148 PIULIPS TECHNICAL REVIEW VOL. 16, No. 5-6

A RADIO SONDE FOR METEOROLOGICAL OBSERVATIONS

by A. HADER *) and M. van TOL.

Radio sondes, as nowadays commonl y employed by meteorological services, mus, s(!/.isfrfairly stringent. requirements as regards accurac)', and must. at the same time be as liglï andinexpensive as possible. The sonde developed by Philips in collaborution. ,<'it.h the RoyalDutch Mvteorological Instiuue and described in this article embodies several net" /<'(111.'1'1'5 ofiniporuuu:c in the [uljilmeni of these requirements.

Others, apart, from the writers, icho assisted in the development of this sonde nere ]\fessrs.R. J. Ritsma and H. J. A. Vesscur of the Royal Diuch. Meteorological l nsuuite, ]\11'. J. L.iVf. Reijruler s of tlu: Philips Measuring Equipment Derelopment Dept., an d Messrs. 11.van Sucluelcn. and D. J. IJ. Admiraal of the Ph.ili ps Research Labor at ories in Eindhooen,

621.396.91

The radio sonde

The study of physical conditions in the upper-air,i.e. in that part of the atmosphere not within directrange of measuring instruments on the ground, isan essential branch of modern meteorology. Know-ledge so acquired is particularly valuable in thepreparation of weather forecasts.The principal quantities to be measured are the

temperature and the humidity as a function ofaltitude, but accurate pressure measurements arealso required as a means of determining the altitude.Observations from aircraft enable us to record thesedata at altitudes up to about 10 km, but arc veryexpensive and can be carried out only in the vicinityofan aerodrome; hence another method, i.e. balloon-borne measuring apparatus is preferred. Balloons

*) Royal Dutch Meteorological Institute, De Rilt, Holland.

havc already been used for many years by meteoro-logists as a means of measuring wind velocity, butother information concerning the physical conditionof the atmosphere can also be obtained from anascent, by attaching recording equipment for pres-sure, temperature and humidity below the balloon.The latter, filled with hydrogen, climbs at a constantrate of about 5 m/sec., and bursts on reaching analtitude of 15-25 km. A parachute then opens tolower the recording equipment (soncle) slowly to theground.It is necessary to search for, and recover the sonde

when once it has reached the ground, a task whichmay often present some difficulty, particularly insparsely populated areas. Other serious disadvanta-ges of this method are that the results of theobservations cannot be ascertained until, and are

NOVEMBER-DECEMBER 1954 RADIO SONDE 149

never known unless, the sonde is found; hence aperiod of hours, or even of days may well elapsebetween the recording and the actual reading ofthe data.

However, in about the year 1930, a method wasevolved whereby the results of the measurementscan be ascertained whilst still being gathered, thatis, whilst the sonde is still airborne.

This method is that of the "radio sonde", nowemployed by most meteorological services. Inprinciple, a radio sonde comprises two individualunits. One of these, the measuring unit, containsinstruments sensitive to temperature, humidity andpressure, and a circuit whose function is to pass theinformation to the other unit, that is, the trans-mitter. The latter is a small short-wave transmitter;the signal proceeding from this transmitter andreceived by equipment on the ground contains theresults of the measurements in code form.

The radio sonde as a whole must satisfy one ortwo fundamental requirements, the most importantof which is that it shall be sufficiently accurate (theaccuracy considered ideal by meteorologists hasnot yet been attained by any radio sonde). It islikewise very important that the sonde be as lightas possible, so that it will reach the highest possiblealtitude; this severely restricts circuit design, notleast owing to the consequent limitation of batterycapacity (the batteries make up a considerableproportion of the total weight). Hence the availablepower supply is very small.

To he suitable for regular, and at the same timeeconomical use, the radio sonde must be inexpensive.In the Netherlands, the probability that it will nothe recovered after an ascent is about 50 %, and evenif recovered it may he seriously damaged (despitethe parachute). In view of the latter possibility itis of course desirable that the radio sonde be mecha-nically robust.

