A/W
Core formulae and principles
Quick reference of typical detector parameters
R is calculated from the η in the table using R = η·λ/1.23984; the ‘BLIP D*’ column is computed in real time by this tool for a 300 K blackbody, a hemispherical field of view (2π) and the tabulated η. At 905 nm and 1550 nm the 300 K background photon flux is almost zero, so the BLIP limit ceases to be binding — devices in these wavebands are governed by Johnson / shot noise rather than by the background limit.
| Detector | Spectral response band | R @ λ (typical η) | Typical NEP | Typical D* | BLIP D*(300 K、2π) |
|---|---|---|---|---|---|
| Si PIN | 320–1000 nm | 0.584 A/W @ 905 nm(η≈80%) | ≈1×10⁻¹⁴ W/√Hz | ≈1×10¹³ Jones | 1.3×10¹⁹ (background negligible) |
| InGaAs PIN | 900–1700 nm | 1.000 A/W @ 1550 nm(η≈80%) | ≈1×10⁻¹⁴ W/√Hz | ≈1×10¹³ Jones | 6.2×10¹⁴ (background negligible) |
| InSb photovoltaic | 1–5.5 µm | 2.145 A/W @ 3.8 µm(η≈70%) | ≈1×10⁻¹³ W/√Hz | ≈1×10¹¹ Jones | 3.5×10¹¹(PV) |
| MCT photoconductive | 3–12 µm | 5.646 A/W @ 10 µm(η≈70%) | ≈1×10⁻¹¹ W/√Hz | ≈3×10¹⁰ Jones (room temperature) | 3.0×10¹⁰(PC) |
| Thermopile (thermal detector) | Wavelength-independent | Thermal effect; no QE concept | ≈1×10⁻⁹ W/√Hz | ≈1×10⁸–10⁹ Jones | Not applicable (not a photon detector) |
Worked examples
Example 1: InGaAs at 1550 nm, η = 80%, A = 0.785 mm², i_n = 20 fA/√Hz
- R = η·λ/1.23984 = 0.80 × 1550 / 1239.84 ≈ 1.000 A/W(consistent with the datasheet value of 0.95–1.0)
- NEP = i_n/R = 20 fA/√Hz ÷ 1.000 A/W ≈ 2.00×10⁻¹⁴ W/√Hz(20 fW/√Hz)
- D* = R·√A/i_n = 1.000 × √0.00785 cm² ÷ 20 fA/√Hz ≈ 4.43×10¹² Jones
- Minimum detectable power at 1 kHz bandwidth = NEP·√1000 ≈ 0.63 pW
Example 2: noise budget of a 10 µm MCT photoconductor (Johnson noise dominates)
- R = 0.70 × 10 / 1.23984 ≈ 5.646 A/W
- Johnson: i_J = √(4kT/R_sh) = √(4×1.38×10⁻²³×300/50) ≈18.20 pA/√Hz
- Dark-current shot noise: i_dark = √(2e·I_d) = √(2×1.6×10⁻¹⁹×10⁻⁶) ≈ 0.57 pA/√Hz
- Total noise: √(18.20² + 0.57² + 5.0²) ≈18.89 pA/√Hz, with Johnson noise accounting for 92.9% of the power
- NEP = 18.89 pA/√Hz ÷ 5.646 A/W ≈ 3.35 pW/√Hz;D* ≈ 2.65×10¹⁰ Jones
Example 3: BLIP limit of a 10 µm MCT detector (300 K, hemispherical field of view)
- Background photon flux: Q_B = M_q(0→10 µm, 300 K)·sin²90° ≈4.90×10¹⁷ cm⁻²·s⁻¹(Planck integral)
- BLIP(PC):D* = λ/(2hc)·√(η/Q_B) ≈ 3.01×10¹⁰ Jones
- For photovoltaic (PV) operation: D* = λ/(hc)·√(η/2Q_B) ≈ 4.26×10¹⁰ Jones, exactly √2 times higher
- Measured (with the i_n = 18.89 pA/√Hz from Example 2): D* = 2.65×10¹⁰, which is 88% of the BLIP value — the device is close to the background limit
Example 4: back-calculating quantum efficiency from a datasheet responsivity
- η = 1.23984·R/λ[µm] = 1.23984 × 0.95 / 1.55 ≈ 76.0%
- Si:η = 1.23984 × 0.60 / 0.905 ≈ 82.2%
- Both lie within the typical ranges of their respective materials (InGaAs 70–85%, Si 80–90%), so the datasheet figures are self-consistent
Common misuses
Using D instead of D* to compare devices of different areas
NEP is proportional to √A and D = 1/NEP is inversely proportional to √A, so a large-area device has a large NEP and a small D.Comparisons across devices and areas must use D* (normalised to 1 cm², with D* = D·√A). When selecting a detector, first check whether a quoted ‘detectivity’ is D or D* — the two can differ by a factor of √A (a factor of 10 for a 100-fold difference in area).
