Errors in minimum resolvable temperature difference charts

Errors in minimum resolvable temperature difference charts

Infrared Physics, 1977, Vol. 17. pp. 375-379. Pergamon Press. Printed in Great Britain ERRORS IN MINIMUM RESOLVABLE TEMPERATURE DIFFERENCE CHARTS WI...

262KB Sizes 1 Downloads 47 Views

Infrared Physics, 1977, Vol. 17. pp. 375-379. Pergamon Press. Printed in Great Britain

ERRORS

IN MINIMUM RESOLVABLE TEMPERATURE DIFFERENCE CHARTS WILLIAML. WOLFE

Optical

Sciences

Center,

University (Rrceiwd

of Arizona, 14 March

Tucson,

Arizona

85721. U.S.A.

1977)

INTRODUCTION The test chart or target usually employed in assessing the minimum resolvable temperature difference (MRTD) of infrared devices as a function of frequency consists of several groups of four bars. Each group has a different fundamental spatial frequency. The bars have an aspect ratio of 7 : 1. The test target is generally made from two plates operated at slightly different temperatures and located one behind the other. The bars are slots in the front plate. Figure 1 is a schematic diagram of such a device. The sensor then views alternately the front plate and the bars. The output signal-to-noise ratio is recorded as a function of the temperature difference between the rear plate and the front plate. The usual assumption is that the front and rear plates have the same emissivity. For many reasons this assumption is never quite true. For example, there is a reasonable cavity effect causing an enhancement of the effective emissivity of the rear plate, and as a consequence, the bars. We have analyzed the errors that accrue from the fact that these two emissivities are different. The results are functions of the plate temperature, the bar temperature, the background temperature, the two emissivities and the spectral band of operation. THE PHYSICAL MODEL We assume that the front plate has a temperature T and emissivity E; the bar has a temperature T + dT, emissivity E + de; and we assume that the background is uniform in temperature with unity emissivity. The analysis is done first for a system that senses

60

0.009

50 0.007

40

0.005

30 20

0.003

IO 0.001 0 -10 -20 -30 -40 -50 -60 -70 -60 -90 -100 200

250

350

309

Temperature,

400

K

Fig. 1. The relative error in bar charts for the full spectrum of power as a function of temperature for different values of dr/dT and a background temperature of 310K. 375

376

WILLI4M

L.

WOLFI

radiation power of all wavelengths equally. This shows the basic principles. A similar expression for a detector that responds equally to photons of all wavelengths is easily obtained from this. Next the more complicated analysis for monochromatic radiation is performed. Interestingly the error for the monochromatic case is independent of whether a power distribution or photon distribution is used. Finally finite-spectrum and non-uniform background cases are considered. FULL

For full-spectrum

quantities

SPECTRlJM

ANALYSIS

one can write L= 71 ‘CUT4 + (I - E)7r_‘ITT;:

where: L= radiance; 0 = Stefan Boltzmann constant; T = target E = target plate emissivity; T, = background temperature. The difference in radiance between bar and plate then is given by:

a quantity

dT defined

by:

4eoT3 dT,, Then

dT,,, is the apparent

error

= 4oT”dT

temperature

+ aT4dc

- aT,4dc

and it is given by:

difference

= dT + a(T4 - T,“) dQ&T3

dT,, The relative

temperature;

+ nT4 de - UT, de

ndL = 4EaT3dT Consider

plate

in temperature

is therefore:

@IT,, - dT)/dT

=

i’

; -;; -Tm [1 - (TB/T)4] dc!dTj

RE = D(T/4)[1

- (T,/T)‘]

where D is the quotient of the two changes and E is assumed to be one. This shows the error is zero when the target and background are the same temperature or when there is no difference in emissivity. Further, the error will get larger as the difference between T and T, gets larger. Figure 2 is a curve for the relative error for different values of dc/dT as a function of T when T, = 310K. It seems obvious that for a photon distribution the error is given by:

;f

, 300

310

,

,

320

330

Temperature,

, 340

1 350

K

Fig. 2. The relative error in bar charts for the full spectrum of photon flux rate for the region 300K to 350K for different values of dc/dT and background temperature of 3IOK.

Errors

in MRTD

charts

311

The error is zero for de = 0 and for T = TB. The first of these results is just satisfaction of the basic assumption. The second is the situation for which the background irradiation exactly makes up for the emissivity difference. The error is larger for larger de and smaller for larger dT. The quantity D might be made smaller by making the effective emissivity of the back plate more nearly equal to that of the front plate. MONOCHROMATIC

In this case the fundamental

ANALYSIS

equation is: L, = LfB + (1 - E) Ly

The change is: dL = La’ xeX(eX - 1)-r dT/T + (LYE - La) de where: x = cJAT;

c2 = 14388 cm K.

The relative error can then be written: RE = (1 - L:/LfB) (xeX(eX - 1)-l)-’

For the Wien approximation

DT

this reduces to: RE = [l - exp(x - xs)] x-l DT

The relative error for the region in which the Planck function must be used can be written: RE = [l - (e, - 1) (eXB- 1)-r (1 - ewX)]x-l DT

These relationships are shown in Figs. 3 and 4 for two values of de/dT and for several wavelengths. Figure 5 shows the same information for temperatures near the background temperature. FINITE

BAND

ANALYSIS

(GRAY

BODIES)

Over any finite band one can write: L,A = L;f

+ (1 - E) Lf;,

The same manipulations lead to the same basic equation for the relative error. This is true because the temperature and emissivity differentiation are independent of wave-

Temperature,

K

Fig. 3. The relative error in bar charts based on the Planck Law for de/dT = 0.001 and for different wavelengths and a background temperature of 310K.

WILLIAM L. WOLFE

378

1

I

I 310

-‘00300

I

320 Temperature,

Fig. 4. The relative

I

I 340

330

3x,

K

error in bar charts based on the Planck Law for dr/dT wavelengths and a background temperature of 31OK.

= 0.01 for different

length for graybodies. Then : 1-

RE = (PI-/x)

I

;:-qex - I)-” di,/ : J-JA

3‘3i.

xe”X”(e”

- f)-‘di. /

DISCUSSION

Most measurements are made with the front plate near ambient temperature (300K-320K). The rear plate usually varies in temperature from a few degrees above the front plate to 10 or 20 degrees above the front plate. fn this region the errors can be reduced by measuring the effective emissivities of the bar chart, so the errof would then depend upon the measLIring instrument. Part of the reason that there is any error at all arises from the fact that most infrared instruments measure differences of irradiance on their entrance pupils rather than diEerences in either temperature or emissivity. An additional contribution to the error is

7

6

304

306

308

Temperature, Fig. 5. Detail

of the relative

error

1

310

1 -2-r-5k A

~

320

K

near the background

tcmperat~lrc

of 31OK

Errors

in MRTD

charts

379

the use of the bar chart as a primary standard. These calculations seem to underline the fact that the charts should be calibrated. Perhaps we are also using the wrong specifications for the performance. Only the apparent MRTD should be specified in terms of the variations of the charts as measured by an appropriate radiometer. The measurement procedure and specifications must be described very carefully. Some charts of this type use the contrast between sky or other ambient radiation and the front plate. These need to be calibrated by their very design, so no additional analysis of the type carried out here is necessary.