In collaboration with the Royal Dutch Meteoro-logical Institute at De Bilt, Philips have developeda radio sonde which embodies a number of uncon-ventional features. A description will now he given,dealing with the items in the following order; the_measuring unit, modulation of the transmitter bythe measuring elements, the transmitter itself, andfinally the results of several trial ascents by thenew sonde.

Temperature, pressure and humidity measurements

To satisfy fully all the requirements of the meteo-rologists, a radio 'sonde would have to be capableof effecting extremely accurate measurements. Inpractice, however, the accuracy is limited to what

is technically and economically practicable. Accor-dingly, the following requirements were taken asa starting point for the project here considered.Firstly, temperature measurements are to be accu-rate to within 0.5 oe, that is, in effect, to within 0.5%,since the temperature in the particular regions ofthe atmosphere where the measurements are carriedout varies on an average between 20° and -70 oe.Secondly, a similar accuracy is required in the mea-surement of pressure. The latter varies between1050 and 70 millibars and the requirement is that ithe measured accurately to within about 5 millibars,i.e. to within 0.5%. Thirdly, possible errors in themeasurement of the relative humidity, which isexpressed as a percentage of the saturation value,must not exceed 5-10 %.The accuracy of such measurements depends

firstly upon the particular measuring element inthe sonde, secondly upon the manner in which thedata are transmitted to the receiver on the ground,and thirdly upon the measuring instruments on theground. In this article we shall consider only thefirst two factors.

Most radio sondes include several moving parts,the purpose of which is to convey the readings ofthe instruments measuring temperature, pressureand humidity mechanically to the transmitter, e.g.through ~ variable capacitance, resistance or self-inductance. Many also include a switching mecha-nism to connect the output of each measuringelement in turn to the transmitter. This mechanismmay be a rotary switch turned by a small windmill(which, in turn, is driven by the downward flowof air relative to the ascending sonde), a clock, orin some cases a small electric motor. Experiencehas shown that the reliability of sondes is oftenimpaired by such mechanical parts.

In the Philips radio sonde the idea of switchingwas abandoned; measurements of the three quanti-ties are transmitted simultaneously as three audio-frequency signals of variable frequency on a singlecarrier. This method necessitates the use of a separ-ate audio-frequency oscillator for each of the quanti-ties measured; altogether, then, three such oscillatorsare required, whereas if the signals were transmittedconsecutively one oscillator would he sufficient.However, the switching mechariism has now beendispensed with, and by using suitable sub-miniaturevalves for the oscillators, the use of three oscillatorshas not increased either the total weight or the priceof the radio sonde compared with earlier types. Theoscillators will be considered presently. As we shallnow explain, the number of moving parts in thesonde has been reduced to a minimum.

150 PHILIPS TECHNICAL REVIEW VOL. 16, No. 5-6

The measurement of tempermureThe temperature measurement can be accomplis-

hed without the aid of any moving parts whateverby means of a temperature-sensmve elementcapable of acting direct upon the oscillator fre-quency, e.g. a resistor sensitive to temperature.This resistor must be made of a material having atemperature coefficient of resistance high enough toensure the desired response. Such a material is tobe found in the N.T.C. resistor, or thermistor 1),that is, a semiconductor having a high negativetemperature coefficient ofresistance. It is found thata single N.T.C. resistor is capable of changing thefrequency of an R-e oscillator by more than a factorof 2 for a decrease in temperature from 20° to-70 oe. For this purpose a straight-wire resistorwas used whose resistance varies from about 30 kQat room temperature to about 500 kQ at -70 oe.The change in the oscillator frequency per 0.5 oe(the desired accuracy) is then about 0.5%' whichcan easily be measured.

