A Spectroradiometer for the Spectral Region of BiologicalPhotosensitivity

Thomas H. Bulpitt, Murray W. Coulter, and Karl C. Hamner

The instrument described was specifically designed for measurement of the spectral irradiance of lightingin experimental plant growth chambers. However, without modification it is usable for experimentallight source measurements and for illumination engineering applications. The spectroradiometer isbasically composed of a motor-driven grating monochromator with suitably adapted multiplier photo-tube, amplifier, and strip chart recorder. The spectral region covered extends from 350 nm in the nearuv to 850 nm in the near ir. Included here is a discussion of the desirable characteristics of spectro-radiometers for this application.

Introduction

In the last few years the need has developed for thequantitative and qualitative measurement of lightsources or arrays used in experimental plant growthchambers. For meaningful description or repetitionof results involving the light responses of plants, it isimperative that irradiance levels and spectral qualitybe precisely specified. It is the purpose of this paperto discuss those characteristics we believe to be de-sirable in such a spectroradiometer and to describe onepossible solution presently in use at UCLA in the De-partment of Botany and Plant Biochemistry.

Currently the most common tool of the biologist formeasuring light is a foot-candle light meter. It isobvious that foot-candle readings give no indication ofthe light composition and also tell very little about theactual quantity of radiant energy unless the light sourcesare extremely standard and well defined. The morecommon utilization of light in plant growth facilitiesinvolves a combination of both fluorescent and in-candescent lamps. Manufacturers can usually providereasonably accurate specifications of the spectral dis-tribution for the various fluorescent sources used, buttheir specifications do not attempt to take into accountthe effect of various reflective surfaces and transparent

The authors were with the Department of Botany and PlantBiochemistry, University of California, Los Angeles, California.T. H. Bulpitt is currently with Douglas Aircraft Co., SantaMonica, California; M. W. Coulter is now with the Departmentof Biology, Texas Technological College, Lubbock, Texas.

Received 18 November 1964.This work was done at the Department of Botany and Plant

Biochemistry, University of California, Los Angeles, and wassupported by a National Science Foundation grant to KarlC. Hamner of that department.

barriers which are frequently necessary in such facili-ties. Fluorescent lamps also frequently show consider-able variation in the amount of energy emitted for mer-cury lines at various temperatures.

In studies of photoperiodic systems, the need for anaccurate and usable spectroradiometer has become es-sential in studies of biological responses. The de-velopment of the instrument described here was ini-tiated because of a complete lack of any commerciallyavailable instrument for this purpose which was ac-ceptable for measurements in the range of biologicalsensitivity and which was reasonably priced. Untilthis time, it has been common to determine light struc-ture and total energy of a given light source inde-pendently. Spectral distribution is frequently de-termined by use of a spectrometer with readings givenin relative units. Such readings, however, are notalways convenient or practical. Total energy readingsare usually determined by use of a thermopile, and theenergy in each spectral region of the light source isextrapolated to the curve obtained for light structure.Such determinations are extremely cumbersome andfrequently inaccurate. The spectroradiometer de-scribed here was designed to cover the spectral rangeof primary importance to the biologist with a sensi-tivity sufficient to provide a resolution appropriate forthe biological systems which might be studied. Thedirect recording of incident radiant power with wave-length augments the usability as well as the accuracyof the instrument.

General Considerations

The measurement of light incident on plant foliagerequires that one important geometric condition be met.The response of the instrument used should indicatethe total irradiance incident on an element of foliage;

July 1965 / Vol. 4, No. 7 / APPLIED OPTICS 793

i.e., the response should be proportional to the cosine ofthe angle from the normal. This characteristic is nottoo important where the source is direct sunlight, butit is very important for the measurement of irradiancefrom extended sources. Plant growth chambers typ-ically consist of extended and mixed light sources andoften reflection from walls contributes significantly tothe usable light. Larche and Schulze' described avery satisfactory cosine receiver consisting of a diffusereflecting spherical cavity, and integrating sphere, witha conical internal baffle. Tea and Baker2 described anintegrating sphere with a lens for an internal baffle.Budde3 recommends an integrating sphere configura-tion with a nearly perfect cosine response.