Forgetting to take the square root for NEP
The unit of NEP is W/√Hz√Hz, and the minimum detectable power = NEP·√Δf. Treating NEP directly as the threshold for a 1 Hz bandwidth underestimates the real value by 1/√Δf: a factor of 31.6 at 1 kHz and 1000 at 1 MHz. This is the most common order-of-magnitude error in lock-in amplifier / weak-signal measurements.
Rejecting a cold-shielded device using the BLIP value for a 300 K hemispherical field of view
The BLIP limit is determined entirely bythe background temperature and field of view, not by an intrinsic property of the device. A cold shield plus filters can reduce Q_B by several orders of magnitude (at 10 µm, from 4.9×10¹⁷ at 300 K down to 8.6×10¹⁰ at 77 K), raising the BLIP limit substantially. Judging a narrow-field system by wide-field figures seriously underestimates the detector’s performance.
Mixing up the PV and PC BLIP formulae
Photovoltaic (PIN/APD) devices use λ/(hc)·√(η/2Q_B), while photoconductive devices (such as MCT photoconductors) use λ/(2hc)·√(η/Q_B);the two differ by √2. Mixing them up systematically under- or over-estimates the limit. The rule of thumb is simple: a device with a PN junction operating at zero bias is photovoltaic; one requiring a bias and relying on photoconductive gain is photoconductive.
Using the wrong units for λ in R = η·λ/1.24
The λ in this formula must be inµmmicrometres. Substituting nm (1550 instead of 1.55) gives an R that is 1000 times too large; substituting mm makes it three orders of magnitude too small (1/1000). This tool calculates entirely in SI (metres) internally, and the input boxes include unit selectors to avoid such errors.
Counting only the dark current in shot noise
The I in √(2eI) is thetotal DC current: dark current + signal photocurrent. Under strong illumination the signal shot noise overtakes the dark-current shot noise and becomes dominant — this is the physical picture of the shot-noise limit (SNR ∝ √P). The ‘③ Noise synthesis’ mode lists the two terms separately, making it easy to identify the crossover point at which the dominant term changes with optical power.
Frequently asked questions (FAQ)
What is the difference between D* and NEP? Which should I use when selecting a detector?
NEP = i_n/R is the ‘optical power density required to produce unit signal-to-noise ratio’ and depends on detector area; D* = R·√A/i_n is NEP normalised to 1 cm² and 1 Hz, and is independent of area.Use D* to compare different detectors and NEP to estimate the detectable power of your own system (then multiply by √Δf to obtain the actual threshold). Datasheets usually quote both; mind the conventions.
How should the unit of D*, cm·√Hz/W (Jones), be understood?
The Jones is the conventional unit name for D*. Dimensionally, D* = √(A·Δf)/NEP: square root of area (cm) × square root of bandwidth (√Hz) ÷ power (W). ‘The higher the Jones figure the better’: 10¹² Jones means a 1 cm² detector can detect a signal power on the order of 10⁻¹² W at a 1 Hz bandwidth. Thermal detectors are around 10⁸–10⁹, good InGaAs around 10¹³, and cooled MCT can reach 10¹³–10¹⁴.
What does BLIP mean, and why is the PV limit √2 higher than the PC one?
BLIP = Background Limited Infrared Photodetection: the detector noise is determined entirely by the shot noise of the background photons themselves, so D* reaches its upper limit for those background conditions. Photovoltaic devices (PIN/APD) have only the shot noise of the background photons; photoconductive devices also exhibit carrier generation-recombination noise, adding a further noise term under the same background and lowering the limit by exactly √2 (the generation-recombination noise of photoconductive devices lowers the limit by √2 relative to photovoltaic devices; see Kruse, Optical and Infrared Detectors, Springer, 1977).
Why does the field of view affect BLIP, and what exactly does a cold shield do?