The thermometer (see fig. 1) is a straight-wireN.T.C. resistor about 10 mm long, which is intro-duced direct into the frequency-controlling networkof an audio-frequency oscillator (R-e oscillator).One advantage of this resistor over the bimetallicstrip thermometer often used in other sondes isthat it is very much quicker in response: it wasfound during trial ascents that the N.T.C. resistorresponds quite noticeably to temperatm'e fluctua-tions not detected by a bimetallic thermometer.Again, the "radiation error", which may he veryappreciable in the case of a bimetallic strip (up to10°C), is very much smaller when a thin wirethermometer is employed. Hence the thermometerwire can be mounted quite freely about 10 cmoutside the housing of the sonde, without anyradiation screen or other cover.

The measurement of pressure

Metal aneroid barometers have been used aspressure-meters in almost all the radio sondesproduced hitherto. However, the lag in responseof such a barometer is very liable to introducesystematic error into the pressure measurement;moreover, the instrument itself is difficult to manu-facture. For these reasons, and also to dispense withmoving parts, anuther method based on the principleof the hypsometer was adopted, at the suggestionof the Meteorological Institute.

A hypsometer is a vessel, opeu tu the atmosphere,

1) E. J. W. Vcrwey, P. W. Haaijman and F.C. Romeyn,Semi-conductors with large negative temperature coeffi-cient of resistance, Philips tech. Rev. 9, 239-24.8, 1947/1948.

contammg a liquid maintained by some means atboiling-point; the boiling -point ofthe liquid dependsupon the pressure of the surrounding atmosphere,that is, it decreases with this pressure. If the preciserelationship between pressure and boiling-point isknown, then, the one can be determined by mea-suring the other (e.g. by means of an N.T.C. resistorimmersed in the liquid). The hypsometer is veryreliable as a pressure-meter provided that the

Fig. 1. The Philips radio sonde. The entire instrument is housedin a cardhoard case, 165 X 120 X 155 mm. The temperature-measuring clement T is a str aig ht-wire N.T.C. resistor 10 mmlong, mounted outside the sonde; the pressure and humiditymeasuring-elements, Pand H respectively, are partlyvisible at the open side of the sonde. Altogether, the sondeweighs 500 grams; it is suspended from a hydrogen-filled balloonby a line about 10 m long, 5.5 ID of which is covered withconducting sleeving to act as an aerial.

liquid used in it is sufficiently pure, since for pureliquids the relationship between boiling-point andpressure is very well established.To maintain the liquid at boiling-point, at which

some of it evaporates, a certain amount of energyis required. This would of course be a big disadvan-tage if batteries were the only source of energyavailable, since these would increase the weight ofthe sonde. However, if we have a liquid whoseboiling-point will remain below the temperature of

NOVEMBER-DECEMBER 1954 RADIO SONDE 151

the surrounding atmosphere throughout the ascent,the energy to keep the liquid boiling will be suppliedby the atmosphere itself. This is in fact practicablesince although the tempCl'ature of the atmospheredrops as the radio sonde ascends, the accompanyingdecrease in pressure causes a similar drop in theboiling-point of the liquid. Curves 1 and 2 in fig. 2show thc relationship between decreasing pressureand decreasing temperature in the atmosphere(these curves are based on the summer and winteraverages, respectively, of observations carried outat various times and places). Accordingly, we mustchoose a hypsometer-liquid whose pressure/boiling-point curve is to the left of curve 2 in fig. 2. At thesame time the two curves should not be too farapart or the liquid will evaporate too quickly andit will be necessary either to insulate the hypsometerheavily, or to fill it with a very large amount ofliquid. Freon, the vapour pressurc of which isindicated by curve 3 in fig. 2, is a liquid which satis-fies these conditions very well indeed.

5,0 4,8 4,0 3,8 3,61ffl)- 80600

Fig. 2. Variation of ternperature with pressure in the "average"atmosphere during summer (curve 1) and winter (curve 2)Line 3 represents the vapour-prcssurc curve of freon. Thepressure of the atmosphere p is plotted on a logarithmic scaleagainst the reciprocal of the absolute temperature T, so thatcurve 3 is then nearly a straight line.

