Sky Brightness Temperature Measurements at 135 GHz and 215 GHz

Zenith sky brightness temperature measurements at 135 GHz and 215 GHz have been made on a semicontinuous basis for a period of seven months in the Gunston Cove area of Northern Virginia. These measurements were made using Dicke receivers with noise figures of 8 dB and 14 dB, respectively. A liquid nitrogen cooled load was u ed to calibrate the measurements. The 215 GHz sky temperature was on the average about 80 K greater than that at 135 GHz. Clouds were found to cause the sky temperatures to fluctuate as much as 150 K in a few minutes. Graphs are presented to outline general trends of the data as well as representative days, including the blizzard on February 11, 1983. In addition, empirical relations between precipitable water vapor, atmospheric water density at the surface, sky brightness temperatures, and zenith attenustion are given for visually clear days.


I. INTRODUCTION
T HE EARTH'S ATMOSPHERE has four extremely high fre- quency (EHF) windows around 35, 95, 135, and 215 GHz.These windows are framed by rotational absorption lines of oxygen and water molecules and their transparencies vary with the atmospheric water content.Excellent reviews of the atmospheric propagation in this spectral region can be found in [l] - [3].The two lower frequency windows have been more extensively characterized and utilized due mainly to availability of equipment, However, with the continuing improvement of technology in the miLlimeter wave region the upper two windows are becoming increasingly important for application in radio astronomy, remote sensing, and communications [4] - [ 6 ] .
The authors are with the U.S. Army Night Vision and Electro-optics Laboratory, Fort Belvoir, VA 22060.
In this paper we report the measurement of zenith downwelling radiation at 135 GHz and 215 GHz over a seven-month period from January through July 1983, in the Gunston Cove area of Northern Virginia (longitude,77'07'W;latitude,38'41'N).This radiation is described in terms of an equivalent sky brightness temperature which at these frequencies is simply related t o a black-body source by the expression P = kTAf, where P is the detected power [7].These sky brightness temperatures on clear days are caused almost entirely by atmospheric molecular emissions and can be used to calculate the zenith attenuation or absorption [8] - [ll] .
These measurements were made using an unique uncooled dual frequency Dicke radiometer [12] that provided near simultaneous measurements at the two frequencies (-3 ms switching time).Initially, data were recorded only during the daytime until system stability and reliability were established.Subsequently, they were taken continuously during the week from Monday morning t o Friday evening.Corresponding weather data were obtained from two sources.Dew point, relative humidity, and surface temperature measurements at Washington National Ahport (14 mi NNE) were supplied by the National Weather Service.Precipitable water data for this area were taken from contour maps supplied by the US.
Air Force Environmental Technical Applications Center in A s h e d e , NC. at the Georgia Institute of Technology.It consisted of a Gunn diode local oscillator at 35.83 GHz which was tripled and then subharmonically mixed with the signal [13].The mixer contained two antiparallel point contact diodes [ 141 , [15] .A variable attenuator between the local oscillator and the mixer was set to obtain the maximum sensitivity.The 135 GHz receiver was purchased from Alpha/TRG.It also used a Gunn diode for the local oscillator which was tripled to the fundamental frequency of 135 GHz.The mixer contained two beam lead diodes(ba1anced type) [16] .Both receivers had IF bandwidths of 1 GHz.The receiver noise figures (DSB) were 14 dB for the 215 GHz system and 8 dB for the 135 GHz system.The signals from the squarelaw detectors were sent t o lock-in amplifiers with time constants of 1 second and then recorded on a chart recorder.The chopping rate varied from 400 Hz initially to 300 Hz at the end due to wear in the chopper wheel bearings.The lock-in amplifiers were able to track the observed variations in the chopping rate.
The radiometer and the electronic equipment were housed in a truck which had a 5 cm thick Styrofoam window and zenith reflector.The two receivers were arranged in the configuration shown in Fig. 1.The chopper was a 45 cm diameter polished aluminum wheel which acted both as a mirror and a window, allowing the receivers to alternately view the Dicke load and the sky.The Dicke load was made of Eccosorb CR-117 [I71 and kept near ambient temperature (15-25').The TPX (Poly 4-Methyl-Pentene-1) lens was 28 cm in diameter, causing a beam divergence of about 10 mrad for the 135 GHz signal and 6 mrad for the 215 GHz signal.Two aluminum reflectors, not shown, were used to reflect the sky radiation into the lens.The zenith reflector, mounted outside the truck, reflected the sky radiation 90" to a horizontal direction through the Styrofoam window.The other reflected the sky radiation another go", still horizontally, into the lens.The second reflector could be pivoted to allow the receivers to view the primary calibration load.Other than the zenith reflector and window, the entire arrangement was mounted in a 70 cm X 70 cm X 140 cm aluminum box.

