Microbolometer

1. Theory

• Definition: A microbolometer is a thermal detector: a thin absorbing membrane, suspended about 2 μm above the readout circuit by thin arms, is heated by the infrared radiation of the scene, and the temperature coefficient of its resistance (TCR) turns the small rise of temperature into a change of resistance that is read by the circuit below it.
• Uncooled operation: It needs no cooling of the detector, whereas cooled infrared imagers need more than 10 minutes to cool down; the response is slow (milliseconds) because the heat has to flow through the arms, and the time constant is the heat capacity divided by the thermal conductance.
• Materials: The two most common thermistor materials are vanadium oxide (VOx) and amorphous silicon (a-Si); the TCR is near -2 %/K in the industry, -2.5 %/K for a-Si in one design, and about -4 %/K for a VWOx film in another; a-Si is deposited below about 200 °C and is compatible with CMOS.
• Vacuum: The array is packaged under vacuum so that the gas does not conduct the heat away; the package pressure is below 10${}^{-2}$ mbar with a long-term target near 10${}^{-3}$ mbar in one study, and a loss of vacuum by outgassing is the dominant failure of the pixels in that study.
• Absorber: The membrane and the reflector below it form a quarter-wave cavity that peaks near 10 μm, so that the device works in the 8 μm to 14 μm band, and the absorptance is above 85% in a measured design.
• Pixel and format: Pixels of 45 μm (older devices) shrank to 12 μm in the 2000s and 6 μm in devices of 2025; formats of 320×240, 640×480 and 1024×768 are common.
• History: Honeywell developed the technology from the late 1970s under a classified contract, and it was declassified in 1992.
• Noise and sensitivity: Johnson noise, flicker noise and the fluctuation of the temperature set the noise; a noise-equivalent temperature difference (NETD) below 50 mK is reported for 12 μm and 17 μm pixels at F/1.
• Figures of merit: The TCR, the thermal conductance, the time constant, the voltage responsivity, the NEP, the detectivity, the NETD and the pixel pitch.

2. Working Principle

• Absorption and heating: The membrane absorbs the radiation of the scene (7.5 μm to 14 μm), its temperature follows $Cd\Delta T/dt+G\Delta T=\eta P$, and the rise is $\eta P/G$ for slow signals.
• Resistance change: The thermistor changes its resistance by $\alpha \Delta T$, with a negative TCR for the semiconducting films, and the change is turned into a voltage or a current by a bias.
• Readout: In the array of one source the readout circuit integrates the difference of the current of an active bolometer and of a thermally shorted one; this removes the common part of the current and the effect of the temperature of the substrate.
• Thermal isolation: The long thin arms give the low conductance (20 nW/K in the best measured pixel of 12 μm), and the radiation of the faces sets a floor of about 2 nW/K for this pitch.
• Frame time: One source gives a rule of $1/(3\tau )$ for the frame rate, so that a time constant of a few milliseconds allows video rates.

3. Applications

• Thermal imaging cameras: Uncooled long-wave infrared cameras and cores in the 7.5 μm to 14 μm band, with 640×512 or other formats and no cooler.
• Systems without a cooler: The 12 μm detector of one company operates from -40 °C to +71 °C without a thermoelectric cooler, with a power of 300 mW or less at 30 Hz.
• Defence and civil use: The technology started in defence; the sources do not list the civil uses, which are a background statement.