There is a limited region of the atmosphere whose averagetcmperature (in winter) is very close to the boiling-poin t offreon, so that on cold days the temperature of the air in thisregion may (hop below the boiling-point of the hypsometerliquid. Apparently, then, freon fails to satisfy the condit.iongoverning spontaneous boiling, but despite this the liquid isprevented from "going off the boil" by a second source ofheat within itself. This may be explained as follows. Duringthe ascent of the sonde the entire mass of liquid, the tempera-ture of which cannot of course be any higher then is consist.entwith the continually decreasing boiling-point, must dissipatemore and more of its heat content. This heat helps to promoteevaporation and the power thus made available will be all

the greater, the more rapid the decline of the boiling-point,that is, the higher thc rate of ascent of the sonde. Providedthat the rate of heat transmission tbrough t.be insulationof the boiling-vessel to the relatively cold surroundingsis less than t.he rate at which energy from the heat of the liquidbecomes available, a certain surplus of beat. will always beavailable t.o keep the liquid boiling.

It is found that a relatively small quantity offreon (10 cm"], insulated with 1 cm thickness ofcotton wool, is sufficient to supply the hypsometerduring an entire ascent. The vessel used to containthe liquid is an ordinary radio-valve bulb; the N.T.C.resistor is secured to two lead-in pins through thebulb in such a way as to be completely immersedin the liquid (fig. 3). The resistor, like the air-tempCl'ature measuring element, is part of thefrequency-controlling network of an R-e oscillator.The outer air has access to the liquid through anarrow tube (exhaust stem) at the top of the bulb;a completely open vessel would be unsuitable since,apart from the chance of spilling, it would allowthe freou to evaporate too quickly and so perhapsassume a temperature below the boiling-point. Onthe other hand, access to the surrounding atmos-phere should not be too severely restricted, in viewof the possibility of the liquid overheating.Tests have shown that a pressure variation of

Fig. 3. Hypsometer. This is made from a radio-valve bulbcontaining freon, the boiling-point of which is measured bymeans of an N.T.C. resistor T secured to two lead-in pins in thebulb and completely immersed in the liquid; tbe latter hasaccess to the outer air through a tube (the exhaust stem) atthe top of the bulb.

152 PHILIPS TECHNCAL REVIEW VOL. 16, No. 5-6

5 millibars corresponds to a variation of the boiling-point such as to produce a relative variation of0.2 % or more in the oscillator frequency.

The measurement of humidity

In the Philips radio sonde, as indeed in severalothers, a piece of gold-beater's skin is used as ameans of measuring the relative humidity of theatmosphere. This skin is an animal tissue, thechief constituent of which is keratin, and likehuman hair, expands and contracts with changesin the humidity of the surrounding atmosphere.Unfortunately this material is affected by hysteresiswhen it passes from a humid to a dry state, andthen back to the humid state. For this reason theelectrical resistance of a solution of lithium chlorideis employed instead of gold-beater's skin as ameasure of humidity in sondes of one Americanmake. This is a very sensitive method, but one whichrequires alternating current to make it fully effectiveand moreover, involves the use of rather complexmeasuring equipment; hence gold-beater's skin ispreferred in the radio sonde described here, despitethe above-mentioned defect.

The variation in the length of this skin affectsthe width of the air-gap in the magnetic circuitof a coil (fig. 4) and is thus converted into a varia-tion of the self-inductance of this coil. The coil ispart of an audio-frequency L-C oscillator, thefrequency of which varies by approxim ately a factorof 2 for a variation in relative humidity from 10to 100 %.

The audio-frequency oscillators

As we have already seen, mechanical switching isdispensed with in the Philips radio sonde by virtueof the fact that the three audio-frequency signalscorresponding to the three quantities to be measuredare modulated on one carrier wave and transmittedsimultaneously. These signals reach the groundstation receiver continuously, and the response timeof the recording equipment connected to thisreceiver does not have to be very small since onlygradual variations take place in any ofthe quantitiesmeasured.