CALIBRATION
The primary calibration load consisted of a 35 cm X 35 cm of Eccosorb CR-117 mounted on an aluminum block in contact with a stainless steel dewar.The entire arrangement was sealed in a Styrofoam box.The load was furnished by Georgia Institute of Technology and had a stated accuracy of k10K from a thermocouple readout.In order to conserve liquid nitrogen a secondary calibra-tion load was normally used.This was a 13 cm X 13 cm section of Eccosorb-CV which was placed in an open Styrofoam box f j e d liquid nitrogen.The measured calibration temperatures for the secondary load were found to be 90 K for the 215 GHz system and 96 K for the 135 GHz system.Approximately once an hour this load would be placed in front of the lens causing calibration lines to be recorded.At the same time, the temperature of the Dicke load T,: and the ambient room temperature TA were also noted.For analysis of night data the calibration readings for the preceeding and following days were averaged.
Since the output of the detectors was linear, the temperature at the receiver TR in kelvins was determined from the following equation: where VR is the receiver voltage and m is the slope in K/V.
The slope was determined using the voltage and temperature readout from the calibration line.
TR was not the actual sky temperature since a number of structures of various absorptivity and emissivity were within the radiation path to the receiver.These structures included the two reflectors, the Styrofoam window, and the lens.The effect of each of these objects is accounted for by using the following equation [2j, [IO]: (2) Here T, represents the "output" radiation from the structure, T , is the "input" radiation, TA is the temperature of the structure, and L is the loss factor.The first term on the right corresponds to the energy lost by absorption in the material and the second term is the energy emitted or reflected by the material.This equation assumes that the background energy reflected off the material equals the energy emitted by the material 11 81 .The sky temperature was calculated from TR by using (2) sequentially at each structure.The loss factor for the 5 cm Styrofoam window was measured to be 1.05 for both frequencies and the loss factor for the lens was 1.26 at 215 GHz and 1 .I 5 at 135 GHz.The losses due to the reflectors were negligible.Sidelobe contributions from the receiver horn were not accounted for in the analysis but could add as much as 3 K to the measured temperatures.In addition, water accumulation on the window and reflector affected the readings during rainy weather.The absolute accuracy in these measurements is estimated to be k10 K , due mainly to the uncertainty in the primary calibration standard.