Heating, Resistance and Responsivity

Name	Formula	Related Formulae	Variables/Constants
Heating of the pixel	$C\frac{d\Delta T}{dt}+G\Delta T=\eta P$	$\Delta T=\frac{\eta P}{G\sqrt{1+{\omega }^2{\tau }^2}}$ $\tau =\frac{C}{G}$	$C:heatcapacityofthepixel$ $\Delta T:riseofitstemperature$ $G:thermalconductancetothesubstrate$ $\eta :absorptance$ $P:incidentpower$ $\omega :angularfrequencyofthemodulation$ $\tau :thermaltimeconstant$ $\begin{array}{r}(\eta =0.64,P=100nWandG=2.7\times 10^{-6}W/K\\ give\Delta T=23.7mK,andC=5\times 10^{-10}J/K\\ gives\tau =185\mu \mathrm{s},thesourceprints23.7\\ mKand184,computed)\end{array}$
Thermistor and temperature coefficient of resistance	$\alpha =\frac{1}{R}\frac{dR}{dT}$	$R\left(T\right)=R_0\exp \left(\frac{E_a}{k_{B}T}\right)$ $\alpha =-\frac{E_a}{k_B{T}^2}$	$\alpha :temperaturecoefficientofresistance(TCR)$ $R:resistanceofthethermistor$ $R_0:prefactor$ $E_a:activationenergy$ $k_B:Boltzmannconstant$ $T:temperature$ $\begin{array}{r}(at300K,\alpha =-2\%/Kneeds{E}_a=0.155eV,\\ -2.5\%/Kneeds0.19eVand-4\%/Kneeds\\ 0.31eV;0.25eVgives-3.2\%/K,computed)\end{array}$
Voltage responsivity	${\Re }_V=\frac{\eta V_B\alpha }{4G\sqrt{1+{\omega }^2{\tau }^2}}$	${\Re }_V=\frac{\eta \beta I_b\alpha R}{G\sqrt{1+{\omega }^2{\tau }^2}}$ ${\Re }_V=\frac{\Delta V}{P}$	${\Re }_V:voltageresponsivity$ $\begin{array}{r}{V}_B:biasvoltageofadividerwhoseload\\ equalstheresistanceofthebolometer\\ (thefactor4)\end{array}$ $I_b:biascurrent$ $\beta :fillfactor$ $R:resistance$ $\Delta V:changeoftheoutputvoltage$ $\begin{array}{r}P:incidentpower,thefirstformisfor\\ thedividerandthesecondforacurrent\\ bias\end{array}$ $\begin{array}{r}(V_B=1V,\alpha =0.02,\eta =0.64andG=\\ 2.7\times 10^{-6}W/Kgive1185V/Wwiththefactor\\ 4and4741V/Wwithoutit,against1184in\\ thesource;ameasured290\mu \mathrm{V}for726nW\\ is399V/W,computed)\end{array}$
Thermal conductance	$G=G_{arms}+G_{rad}+G_{gas}$	$G_{arms}=2{\sum }_k\frac{{\sigma }_k{A}_k}{l},G_{rad}=4ϵ{\sigma }_{SB}{A}_{rad}{T}^3$ $\tau =\frac{C}{G},f_{frame}=\frac{1}{3\tau }$	$G_{arms}:conductionthroughthetwoarms$ $G_{rad}:radiationfromthefaces$ $\begin{array}{r}{G}_{gas}:conductionoftheresidualgas\\ (negligibleinvacuum)\end{array}$ ${\sigma }_k:thermalconductivityofthelayerk$ $A_k:itscross-section$ $l:lengthofthearm$ $ϵ:emissivity$ ${\sigma }_{SB}:Stefan-Boltzmannconstant$ $A_{rad}:areaoftheradiatingfaces$ $\begin{array}{r}{f}_{frame}:frameratethattheresponseallows\\ (theruleofonesource)\end{array}$ $\begin{array}{r}(ϵ=1(assumed)andbothfacesgive{G}_{rad}=\\ 15nW/Kfor35\mu \mathrm{m},3.5nW/Kfor17\mu \mathrm{m}and\\ 1.8nW/Kfor12\mu \mathrm{m},computed;thelowest\\ measuredGofa12\mu \mathrm{m}pixelis20nW/K)\end{array}$