Accordingly, there is an audio-frequency oscilla-tor corresponding to each measuring element inthe sonde. Since the sensitive elements in the tempe-rature and pressure measuring devices are resistors,the obvious course was taken in adopting R-Ccircuits for the two associated oscillators. As alreadyexplained, the oscillator coupled to the hygrometer

is of the L-C type. The frequencies of the threeoscillators are adequately spaced; hence the threesignals are readily separated in the receiver.

The frequency bands are as follows:1- 3 kc/s for pressure;4- 8 kc/s for temperature;12-25 kc/s for relative humidity.

Fig. 4. Hygrometer. The variation in length of a piece of gold-beater's skin G is converted into a variation in self-indnctanceby virtue of the fact that the skin controls the width of theair-gap in the magnetic circuit of coil S.

The R-C oscillators must satisfy certain reqUIre-ments:1) The input power must be as low as possible

(with a view to the weight of the batteries).2) The number of components should be as small

as possible (with a view to the price and totalweight of the sonde).

3) All the components of the different networks(apart from the N.T.C. resistors, of course)should remain as far as possible unaffected bytemperature and pressure variations.

4) The circuit should be insensitive to battery-voltage variations.To satisfy requirement 1), sub-miniature valves

NOVEMBEH-DECEMBER 1954 RADIO SONDE l53

are employed. These are very small valves designedto operate with battcries; the type used in theoscillators is the DL 67, which has a diameter of7.5 mm and a length of 35 mm. Valves of the sametype are used in hearing-aids, etc., and were origi-nally developed for military purposes, e.g. forproximity fuses. Fig. 5 shows the DL 67, togetherwith a valve of the type DL 41 also employed inthe sonde. The former are wired direct into thecircuit.

87882

Fig. 5. The five valves used ill the radio sonde. The foursmall valves are battery-operated sub-miniatures type DL 67(length 35 mm, diameter 7.5 mm), three of which are employedin the three oscillators and the fourth as a reactance valve.Centre: the battery-operated valve type DL 41 (dimensions20 X 55 mm) used as the transmi ttiug valve of the radio sonde.

In as far as they relate to the L-C oscillator, theother requirements can be satisfied merely with theaid of standard components (of the smallest possiblesize), but the R-C oscillators present a moredifficult problem. The circuit diagram of an R-Coscillator is shown in fig. 6, from which it will beseen that a triple R-C network is employed asfeedback between anode and control grid (this isthe minimum number of components required toproduce the phase-inversion necessary for oscilla-tion). Small carbon resistors and ceramic or minia-ture mica capacitors are used in the networks.

Temperature variations affect the circuit mainlythrough the carbon resistors (and the batteries),the capacitances being virtually independent of thetemperature. However, the effect of such variationscan be minimized by insulating the circuit as tho-roughly as possible from the surrounding air, soas to prevent the transfer of heat.There is a direct connection between the above

and the fourth requirement, which may be explainedin the following manner. During the ascent of thesonde there is a slight decrease in the battery volt-ages, partly owing to the low capacity of the

batteries, and partly to thc fact that these voltagesalso depend upon the temperaturc. To fully appreci-ate the deleterious effect of this decrease, we mustre-examine the diagram shown in fig. 6.If the circuit is to oscillate, the attenuation

caused by the network of capacitors and resistorsmust be less than the amplification effected by thevalve; at the same time, an unlimited increase in theamplitude of the oscillations must be avoided.Normally, then, such an oscillator circuit willcontain either an incandescent lamp or an N.T.C.resistor, so positioned that the positive feedback ofaltern ating current in the circuit decreases withthe amplitude of the oscillations. However, ampli-tude limitation by this method cannot be employedin our circuit owing to the fact that the A.C. poweravailable is very low. Hence we make use of acertain "natural" restrietion of amplitude arisingfrom the fact that the grid of the valve is conductingduring a small part of each cycle. As a result of this,the control grid acquires a negative bias (the gridcapacitor is charged), which reduces the mutualconductance of the valve. As the amplitude of theoscillatious increases, the voltage on the controlgrid becomes more and more negative, until themutual conductance is so reduced that the amplifi-cation by the valve is only just sufficient to compo-sate for the attenuation by the network. The ampli-tude of the oscillations then remains constant.However, during the time that the grid is con-

ducting, a new conducting link is formed between thegrid and the cathode (effective for A.C. as well asfor D.C.), so that in effect an extra resistance Rsis introd uced in parallel with the terminatingimpedance R of the network (jig. 7); hence a slightchange takes place in the frequency of the os-cillator, which is governed by the capacitance andthe rosist.an ees of the network.