IV. RESULTS
Fig. 2 shows the sky temperature data for a representative day.The weather for this particular day included rain which ended in the morning followed by a clear sky with scattered clouds in the afternoon.This weather is clearly indicated by the sky temperatures at both frequencies; the rain and clouds yield high sky temperatures because of their water content whereas clear skys are relatively cool.Fig. 2 also illustrates some general results.For one, the 215 GHz sky temperature was almost always greater than the 135 GHz sky temperature, as expected.In addition, Fig. 2 shows the rapid fluctuation in temperature which can occur due to clouds.Changes in temperature of as much as 150 K in less than five minutes were sometimes measured.These fluctuations were caused by cumulus or stratus type clouds; high cirrus type clouds had little effect on the sky temperature.x 10-2Tt15 + (36.2 f 4.0).u = 7 percent for this fit.
The analyses for Figs.3-7 following contain only uniform sky data, i.e., days which were visually clear with the possible exception of high cirrus clouds.For these conditions the water density measurements at National Airport produced 263 data points, selected at equal t h e intervals.The precipitable water measurements were taken every 12 h and produced useful data for 92 of the 263 points.For Figs. 3-7 both linear and quadratic least squares fits were compared and the best one chosen.The accuracy of each fit is expressed as the root mean square (rms) deviation of the residuals u given as a percentage of the mean of the dependent (y-axis) variable.
Fig. 3 shows the correlation of the sky temperatures for the two frequencies including all 263 measu,rements.Fig. 4 shows the sky temperatures at both frequencies plotted against the atmospheric water density at the surface.Note in this case the 135 GHz data was best fit by a linear curve while the 215 GHz data was fit by a quadratic.The sky temperatures in Fig. 4 are in general agreement with those presented by Smith [l I ] , however, this graph should be compared to Fig. 5 which shows sky temperatures plotted against the total precipitable water vapor.Here both the 215 GHz and 135 ,GHz data were best fit with quadratic curves.The sky temperatures are better correlated to the precipitable water than the surface water density, as expected [19].It should be noted that the fitted curves in Figs.relatively weak correlation between the surface water density and the total precipitable water vapor for the 92 coincident measurements.Here the fitted curve yields a water scale height of about 2.2 k m .
The zenith attenuation on clear days is equal to the absorption [2] and can be obtained from the sky temperatures by use of the following equation [9], valid for T < 2Np: where Ts is the measured sky temperature, TBB is the cosmic blackbody temperature of 3 K, and T,, is the mean radiating temperature.T,, can be in the range between 270 K-280 K for this geographic area and 270 K was used because it produced slightly better fits.The resulting points were best fit with straight lines as shown in Fig. 7. Also shown in this figure are representative points at 0 cm, 1.44 cm, and 2.87 cm derived from the NTIA millimeter wave propagation model for the U.S. Standard Atmosphere [20].
In Figs. 8 and  The median daily sky temperature difference and the rms deviation in kelvins are shown by month in Table I.Rainy days, in which the temperature difference was very small, were not included in these results.In a very few instances, thunderstorms in particular, the 135 GHz temperature became greater than the 215 GHz temperature.In other instances, again rare? the two frequencies would indicate opposite changes, i.e., one would increase while the other decreased.It is not known if these anomalies are real or if the beams intecepted different active areas in the sky because of spot size or beam misalignment.Before installation in the truck the beams were aligned within 1" of each other.
The minimum and maximum sky temperatures are also shown in Table 1 for these measurements.Although there were more  cold days in January, the coldest sky temperatures occurred on March 25, in the presence of high cirrus type clouds.There is a sharp jump in the minimum sky temperature between May and June, another indication of the short spring in this area.The warmest temperatures were near 300 K at both frequencies during heavy overcast or rainy conditions each month.

TEMPERATURE (K)
The sky temperatures for the blizzard of February 10-1 1, 1983, are shown in Fig. 10.Although the sky temperatures began to increase on the loth, heavy snow fall did not begin until 0700 on the 11 th and lasted about 10 h.In terms of total snowfall, this was the third worst storm in the history of this area. . .This was a "dry" snow which did not collect on the reflector or window.
James  noise) present at the array elements.Thus, knowing the jammer characteristics, one can assess the adaptive array performance.