Noise, Detectivity and NETD

Name	Formula	Related Formulae	Variables/Constants
Johnson noise	$V_J=\sqrt{4k_{B}TR\Delta f}$	$NE{P}_J=\frac{V_J}{{\Re }_V}$ $\Delta f=\frac{1}{2t_I}$	$V_J:Johnsonnoisevoltage$ $k_B:Boltzmannconstant$ $T:temperature$ $R:resistanceofthebolometer$ $\Delta f:noisebandwidth$ $t_I:integrationtime$ $\begin{array}{r}NE{P}_J:noise-equivalentpowerofthe\\ Johnsonnoise\end{array}$ $\begin{array}{r}(R=250k\Omega gives64nV/\sqrt{\mathrm{Hz}}andNE{P}_J=\\ 5.4\times 10^{-11}W/\sqrt{\mathrm{Hz}}at1184V/W;R=91.5k\Omega in\\ 17Hzgives160nV,againstabout200nV\\ inthesource,computed)\end{array}$
Temperature-fluctuation noise	$\Delta T_{rms}=\sqrt{\frac{k_B{T}^2}{C}}$	$NE{P}_{TF}=\sqrt{4k_B{T}^{2}G\Delta f}$ $NE{P}_T=\sqrt{NE{P}_J^2+NE{P}_{TF}^2}$	$\begin{array}{r}\Delta T_{rms}:rmsfluctuationofthetemperature\\ ofthepixel\end{array}$ $C:heatcapacity$ $\begin{array}{r}NE{P}_{TF}:noise-equivalentpowerofthe\\ exchangeofheatwiththesubstrate\\ (standardform,notcopiedfromasource)\end{array}$ $NE{P}_T:totalofthetwoinquadrature$ $k_B:Boltzmannconstant$ $\begin{array}{r}(C=5\times 10^{-10}J/Kgives50\mu \mathrm{K};G=2.7\times 10^{-6}\\ W/Kgives3.7\times 10^{-12}W/\sqrt{\mathrm{Hz}}andG=20nW/K\\ gives3.2\times 10^{-13},computed;thesource\\ prints5.7\times 10^{-12}andatotalof5.5\times 10^{-11})\end{array}$
Flicker noise and total noise	$V_T^2=V_J^2+V_{1/f}^2+V_{TF}^2$	$V_{1/f}^2=K_{1/f}\ln \left(\frac{f_2}{f_1}\right)\left(\mathrm{form}\mathrm{unclear}\right)$ $f_2=\frac{1}{2t_I}$	$V_T:totalnoisevoltage$ $V_{1/f}:flickernoise$ $V_{TF}:voltagenoiseofthetemperaturefluctuation$ $K_{1/f}:coefficientoftheflickernoise$ $f_1,f_2:limitsoftheband$ $\begin{array}{r}{t}_I:integrationtime,thelogarithmicform\\ of{V}_{1/f}ismyreadingofagarbledequation\end{array}$ $\begin{array}{r}(K_{1/f}=10^{-11}isthevalueofthedesign\\ study;thetemperaturenoiseisstatedto\\ bemuchsmallerthantheJohnsonandthe\\ flickernoise;ahighermodulation\\ frequencylowerstheflickernoise)\end{array}$
Detectivity	$D^{\ast }=\frac{\sqrt{A_D\Delta f}}{NEP}$	$NEP=\frac{V_n}{{\Re }_V}$ $D^{\ast }=\frac{{\Re }_V\sqrt{A_D\Delta f}}{V_n}$	$D^{\ast }:specificdetectivity$ $A_D:areaofthedetector$ $\Delta f:bandwidth$ $NEP:noise-equivalentpower$ $V_n:noisevoltage$ ${\Re }_V:voltageresponsivity$ $\begin{array}{r}(A_D=35\times 35\mu {\mathrm{m}}^{2}andNEP=5.5\times 10^{-11}give\\ 6.4\times 10^7\mathrm{cm}\mathrm{H}{\mathrm{z}}^{1/2}/W;themeasuredNEPof\\ 2.13\times 10^{-10}and{D}^{\ast }=3.86\times 10^{8}need{A}_D=\\ 6.8\times 10^{-3}\mathrm{c}{\mathrm{m}}^2,anareaof820\mu \mathrm{m}square,all\\ computed)\end{array}$
Noise-equivalent temperature difference	$NETD=\frac{4\left(F^2+1\right)V_T}{{\tau }_0{A}_D{\Re }_V{\left(\Delta P/\Delta T\right)}_{{\lambda }_1-{\lambda }_2}}$	$NETD=\frac{4F^2{V}_T}{{\tau }_0{A}_D{\Re }_V\left(\Delta P/\Delta T\right)}$ $NETD=\frac{V_T}{\partial V/\partial T_{scene}}$	$\begin{array}{r}NETD:temperaturedifferenceofthescene\\ thatgivesasignalequaltothenoise\end{array}$ $F:F-numberoftheoptics$ $V_T:totalnoisevoltage$ ${\tau }_0:transmittanceoftheoptics$ $A_D:areaofthedetector$ $\begin{array}{r}{\Re }_V:responsivity,\Delta P/\Delta T:changeofthe\\ emittedpowerperunitareawiththe\\ temperatureofthesceneintheband,the\\ firstformisthethesisandthesecond\\ theconferencepaper\end{array}$ $\begin{array}{r}(\Delta P/\Delta T=2.62\times 10^{-4}W/c{\mathrm{m}}^2\mathrm{K}for8to14\mu \mathrm{m}\\ at295K;withF=1,{\tau }_0=1(assumed)and\\ 12\mu \mathrm{m},50mKneedsNEP=V_T/{\Re }_V=2.4pWin\\ thefirstformand4.7pWinthesecond,\\ computed)\end{array}$