ruulII

80601

Fig. 6. Circuit diagram of R-C oscillator, two of which areemployed in the radio sonde. The measuring resistor is drawnin bold lines.

154 PHILlPS TECHNICAL REVIEW VOL. 16, No. 5·6

Although insignificant in itself, this frequencyshift is governed by the battery voltage, since theamount of the extra resistance depends upon thisvoltage. The reason for this as follows. Followinga reduction of the battery voltage, and thus areduction of the mutual conductance of the valve,a relatively smaller negative grid voltage is sufficientto bring about the further decrease in mutual·conductance then required. Moreover, the gridcurrent is also somewhat reduced during this process

t~:I:,:J,:JRs______ :J

Fig. 7. Equivalent circuit; of the grid circuit of the oscillatorvalve in an R·C oscillator, showing the apparent resistanceRB resulting from the flow of grid current. R is the total ter-minating impedance of the filter.

When the grid A.C. voltage is high, the grid currentis, to a first approximation, proportional to it,so that Rs is then constant. This proportionalitydoes not exist however at the low grid voltageswhich occur in the present circuit. In the case hereconsidered, then, Rs depends upon the grid current,that is, upon the battery voltage.

To minimize this effect two measures were adop-ted. Firstly, a high-value resistor was included inthe screen grid circuit to moderate the effect ofbattery-voltage variations upon the screen gridvoltage, that is, upon the mutual conductance ofthe valve. Secondly, the control grid was given aninitial negative bias such as to reduce the gridcurrent as far as possible without causing toogreat a decrease in amplification. Since the heatervoltage of the DL 67 is only 1.25 volts whereas theavailable battery voltage is 2.3 V, the required biasis readily extracted from the difference between thetwo by including a resistor in the cathode circuit.

As it happened, the battery effect persisteddespite these two measures to prevent it. However,it. was found that the frequency shift is also gover-ned by the anode resistance, and since, amongstother things, the decrease in the anode-batteryvoltage causes a frequency shift opposed to thatresulting from the decrease in the voltage of theheater battery, the entire effect can be practicallyeliminated by choosing a suitable anode resistance.Experiments have shown that in this way the

frequency shift caused in each oscillator by a decrea-se of about 10% in the two battery voltages can belimited to less than 0.2%.

In practice the decrease in voltage during anascent is invariably less than 10% 2) and, as willhe seen from the accuracy requirements alreadydefined for the pressure, temperature and humidity,the frequency shift is then sufficiently small.

The transmitter

Itwas at first intended to employ 1 metre as thewavelength of the high-frequency carrier to bemodulated by the three audio-frequency signals,The position of the sonde could then be determinedat regular intervals during the ascent by means ofa radio-theodolite in order to measure wind-veloci-ties, as was, in fact, originally the principal objectof such balloon ascents. However, since the bearingsso obtained become rather inaccurate when once thesonde has travelled any considerable distance fromthe ground-station (>50 km), and also owing to thedifficulty of designing a simple, one-valve high-frequency oscillator suitable for modulation at sucha short wavelength, the idea of position-finding wasabandoned for the time being and a wavelength of11 metres (28 Mc/s) was adopted for the carrierwave (this wavelength is employed in several otherradio sondes). The battery valve type DL 41 (seefig. 5) is used for the transmitter, since it suppliessufficient power for our purpose (about 100 mW).