The background photon flux incident on the detector follows Q_B ∝ sin²θ (θ being the field-of-view half-angle). Narrowing θ from 90° (full 2π hemisphere) to 30° reduces Q_B to one quarter; narrowing it to 10° leaves only about 3.0%. A cold shield plus a narrowband filter reduces both the field of view and the waveband, so Q_B can fall by several orders of magnitude and the BLIP limit rises accordingly — standard practice in infrared systems. The ‘④ BLIP limit’ mode lets you change θ directly to see the effect.
Why does cooling improve D*?
Two mechanisms: (1) a lower background temperature makes Q_B plunge — at 10 µm from 4.90×10¹⁷ cm⁻²·s⁻¹ at 300 K to 8.61×10¹⁰ cm⁻²·s⁻¹ at 77 K (about 5.7 million times lower) — raising the BLIP limit; (2) the detector’s own dark current and Johnson noise decrease with temperature. For mid- and far-infrared detectors the first mechanism is usually dominant, which is why MCT and InSb detectors are routinely cooled.
Which dominates, Johnson noise or shot noise, and how can I tell quickly?
Look at the orders of magnitude: Johnson noise √(4kT/R_sh) depends only on R_sh and temperature, while shot noise √(2eI) is set by the current. Photovoltaic devices have R_sh in the GΩ range, so Johnson noise is only at the fA level and shot (dark-current) noise or amplifier noise usually dominates; photoconductive devices have R_sh of just tens of ohms, so Johnson noise can reach the pA–nA range and often dominates (92.9% in this tool’s default example). The bar chart in ‘③ Noise synthesis’ gives the power fraction of each source directly.
Can quantum efficiency exceed 100%?
Not in a detector without internal gain: one photon can generate at most one electron–hole pair. If back-calculating from R gives η > 100%, check three things: whether R is in mA/W (a factor of 1000 difference), whether λ is in µm, and whether the device is an APD (avalanche gain means the ‘responsivity’ includes gain, so the resulting η is a gain-equivalent value rather than the material QE).
Why is the D* of thermal detectors (thermopiles, pyroelectric detectors) so much lower than that of photon detectors?
The mechanisms differ: photon detectors convert photons directly into carriers and respond quickly (ns–µs), with D* reaching 10¹¹–10¹³; thermal detectors first absorb heat and rise in temperature and then measure the temperature change, responding slowly (on the ms scale) and with a high thermal-noise floor, so D* is typically only 10⁸–10⁹. The advantage of thermal detectors is a wavelength-independent response (covering the whole infrared range without cooling), making them suitable for broadband radiometry rather than weak-signal detection.
I cannot reproduce the D* quoted in a datasheet — what should I do?
First align the conventions: (1) field of view (datasheets often assume a 2π hemisphere or a specific f-number); (2) background temperature (300 K or a cooled temperature); (3) whether a cold shield / filter was used; (4) measurement bandwidth and frequency (1/f noise raises the measured value at low frequencies). Use this tool’s ‘④ BLIP limit’ mode to recompute the limit with the datasheet’s conventions, then check whether the measured value falls within it — as long as D* ≤ D*_BLIP the result is physically self-consistent.
What is the relationship between this tool and the site’swavelength / frequency / photon energy converter?
The site’swavelength / frequency / photon energy converterperformspoint-value conversion (1064 nm ↔ 281.76 THz ↔ 1.165 eV); the photon energy needed by this tool is just one part of that. This tool’s focus isdevice metrics: obtaining R from λ and η, NEP/D* from noise, and the BLIP limit from the background. Used together: first confirm the wavelength convention with the point-value converter, then calculate device metrics with this tool.
Related tools
- Overview of optical calculation tools — the entry point for all calculators.
- Wavelength / Frequency / Photon Energy Converter — point-value conversion to confirm the wavelength convention.
References and data sources
- R. J. Keyes (ed.), Optical and Infrared Detectors, Springer, 1977 — the chapter by P. W. Kruse gives the standard derivation of the PV / PC BLIP limits and the √2 factor for generation-recombination noise.
- E. L. Dereniak & D. G. Crowe, Optical Radiation Detectors, Wiley, 1984.
- CODATA 2018 recommended values: c = 299792458 m/s, h = 6.62607015×10⁻³⁴ J·s, e = 1.602176634×10⁻¹⁹ C, k = 1.380649×10⁻²³ J/K.
- Typical NEP / D* values are compiled from technical documents and device datasheets from Thorlabs, Newport and others; refer to the specific model’s specification sheet for actual selection.