Effect of Jammer
It is shown that the output SINR of an adaptive array is a function of jammer power.The output SINR decreases with increasing jammer power and, for a wide-band jammer, it eventually goes to zero.Unlike continuous wave (CW) jammers, a wideband jammer does not go through power inversion.Instead, as the jammer power is increased, the INR at the array output undergoes oscillations.where CP is the covariance matrix of the signals present in the array elements, and S is the reference correlation vector.In the presence of a single CW desired signal and m jammers, the covariance matrix CP can be written as where u2 is the thermal noise power in each antenna element, I is an N X N identity matrix, CPj is the covariance matrix due to the jammers, Ad is the desired signal amplitude, u d is the desired signal vector, and superscripts asterisk and T denote complex conjugate and transpose, respectively.In (2), it is assumed that the thermal noise voltages from the array elements are Gaussian with zero mean and variance u2 and are uncorrelated with each other and with other signals incident on the array.The desired signal is assumed to be a narrow-band signal uncorrelated with jammers.Further, assuming that the reference 0018-926X/84/0900-0933$01 .OO 0 IEEE Fig. 1. 135 GHz/215 GHz Dicke radiometer.
Power on the Performance of Adaptive Arrays INDER J. GUPTA, MEMBER, IEEE Abstract-The effect of jammer power on the performance Of adaptive arrays is studied.It is shown that the output signal-tointerference-plus-noise ratio (SINR) of an adaptive array is a fnnction of jammer power.In the presence of a wide-band jammer, the output SINR of the array decreases with an increase in jammer power and eventually goes to zero.Unlike continuous wave ( C W ) jammers, a wide-band jammer does not go through power inversion.Instead, as the jammer power is increased, the interference-to-noise ratio (INR) at the array output shows oscillations.For large jammer power, the output INR increases with 811 antenna arrays have been receiving a great deal of attention in the areas of radar, sonar, communication, and seismic systems.One of the primary reasons for this fast growing interest is the protection it provides against undesired signals (jammers, unintentional R F interferences, etc.).An adaptive array automatically steers pattern minima onto sources of undesired signals, thereby reducing undesired signals at the array output.The stronger an undesired signal, the lower is the pattern minimum, and thus the undesired signals go through a power inversion [l] , [2] and the output signal-to-interferenceplus-noise ratio (SINR) does not vary much with jammer power.These concepts of steering pattern minima and power inversion apply t o narrow-band signals, but not to wide-band signals where different frequency components arrive with different phases at the array elements; and consequently a pattern minimum at one frequency may not be a pattern minimum at another frequency.Thus, it is difficult to assess the output SINR and the output interference-to-noise ratio (INR) as jammer power varies.In this paper, analytic expressions for the output SINR and INR of an adaptive array in the presence of wide-band jammers are derived.The expressions give the output SINR and INR in terms of the eigenvalues and the corresponding eigenvectors of the covariance matrix of undesired signals (jammers and thermal Manuscript received November 14, 1983; revised April 26, 1984.The author is with the ElectroScience Laboratory, Department of Electrical Engineering, The Ohio State University, 1320 Kinnear Road, Columbus, OH 43212. Beyond a certain jammer power level, the output INR monotonically increases with increasing jammer power.At this point the array is fully constrained and the array output SINR drops sharply.The expressions for the output SINR and INR are derived in Section 11.Adaptive array performance in the presence of a CW signal is discussed in Section 111.Section IV deals with a wide-band jammer.11.OUTPUT SINR OF AN LMS ADAPTIVE ARRAY The steady state weight vector W of the least mean square (LMS) adaptive array due to Widrow et ai.[ 3 ] is given by w = CP-l s (1) 9 the hourly sky temperatures are presented in terms of frequency of occurrence by month for February through July.Only complete 24 hour cycles, generally Monday through Friday morning, were included in this data set.Both these figures indicate that this area had a short spring in 1983; February through May represent winter months and June and July represent summer.

TABLE I MONTHLY
SKY TEMPERATURE RESUL.TS c 0 1 I .
Miller was born in St. Louis, MO, in 1943.He received the B.S. degree in physics, the B.A. degree in mathematics in 1964, and the M S .degree in physics in 1966 from Texas A & M University, College Station.He worked for Texas Instruments, Dallas, TX, from 1966 to 1968 primarily on nondestructive characterizing of silicon epitaxial layers.In 1968 he joined the staff of the U S .Army Night Vision and Electro-Optics Laboratory, Fort Belvoir, VA.His areas of work have been optically pumped laser material development, injection laser growth and fabrication, and millimeter wave detector and source research.He is currently engaged in infrared device research.Mr. Miller is a member of the American Physical Society.Richardson was born in Washington, DC, on June 2, 1958.He is currently in the electrical engineering program at Northern Virginia Community College, Annandale, VA.Since 1983 he has been a co-op student trainee at the Night Vision and Electro-Optics Laboratory, Fort Belvoir, VA, where he has been engaged in millimeter wave propagation studies.
E.Phillip H.