Fig. 8 shows the circuit diagram of the high-frequency oscillator; the aerial is a conducting sleeveone half-wavelength in length round the wire onwhich the sonde is suspended from the balloon. Inprinciple, either amplitude modulation or frequencymodulation could be employed; the latter invariablynecessitates the use of an intermediate stage, where-as, in theory at least, the former enables the modula-ting voltage to be applied direct to the control gridof the transmitter valve. However, in the case hereconsidered the output power of the oscillators isinsufficient for direct application, and since for thisreason alone it is essential to include an intermediatestage, frequency modulation is preferred, by reasonof the fact that the signal-to-noise ratio is very muchmore favourable. Accordingly, this method ofmodu-lation was adopted.The three signals of the "pick-up" oscillators,

separated by intervals between the frequency bands,are mixed in a potentiometer circuit (see fig. 8)

2) In fact, the batteries are accumulators, specially designedfor use in radio sondes, the voltage of which dependsonly very slightly upon temperature.

NOVEMBER-DECEMBER 1954 RADIO SONDE 155

Fig. 9. a) Electronic section of the radio sonde. The three oscillators are mounted onthc left-hand panel and the transmitte.r on the right-hand panel. b) In the assembledsonde the two panels lie back to back and are enclosed in a cardboard cover. Themeasuring elements and batteries are connected to the panels by flexible leads.

and are applied to the grid of a "reactance" valve(again a DL 67), the principle of which has alreadybeen described in this Review 3); the effect of such avalve is the same as that which would be produced

5/'./01

Fig. 8. Circuit diagram of the high-frequency oscillator con-taining the reactance valve. The three audio-Frequency signalsare mixed in a potentiometer and applied to the grid of thereactance valve B. The frequency of the H.F. oscillator 0 isgoverned by the voltage on this griel. The oscillator is con-nected across a capacitance C to the ),/2 aerial A. Thefrequency of the oscillator can be adjusted by means of trim-mer T just before the ascent of the sonde.

by a capacitance or self-inductance (in this case acap acitan ce) in parallel with thc high -frequencyoscillator circuit, its value being governed by the

3) For exemple, see Th . .J. Weyers, Frequency modulation,Philips tech. Rev. 8, 42-50, 1946 (particularly page 47).

80598

a

negative voltage on the grid of the valve. Thisvoltage, which varies at audio frequency with thefrequencies of the threc oscillator-signals, thus modu-lates the frequency of the oscillatory circuit.

One factor affecting the accuracy of the meteoro-logical data as received on the ground is the frequencysweep of the transmitter; the greater this sweep, thestronger the audio-frequency signal in the receiver.In this case, a maximum frequency sweep of 25 kc/swas adopted as being the most consistent with theaccuracy requirements imposed and the possibilitiesof the circuit.

Constructional details of the sonde; features of theground-station equipment

The high-frequency section is completely screenedfrom the remainder of the sonde, to eliminate possib-le interference; the carrier-wave frequency can bechanged slightly just before an ascent by means ofa small trimming capacitor provided for the purpose.The transmitter and the oscillators are together

accommodated in a robust cardboard cover; thisgreatly increases the probability that they will berecovered intact. Fig. 9 shows the interior structureof the sonde.A brief description of the ground-station receiving

equipment will suffice. In principle, this may com-prise a frequency-modulated receiver with threeaudio-frequency bandpass filters, one for each ofthe quantities to be measured, and three frequencymeters to determine the frequencies of the signalspassed by the filters. Preferably, the meters usedto measure the frequencies should be ofthe recording

b

156 PHILIPS TECHNICAL REVIEW VOL. 16, No. 5-6

type. Fig. 10 shows the recervmg equipment atpresent installed in the Royal Meteorological Insti-tute at De Bilt.

Fig. 10. Photograph of the ground-station receiving equipmentas installed in the Royal Dutch Meteorological Institute. TheF.M. receiver is seen at the centre, and the recording meteron the right. At the extreme left is the calihrating equipmentwhich supplies a constant comparison-frequency. Calibrationis preferably effected with the aid of a cathode-ray oscillo-graph.