Following the light receiver and integrator is themonochromator. Monochromators of many configura-tions can be adapted to this function. They may beeither refractive or diffractive, or they may employwedge spectral interference filters. Whatever type isused, the stray light must be compensable. Notethat we do not say that stray light must be negligiblylow. Provided that stray light is independent of wave-length over our measurement range, it can be com-pensated by electronic zero suppression. This is nottrue of all monochromator applications. Tests suchas those described by Poulson4 should always be per-formed as required to qualify equipment for a specificapplication. For our purpose the low stray light in-herent to double monochromators is not entirelyjustified, although certainly not undesirable. If dif-fraction grating or interference wedge monochroma-tors are used, suitable broadband filters may be neededto remove unwanted spectral orders.

Whatever the type of monochromator used, con-sideration must be given to the spectral bandwidthdesired. It has been our experience that in the mea-surement of spectral energy from light sources a band-width from 5 nm to 10 nm is most satisfactory. Nat-urally, the bandwidth should be narrowed if there isreason to believe that the response under study cor-responds to a narrower bandwidth.

Selection of a photodetector to cover the desiredspectral region of 350-850 nm depends not only on de-tector characteristics, but on over-all system con-siderations as well. Photoelectric, photoconductive,photovoltaic, thermoelectric, and thermoconductivedevices can all be used for the specified spectral region.Thermoelectric detection has an appeal to those in-terested in devices which maintain a high calibrationaccuracy. However, thermoelectric detectors havelow voltage responsivities and require sophisticatedamplification techniques. Indeed, at low light levelsthe signal may not be large enough to pull out of theJohnson noise generated by the detector. Thermistorbolometers are thermoconductors offering higher re-sponsivities but greater noise and a tendency for re-sponsivity to change with ambient temperature. Thisis not too serious if the detector is used only for nullingpurposes. Photoconductors such as lead sulfide arealso more suited to nulling applications. Applicationsrequiring calibration of the detector, or calibration of

optics and detector, preclude the use of photoconductors,since, in general, responsivity and spectral response arestrongly dependent on temperature. Photovoltaiccells exhibit the same temperature sensitivities asphotoconductors. One of these photovoltaic devices,the silicon solar cell, possesses these traits to a lesserdegree than most. It has a usable spectral range of400-900 nm and has been successfully employed inradiometric devices. Reasonable stability of respon-sivity and spectral response are also found in photo-electric devices (phototubes and multiplier phototubes)when the long wavelength cutoff is avoided. Thereis but one photocathode material (cesium oxide, silver)which has the broad response (350-850 nm) required byour apparatus. This is the so-called S-l response.Unfortunately, the quantum efficiency of this materialis extremely low (on the order of 0.4% at 800 nm) re-sulting in poor signal-to-noise ratios for low irradiancelevels.

Treatment of the electronic signal after detection isa rather important step in obtaining a meaningfulmeasurement. For many detectors this treatmentconsists of high-gain voltage amplification. For somedetectors the problem may simply consist of assuringthe proper operation of the detector to obtain a linearrelation between irradiance and signal. Linearity isalso an important property of amplifiers. Wherehigh gain and wide dynamic range are required, theselection of a good amplifier is imperative. Amplifiershaving high gain frequently do not use enough feed-back. The result is poor gain stability and linearity.This is a fault of widely used amplifiers of thermocouplevoltages. Measurement of steady light sources maybe accomplished with either ac or dc amplification.Because of the low signal-to-noise ratio from mostdetectors at low light levels the ac method is superior.It increases signal-to-noise ratio by narrowing the noisebandwidth. Radiation is chopped and detected. andthe signal is amplified and synchronously rectified.Synchronous rectification is preferred to narrow-bandfiltering for two reasons. The bandwidth achievedwith synchronous rectification is one cycle, a difficultaccomplishment for filtering. In addition, synchro-nous rectification tends to average out the random na-ture of the remaining one cycle of noise rather than in-dicate a noise signal. Direct current amplification isfeasible with some detectors when the irradiance levelsare sufficiently high. This method will be treatedin detail later.

Finally, a desirable, though by no means essential,part of a spectral instrument of this type is a recorder.A great deal of tedious labor can be eliminated, and insome cases perhaps a degree of accuracy can be intro-duced. Use of a recorder also provides advantages inthe speed with which a spectral analysis can be ob-tained. For many uses the recorder response recordmay be used directly without additional treatment ofthe data. It is frequently sufficient to adjust the am-plitude and compare curves with overlays. In ap-plications in which greater accuracy is desired, cor-rection factors should be applied at many points along

794 APPLIED OPTICS / Vol. 4, No. 7 / July 1965

S H

S2/

MPT

Fig. 1. Optical diagram (shown in calibrate position): flight source, R-receiver, Ffilters, S1-entrance slit, M-plain mirror, M2,3-spherical mirrors, G-grating, S2-exit slit,

MPT-muliplier phototube, D-grating drive screw.

the spectral curve. Treatment of spectral line ir-radiance is usually easier and more accurately accom-plished from a recording.