Response Time, Absorber and Scaling

Name	Formula	Related Formulae	Variables/Constants
Time constant and frame rate	$f_{frame}=\frac{1}{3\tau }$	${\tau }_{max}=\frac{1}{3f_{frame}}$ $\tau =\frac{C}{G}$	$\begin{array}{r}{f}_{frame}:frameratethatthethermalresponse\\ allows(thefactor3istheruleofone\\ source)\end{array}$ $\tau :thermaltimeconstant$ $C:heatcapacity$ $G:thermalconductance$ $\begin{array}{r}(\tau <7msgives48Hz,\tau <2msgives167\\ Hzand185\mu \mathrm{s}gives1.8kHz;60Hzneeds\tau \\ <5.6msand30Hzneeds\tau <11ms,\\ computed)\end{array}$
Quarter-wave absorber cavity	$d=\frac{{\lambda }_0}{4n_{gap}}$	${\lambda }_0\approx 10\mu \mathrm{m}$ ${\eta }_{peak}\to 1\left(\mathrm{ideal}\mathrm{cavity}\right)$	$\begin{array}{r}d:heightofthegapbetweentheabsorbing\\ membraneandthereflectoronthereadout\\ circuit\end{array}$ ${\lambda }_0:wavelengthofthepeakabsorption$ $n_{gap}:refractiveindexofthegap(1forvacuum)$ $\begin{array}{r}{\eta }_{peak}:absorptanceatthepeak,theequation\\ isthestandardquarter-waveruleandwas\\ notcopiedfromasource\end{array}$ $\begin{array}{r}({\lambda }_0=10\mu \mathrm{m}givesd=2.5\mu \mathrm{m},the\\ simulatedgapofthethesis;itstargetof\\ 2.2\mu \mathrm{m}isthequarterof8.8\mu \mathrm{m};the\\ absorptanceisabove85\%measuredand\\ 95\%to98\%simulated,computedford)\end{array}$
Scaling with the pitch	$\frac{NET{D}_2}{NET{D}_1}=\frac{A_{D1}{\Re }_1{V}_{T2}}{A_{D2}{\Re }_2{V}_{T1}}$	$\frac{A_{D2}}{A_{D1}}={\left(\frac{p_2}{p_1}\right)}^2$ $\frac{NET{D}_2}{NET{D}_1}={\left(\frac{p_1}{p_2}\right)}^2\left(\Re ,V_T\mathrm{fixed}\right)$	$NET{D}_{1,2}:valuesfortwopixels$ $A_D:detectorarea$ $\Re :responsivity$ $V_T:totalnoise$ $\begin{array}{r}p:pitchofthepixel,thelastformholds\\ onlyiftheresponsivityandthenoisedo\\ notchange,whichisnotthecasein\\ practice\end{array}$ $\begin{array}{r}(17to12\mu \mathrm{m}isanarearatioof0.50,45\\ to12\mu \mathrm{m}is0.071and12to6\mu \mathrm{m}is0.25,\\ computed)\end{array}$

Important Questions

Ques: Which numbers of the design paper are reproduced, and which are not?

With $\eta =0.64$, 100 nW and $G=2.7\mu W/K$ I obtain 23.7 mK and, for $C=5\times 10^{-10}$ J/K, 185 μs, as printed. The responsivity 1185 V/W (1184 printed) needs the factor 4 of the divider; without it, 4741 V/W (computed). The Johnson NEP of 5.4×10${}^{-11}$ is reproduced, but in W/√Hz and not in W. The resistance change of 474 mΩ is not: $250\mathrm{k\Omega }\times 0.02\times 23.7\mathrm{mK}=118$ Ω (computed), and 474 is four times that. The fluctuation NEP (5.7×10${}^{-12}$ printed) I get as 3.7×10${}^{-12}$ (computed). The abstract’s 1×10${}^{-11}$ and the conclusion’s $D^{\ast }\approx 4\times 10^{10}$ disagree with the body, which has 5.5×10${}^{-11}$ and a table maximum of 3.86×10${}^8$ ($\mathrm{cm}\mathrm{H}{\mathrm{z}}^{1/2}/W$). Reference: University of Central Florida, design of a VO2/NiCr microbolometer (DSS 2014).

Ques: Is the measured responsivity of the pixels consistent with the prediction?

