth-and-shape test. Deep and still falling at 1700 nm is polystyrene. Deep with a recovery is PET. Barely moving is polypropylene. At roughly 2.3 nm per pixel across the broadband range, each of those features is still sampled by ten or more pixels, which is why the call survives a noisy conveyor.
Traces digitized from the d-vision measurement shown above and re-keyed to the Axiom Optics palette for legibility. Read them as the shape of the result, not as a calibration reference.
The same pattern holds well beyond recycling. Any time two materials differ in bond chemistry rather than in color, a SWIR spectrum separates them where a camera cannot, and it does so without contact, preparation or consumables.
07. Specifying a unit
Five questions to answer before you buy
1. Which bands do you need?
Map your target chemistry to the band map above. If everything you care about sits below 1700 nm, standard InGaAs is the right answer and you avoid the cooling penalty entirely.
2. How narrow is your narrowest feature?
Aim for eight to ten pixels across it. That single number usually decides the grating, and therefore the spectral span you can keep.
3. How much light do you have?
Source brightness, fiber core diameter and sample attenuation set your integration time, which in turn sets your throughput. Weak signals push you toward longer integration and careful dark referencing.
4. Bench, line or field?
Inline mounting raises questions about vibration, ambient light rejection, temperature stability and how often you can realistically take a reference.
5. What happens to the data?
A 16-bit spectrum every few milliseconds adds up quickly. Decide early whether the analysis lives on the host, in a PLC, or somewhere further upstream.
Not sure yet?
That is the normal starting point. Axiom Optics works through these five with customers regularly, and can usually tell within one conversation whether a compact InGaAs unit fits or whether the application needs something else.
08. Frequently asked
Common questions about USB SWIR spectrometers
Why not just use a silicon spectrometer with an extended range?
Silicon runs out of band gap. Quantum efficiency falls away steeply approaching 1100 nm and is effectively zero beyond it, so no amount of gain or integration time recovers a signal that never reached the detector. Every band that makes SWIR worth doing, the O-H first overtone at 1450 nm and the C-H first overtones at 1660 to 1750 nm, sits beyond that wall.
What resolution do I actually need?
Work backwards from the narrowest feature you have to resolve, and aim for at least eight to ten pixels across it. Broad overtone bands are typically 30 to 60 nm wide, which the broadband 500 to 1700 nm configuration handles comfortably at roughly 2.3 nm per pixel. Laser line characterization is the opposite case and pushes you toward the narrowest span available.
Does a USB spectrometer need cooling?
Standard InGaAs covering 900 to 1700 nm runs at room temperature with low enough dark current for most work, which is what makes bus-powered operation possible. Cooling becomes necessary when you extend the cut-off past 1700 nm or run long integration times, and it brings power and size penalties that work against a compact module.
Can it measure through packaging?
Often, yes. SWIR light penetrates further than visible light and is far less bothered by pigment and printing, which is why polymer sorting works on colored and printed material. Highly scattering or metallized packaging is the practical limit. It is worth testing your actual samples rather than reasoning from the material alone.
How does it connect to my existing setup?
Light in through a standard fiber, data and power out through one USB cable. There is no external supply, no scanning stage and no moving parts, so the same instrument can sit on an optical bench in the morning and go out to a production line in the afternoon.
In summary
A convergence worth using
Compact USB SWIR spectrometers based on InGaAs sensors represent a genuine convergence of detector physics, miniaturized optics and digital electronics. By enabling high speed, high sensitivity spectral measurement across 500 to 1700 nm, they are changing how work gets done in manufacturing, agriculture, recycling and biomedical diagnostics. Their combination of portability, robustness and analytical capability makes them an increasingly indispensable tool for real time spectroscopy in both laboratory and field environments.
Talk through your measurement
Axiom Optics supplies the d-vision VIS to SWIR spectrometer from Sphere Photonics alongside a full range of infrared imaging and sensing instruments. Send us the sample, the band and the constraint, and we will tell you what fits.