An Instrument

With the above considerations in mind, the construc-tion of a spectroradiometer was undertaken. Theinstrument was to be fabricated at a total cost of $1500(excluding labor). This goal was attained.

Figure shows the optical diagram. The mono-chromator is a Farrand grating monochromator havinga focal length of 160 mm. This instrument was chosenfollowing the example of Teubner et al.5 who have shownits usefulness in this type of application. In front ofthe entrance slit is mounted a 7.62-cm spherical receiverinternally coated with smoked magnesium oxide. Theentrance aperture of the sphere is a 2.54-cm diam Vycorwindow which is normally positioned on top to receiveirradiance from an array of lights. At right angles tothis is the exit aperture. The sphere is so mountedthat it swivels about an axis through the exit aperture.This allows the entrance aperture to be turned to ahorizontal position for calibration with a standardlamp. The sphere itself is made of two spun aluminumhemispheres fitted together. The sphere mount alsohouses a filter change mechanism consisting of a Vycorglass (Corning CS 9-54) and a short-wavelength cutoffglass (Corning CS 3-68) which are interchanged by asolenoid in the space between the integrating sphereand the entrance slit of the monochromator. Theshort-wavelength cutoff glass removes the second orderof short wavelengths when the monochromator is setat wavelengths beyond 620 nm.

The grating of the monochromator is a replica gratinghaving 550 grooves per mm, and it is blazed for 550nm. The grating is rotated by a lead screw bearingon a ball set in an indentation in the back of the grating.

The lead screw is driven through a connecting shaftand a slip coupling by a reversible synchronous motorwith clutch (Hurst PC-DA-1) at 1 rpm. Each revo-lution corresponds to a spectral range of a little morethan 300 nm so that the range of 350-800 nm iscovered in a little less than 100 sec. The shaft isfitted with cams to actuate snap action switches whichstop travel at each end and change the filters at 620nm. The monochromator chosen was expressly builtby Farrand for the spectral region 400-700 nm, andthese are the limits of the marking on the dial. How-ever, it was found that extensions to shorter and longerwavelengths were possible without change.

At the exit slit was mounted a red sensitive (S-1) end-on multiplier phototube (MPT), the RCA 7102, havingten stages of electron amplification. It is operated at800 V by a 0.1% regulated multiplier phototube supply(Applegate Model 123A). The MPT is mounted inan aluminum tube which seals light tight to the mono-chromator at the exit slit with a separation betweenthe face of the MPT and the exit slit of approximately2.5 cm. Magnetic shielding of the MPT was notfound to be necessary. The socket and voltage dividerdynode resistor string were epoxy-potted into a re-movable end section of the aluminum tube. Thehigh-voltage supply and the anode signal current sharea common connector in this section. A shielded cablesupplies the voltage from and conveys the signal tothe console unit. The high-voltage supply is containedin the console.

The combination of integrating sphere spectraltransmission, monochromator spectral transmission(including grating efficiency), and MPT photocathodespectral sensitivity determine the over-all systemspectral response. Figure 2 shows this combinedspectral response. The fortunate result of this com-bination is a very flat response between 400 nm and800 nm. The rise at 350 nm is due to the intrinsicresponse peak of the photocathode material, and thedropoff beyond 800 nm is due to the lower gratingefficiency in combination with the photocathode long-wavelength cutoff. The high efficiency of the gratingin the middle of the visible tends to compensate for thelow photocathode response.

The choice of d amplification was dictated bysimplicity and low cost. But the choice of the methodof d amplification is frequently not considered. Anacceptable practice is to drop the anode current througha resistance value which is small compared with the

1,1_I

z

0zz

350 400 450 500 550 600 650 700 750 800 850

NANOM ETERS

Fig. 2. System response.