No. The design study predicts 1184 V/W, but a measured signal of 290 μV for 726 nW is 399 V/W (computed), a third; the best NEP is 2.13×10${}^{-10}$ against 5.5×10${}^{-11}$ predicted. A later device with a VOx-Au film and nichrome arms gave 1399 V/W at 35 Hz and 1 V with an absorber and 817 V/W without it, a ratio of 1.71 (computed); its noise of 0.67 μV is about four times the 160 nV that I compute for 91.5 kΩ in 17 Hz, while the paper quotes 200 nV and ‘2 to 3 times’. The absorptance used for the 726 nW is not given, so the gap cannot be assigned. Reference: University of Central Florida, design of a VO2/NiCr microbolometer (DSS 2014); University of Central Florida, microbolometers with a VOx-Au film and a metamaterial absorber (DSS 2016).

Ques: What TCR values do the sources give, and what do they imply for the activation energy?

The sources give -2 %/K (Wikipedia and the design study), -2.5 %/K for a-Si and about -4 %/K for VWOx (the thesis, at optimum resistance). With $\alpha =-E_a/k_B{T}^2$ these are 0.155, 0.19 and 0.31 eV at 300 K (computed), and the 0.25 eV used in the simulation of the thesis gives -3.2 %/K (computed). One paper states that the activation energy is half of the band gap; for a band gap of 1.12 eV that would be 0.56 eV and -7.2 %/K (computed), so the rule does not fit the amorphous film, whose conduction is not intrinsic, and I did not use it. The review gives no TCR for VOx. Reference: Wikipedia, Microbolometer; Gupta et al., Microbolometer, Defence Science Journal (2013); Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); University of Central Florida, design of a VO2/NiCr microbolometer (DSS 2014).

Ques: Why do the sources give different forms of the NETD?

The thesis writes $4(F^2+1)V_T/({\tau }_0{A}_D\Re \Delta P/\Delta T)$ and the conference paper $4F^2{V}_T/(\dots )$; at $F=1$ they differ by a factor of 2 (8 against 4), and at $F=2$ by 1.25 (computed). For a Lambertian scene the power on the pixel scales as ${\sin }^2\theta =1/(1+4F^2)$, which gives the factor $1+4F^2$ (5 at $F=1$); this is my derivation and not a statement of the sources. The reconstruction of both forms from garbled text is not certain, and the symbol $\Delta P/\Delta T$ is a power per unit area, 2.62×10${}^{-4}$ W/(cm${}^2$K) for 8 μm to 14 μm at 295 K, so the formulas also need the area of the pixel. Reference: Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); Microbolometer package vacuum and reliability, ESREF 2015, session E2, paper 3; Gupta et al., Microbolometer, Defence Science Journal (2013).

Ques: How low can the thermal conductance be, and what does the vacuum do?

The radiation from both faces sets a floor of $8ϵ{\sigma }_{SB}{p}^2{T}^3$: 15 nW/K for 35 μm, 3.5 for 17 μm and 1.8 nW/K for 12 μm (computed, $ϵ=1$ assumed). The thesis measured 20 nW/K for a 12 μm pixel, eleven times the floor, and simulated 25.8 nW/K. The design study printed a radiative conductance of 30 nW/K for 35 μm, twice my value for two faces. The packages are held below 10${}^{-2}$ mbar, while the later device was measured at 20 mTorr, which is 0.027 mbar (computed), higher than that limit; the paper gives no gas conductance. Outgassing at 150 °C raised the pressure and lowered the responsivity by 22.5% with the TCR down only 2.8%. Reference: Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); University of Central Florida, design of a VO2/NiCr microbolometer (DSS 2014); University of Central Florida, microbolometers with a VOx-Au film and a metamaterial absorber (DSS 2016); Microbolometer package vacuum and reliability, ESREF 2015, session E2, paper 3.

Ques: Is a time constant of 7 ms compatible with a frame rate of 60 Hz?

The 12 μm detector is specified with a time constant below 7 ms, and its NETD below 50 mK at F/1 and 60 Hz. With the rule $f=1/(3\tau )$ of the thesis, 7 ms allows only 48 Hz, and 60 Hz needs $\tau <5.6$ ms (computed). The specification is a bound, so the true value may be shorter, and the rule is one source’s, not a law. The thesis measured under 2 ms (167 Hz by the rule), and the design study 185 μs (computed), while an a-Si design is quoted at about 10 ms. A time constant of 2 ms with 20 nW/K means a heat capacity of 4×10${}^{-11}$ J/K (computed). Reference: Jakobson et al., 12 micron pitch VOx microbolometer, International Image Sensor Workshop (2025), paper P28; Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); Gupta et al., Microbolometer, Defence Science Journal (2013); University of Central Florida, design of a VO2/NiCr microbolometer (DSS 2014).