Testing the radio sonde

To test the accuracy of the radio sonde describedin this article, the Dutch Meteorological Institutecarried out several "twin-ascent" trials, that is, theysent up two such sondes suspended from one balloon,and compared the temperatUl'e and pressure datatransmitted by these instruments. It was found asa result of these trials that the pressure measure-ments of the two sondes agreed to within 5 millibarsup to an altitude of 5000 metres, and that thedifferences at altitudes up to 10000 metres did notexceed 10 millibars. The temperatures so measuredagreed to within 0.5 DC.Although the number of observations carried out

during these trials is too small to justify any definiteconclusions, the results show that the pressuremeasurement is at least as accurate, and the tempe-rature measurement about twice as accurate, asthose quoted in the literature for other sondes.Moreover, it is found that the results produced bythe hypsometer at the tempel'atures occurring with-in its practical range of operation are entirelyindependent of the temperature: hence the pressuremeasurements of the sonde do not require correctionto compensate for possible temperature fluctuations.Again, it appeared from the twin-ascent trials

that the new sonde is very convenient to handle,and that it is not easily damaged. Another importantfeature of the sonde revealed by these trials is itsrelative lightness; the combined weight of two ofthe new sondes is still less than that of one of thetype hitherto employed.

A number of the new instruments were recoveredafter ascents, and there were very few instances ofdamage to the electronic equipment contained inthem. In fact, one of the advantages of the newsonde is that the ternperature and pressure calibra-trions are not affected by damage other than to themeasuring resistors, since they are governed solelyby the electrical properties of the sonde. This wasverified by re-calibration of the sondes recovered.

Another point worth mentioning in connectionwith the calibration of the new sonde is that this canbe accomplished without exposing the instrumentas a whole to the effects of cold at the low calibra-ting temperatures. It is sufficient to plot the resis-tance v. temperature curves of the resistors, and thefrequency curves of the oscillators as a function ofthe partienlar values of the variable measuring ele-ments.

Summary. The instrument described as a "radio sonde" isused as a means of observing physical conditions in the atmos-phere; in it, elements for measuring ternperature, pressure andhumidity are combined with a short-wave transmitter emittinga signal which contains the information gathered by the threeelements. All this equipment is attached to a balloon, which isthen sent aloft. Such a sonde must satisfy a number of string-ent requirements, the most important of which are that itshall be accurate, light and inexpensive. The sonde describedin this article differs from earlier types in that it transmitsthe three items of information simultaneously and so requiresno switching mechanism. Each measuring clement controlsa small oscillator, the frequency of which is influenced by theparticular variable to be measured. Pressure-variations aretransmitted in the frequency band between 1 and 3 kc/s,temperature-variations in the band between 4 and 8 cis, andvariations in humidity in the band between 12 and 25 kc/s.To minimize the number of moving parts required in the sondean N.T.C. resistor, included in the appropriate R-C oscillatorcircuit, is employed as a thermometer, and the pressure isascertained from the boiling-point of freon, which is likewisemeasured by an N.T.C. resistor. The relative humidity isestablished with the aid of a piece of gold-beater's skin, varia-tions in the length of which affect the self-inductance of a coilforming part of an L-C oscillator. The description of theR-C oscillators includes an account of bow the frequencies ofthese oscillators are rendered independent of the batteryvoltage, which decreases to some extent during the ascent ofthe balloon. Sub-miniature valves are used in the oscillators;altogether, the sonde contains four of these valves. Frequencymodulation is employed, the wavelength being 11 metres andthe maximum frequency sweep 2S kc/so The total outputpower of the transmitter valve (the battery-operated DL 41)is about 100 mW. Particulars of the mechanica! design of thesonde and the principles of the ground-station receiving equip-ment are given in brief. Finally, the results of a number ofpractical tests on the new radio sonde are described.

NOVEMBER-DECEMBER 1%4 157

SEASONING OF SODIUM LAMPS

The photograph shows a group of sodium lamps during the "seasoning" process - a stage in the manufacture during whichthe sodium is distribnted in droplets along the discharge-tube, in order to ensure a uniform brightness in the finished lamp.