July 1965 / Vol. 4, No. 7 / APPLIED OPTICS 795

Fig. 3. Operational amplifier logic circuit schematic for amultiplier phototube amplifier.

dynamic plate impedance, a criterion for linearity.A temptation, however, is to use a very large value ofresistance to increase the signal voltage. This may beparticularly true in the case of a low response photo-cathode, such as the S-1. The voltage across theresistive load is then measured by a precision circuit,such as that described by Budde6 or amplified andsuitably indicated or recorded. The trouble maybe that in using the preferable low value of load, theamplifier may have insufficient sensitivity for the S-1.Consider the logic circuit schematic of Fig. 3.

The amplifier, -A, represents an operational amplifierhaving high gain, high input impedance, and low out-put impedance. Used in this way the operationalamplifier allows us to obtain the same voltage real-ized by the aforementioned high load impedance, butat the same time presents a low impedance load to thephototube. This is done by using a high impedancefeedback on the amplifier. The load impedance offeredto the phototube is the feedback impedance dividedby the open-loop gain of the amplifier. The reducedinput impedance also minimizes the time constanttypical of high input impedance amplifiers when fedby a capacitive coaxial cable. Voltage output of suchan amplifier is the product of phototube current andfeedback resistance. No significant nonlinearity isintroduced by a high-gain amplifier used this way.This type of amplifier is recommended by Baker andWyatt7 for phototube current measurement. Notethe provision for a small variable positive current tobe injected at the input to remove phototube darkcurrent from the feedback loop. Figure 4 is the actualcircuit used within the amplifier logic circuit symbol.Figure 5 shows the complete spectroradiometer systemconsisting of receiver, amplifier/control console, andHeath recorder.

Calibration of the instrument for spectral irradianceis based on a National Bureau of Standards standardof spectral irradiance.8 The lamp used is a quartz-iodine tungsten filament lamp operated at 6.50 Afrom a variable-transformer ac power supply. Thepower dissipated is approximately 180 W. Spectralquality is similar to that of a 3000° blackbody. Thechoice of this lamp over a color temperature standardis dictated by the extended range of our measurements.The deviation in spectral distribution of a color tem-

perature standard from a blackbody is only about4 2% over the region 400 nm to 700 nm. Outsideof this region the deviation becomes more serious.

Performance

Through most of the spectral range of this instrumentthe response is approximately 2 nA//uW/cm2 for a10-nm bandwidth. This response has varied no morethan 10% over a period of six months as determinedby recalibration. Variation is primarily a function ofhigh-voltage regulation and multiplier phototubegain stability. Dark current of the particular RCA7102 multiplier phototube used was found to be ap-proximately 2 X 10-8 A at an ambient temperatureof 250 C. For most measurements thus far, the full-scale signal currents have ranged from 0.2 MA to 2.0AA, and dark current has not been troublesome. Forlow levels of illumination it has been necessary to coolthe phototube by packing ice around the phototubehousing. Condensation on the face of the phototubeis prevented by introduction of a small amount ofdesiccant, silica gel or calcium sulfate, into the mono-chromator housing. Linearity of system response hasbeen checked over two decades of multiplier phototubeanode current. Response was found to be acceptablylinear within 2%.

Summary

The several considerations relating to the design ofa spectroradiometer for the region of biological photo-sensitivity have been discussed. The scope of thetreatment was such that all considerations could notbe discussed in detail. The purpose was to present acomprehensive picture so that a selection of components

INU S. I1

Fig. 4. Operational amplifier circuit.

Fig. 5. Spectroradiometer consisting of (from right to left)the receptor unit, the amplifier/control console, and the recorder.

796 APPLIED OPTICS / Vol. 4, No. 7 / July 1965

might be made to suit the particular measurementproblem. Our selection was by no means the onlysuitable one. In addition to the technical considera-tions above, the choice was influenced by the practicalconsiderations of cost, size, and convenience of opera-tion. The instrument described has served well inapplying a combined quantitative and qualitative labelto the illuminant parameter in studies of phytobiologicalphotosensitivities.

We are indebted to J. Archer, Los Angeles CityCollege, Ernest Bellows and Harry Tilson of the SpaceBiology Laboratory for technical assistance in con-structing the instrument.

References

1. K. Larche and R. Schulze, Z. Tech. Physik 23, 114 (1942).2. P. L. Tea and H. D. Baker, J. Opt. Soc. Am. 46, 875 (1956).3. W. Budde, Appl. Opt. 3, 939 (1964).4. R. E. Poulson, Appl. Opt. 3, 99 (1964).5. F. G. Teubner, S. H. Wittwer, R. S. Lindstrom, and H.