Ques: How do the NETD of 12 μm and 17 μm pixels compare?

The 12 μm VOx detector has a temporal NETD below 50 mK at F/1, 60 Hz and 7.5 μm to 14 μm, and its noise target of two digital levels of 25 mK is the same 50 mK (computed). A 17 μm product is quoted with a typical sensitivity of 25 mK at F/1, 300 K and 30 Hz; the page does not call it NETD. The area ratio is 0.50 (computed), so a fixed responsivity and noise would double the NETD, but the conditions differ in frame rate and in the definition, and the university devices reached 200 mK. The sources do not explain how the 12 μm detector keeps its figure. Reference: Jakobson et al., 12 micron pitch VOx microbolometer, International Image Sensor Workshop (2025), paper P28; Lynred, SPIE DCS abstract on a-Si and oxide thermistors; Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); University of Central Florida, microbolometers with a VOx-Au film and a metamaterial absorber (DSS 2016).

Ques: How large is the noise of the temperature fluctuation against the others?

For $C=5\times 10^{-10}$ J/K the rms fluctuation of the temperature is $\sqrt{k_B{T}^2/C}=50\mu \mathrm{K}$ (computed), and the equivalent power is $\sqrt{4k_B{T}^{2}G}=3.7\times 10^{-12}W/\sqrt{\mathrm{Hz}}$ for 2.7 μW/K and 3.2×10${}^{-13}$ for 20 nW/K (computed). A NETD of 50 mK at 12 μm needs a total NEP of 2.4 to 4.7 pW in the bandwidth, depending on the form (computed). The review states that the temperature noise is generally much smaller than the Johnson and the flicker noise, which fits these numbers. As extracted, the thesis equation for the fluctuation noise carries $T$ and not $T^2$, probably a misprint or an extraction error, and the flicker term was not given in a usable form. Reference: Gupta et al., Microbolometer, Defence Science Journal (2013); Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); University of Central Florida, design of a VO2/NiCr microbolometer (DSS 2014).

Ques: What does the quarter-wave cavity fix, and how do the sources differ?

The gap $d={\lambda }_0/4n$ is 2.5 μm for a peak at 10 μm in vacuum (computed). The thesis targets 2.2 μm (a peak at 8.8 μm, computed) and simulates 2.5 μm; Wikipedia says about 2 μm. The review states a peak near 10 μm, and the membrane of the thesis, 200 nm of Si${}_3$N${}_4$, reaches 95% to 98% simulated and above 85% measured. The device with an absorber of gold squares raised the responsivity by 1.71 at 35 Hz but the heat capacity from 1.6 to 4.1 nJ/K (2.56, computed), so that the time constant grew from 1.28 to 3.28 ms (computed), in line with the stated factor of 2.5. Reference: Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); Wikipedia, Microbolometer; Gupta et al., Microbolometer, Defence Science Journal (2013); University of Central Florida, microbolometers with a VOx-Au film and a metamaterial absorber (DSS 2016).

Ques: Can the NETD be checked from the array responsivity?

The NETD is the noise divided by the response to the scene, $V_T/(\partial V/\partial T_{scene})$. The thesis gives a mean NETD of 227 mK and a mean responsivity of 7.71 mV/K for one design, which implies a noise of 1.75 mV (computed); with a peak of 203 mK it is 1.57 mV. The responsivity per kelvin of the scene is not the responsivity per watt, since it includes the optics, the area of the pixel and the factor $\Delta P/\Delta T$. The best pixel was reported at 37 kV/W with 473 μV, which implies 13 nW (computed), and the units of the array and of the pixel must not be mixed. Reference: Middle East Technical University, doctoral thesis on VWOx microbolometer arrays (open.metu.edu.tr, handle 11511/93275); Jakobson et al., 12 micron pitch VOx microbolometer, International Image Sensor Workshop (2025), paper P28.

Related pages in this vault: Thermopile Infrared Detector, Colloidal Quantum-Dot Photodetector (CQD), Graphene and 2D-Material Photodetector, PN Photodiode, CCD Pixel (MOS Photogate), Thermal and Electro-Thermal Simulation.