Archer, Proc. Am. Soc. Hort. Sci. 82, 619 (1963).6. W. Budde, Appl. Opt. 3, 69 (1964).7. D. J. Baker and C. L. Wyatt, Appl. Opt. 3, 89 (1964).8. R. Stair, W. E. Schneider, and J. K. Jackson, Appl. Opt. 2,

1151 (1963).

tOff Optics and~e Opticists

This column is compiled partly from information sent by our Reporters in various centers of opticsacross the continent, but the Managing Editor welcomes news from any source

Science increases our power in proportionas it lowers our pride

Claude Bernard 1813-1878

More accurate spectral radiance calibrations .... A spectro-radiometer that makes possible a fivefold increase in the accuracyof calibration of spectral radiance standards-up to now inferiorto that of standards representing almost any other widely usedphysical quantity-has been developed at the NBS Institute forBasic Standards. The increased accuracy is especially importantto pyrometry, spectroscopy, and astrophysics. The new instru-ment, designed by H. J. Kostkowski, D. E. Erminy, and A. T.Hattenburg, extends the range of calibration of 2000 I Effortsare being made to standardize the calibration procedure, to ex-tend the wavelength range to 2.5 JA, and, in order to complete thiswork as soon as possible, only a limited number of special cali-brations are being offered at this time. Persons who requirecalibrations below 250 nm and those whose research problemsexhibit a need for a smaller uncertainty in spectral radiance thanis available commercially (about 5 percent) should write to Dr.Kostkowski, Temperature Physics Section, National Bureau ofStandards, Washington, D.C. 20234.

This journal will feature Optics at the NBS in its January 1967issue under the feature editorship of I. C. Gardner, past memberof the NBS staff and past president OSA.

A two-week course in electromagnetic measurements and stand-ards will be conducted 9-20 August at the Boulder Laboratoriesof NBS. The course will consist of some forty lectures by physi-cists and engineers of the NBS Radio Standards Laboratory,and it is being offered in cooperation with the University ofColorado. Major topics to be covered are: a review of electro-magnetic theory, lumped circuit parameters, free space fieldstrength, generalized waveguide theory, high-frequency wave-guide parameters, radio-frequency properties of materials, radio-frequency noise, and time and frequency. Interested personsshould address the Bureau of Continuation Education, Rm. 328,University Memorial Center, University of Colorado, Boulder,Colorado 80304. Registration will close 15 July.

Other news from NBS-Boulder concerns the Joint Institute forLaboratory Astrophysics Visiting Fellowship Program, whichprovides an opportunity for persons actively contributing to re-search, study, or teaching in theoretical astrophysics, low-energyatomic physics, and other related areas to come to JILA to con-tinue their studies and research so that there is a continuousinterchange of thinking in the general area. In addition to theten or so Visiting Fellows each year, there are permanent Fel-lows, and one of these-recently appointed-is Roy H. Garstang,formerly of the University of London Observatory, to be pro-fessor in the Department of Physics and Astrophysics of theUniversity of Colorado and permanent Fellow of JILA.

Glass research at NBS .... A change in the method of determiningthe strength of flat glass has been recommended as a result ofcomparison tests of midpoint and two-point loading. A two-point loading test gives results which are more representative ofthe strength of the glass than the currently used midpoint load-ing test. The work has been done by M. J. Kerper, T. G. Scuderi,and E. H. Eimer, and will be published in the Proc. of ASTM.... Studies of the effects of heat treatment on the properties ofglass have shown that the relation between refractive index andYoung's modulus is a sensitive indicator of the nucleation andseparation of submicroscopic phases. E. H. Hamilton found thisrelation, and more details will be found in J. Am. Ceramic Soc.47,167 (1964).

Thermal emittance of ceramic oxides for space applications ....Reliable thermal emittance data are needed in the space programfor high-melting-point ceramic oxides. NBS scientists havedeveloped methods and equipment to provide these data from1200-1800'K over the wavelength region 1-15 p. This study, byH. E. Clark and D. G. Moore, was sponsored by NASA and willbe published in the Proceedings of the Symposium on theThermal Radiation of Solids.

Effect of temperature on optical windows .... The effect oftemperature on the vacuum-uv transmittance characteristics ofseveral common optical windows has been studied by A. H.

July 1965 / Vol. 4, No. 7 / APPLIED OPTICS 797