Altitude-referenced Atmospheric Measurementsusing a Rocket

E80 Section 1 Team 3Siddarth Srinivasan, Emily Swindle, Michelle Lanterman, Duncan Crowley

4 May 2016

Abstract

In this paper, a method for making experimental atmospheric measurements refer-enced with altitude using a rocket with a measurement payload is examined. Flightpath and altitude measurements from an Inertial Measurement Unit (IMU) and Abso-lute Pressure Sensor are compared with expectations from OpenRocket simulations. Intwo launches with a G79W rocket motor, we reached consistent altitudes of 480±5 and506±5 meters which closely matched our simulated apogee heights of 510 meters. Usingthis IMU, the flight path of the rocket was simulated with reasonable error, yielding a436 ± 1000 meter apogee for that first flight. Since a consumer grade inertial measure-ment unit was used, it was confirmed with previous literature that angular measurementnoise was the greatest contributor to rocket position uncertainty. In addition, pollutioneffects of rocket launches as well as weather conditions are explored through the use of adust sensor and humidity sensor. Through a method of integration, more dust particlesare observed on the descent than the ascent of the rocket. Humidity is found to increaseas a function of altitude; however, no quantitative relationship is found because of poorresponse time in capacitive relative humidity sensors.

Introduction and Goals

Rockets can be outfitted with sensors andused to make useful atmospheric measure-ments. For our experimental engineeringcourse, we built a rocket equipped with sen-sors with the engineering goal of tracking theposition of a rocket in space throughout itsflight, and the scientific goal of assessing anyatmospheric impacts of a rocket launch. Totrack the rocket in space, we used an inertialmeasurement unit (IMU) system in combina-tion with an absolute pressure sensor whichmeasured altitude as a function of pressure.For our scientific goal of measuring the at-mospheric impacts of a rocket launch (suchas pollution), we measured dust concentra-tion and relative humidity as a function ofaltitude before and after a great number ofrocket launches had been conducted. In thisreport, we describe the sensors used, theirrespective calibration techniques, our rocketconstruction method, flight modeling, and

discuss the results from our rocket launches.

Sensors

We equipped our rocket with four sensors tomeet our scientific and engineering goals.

IMU

Our IMU has 4 accelerometers (X, Y, Z, HighG) to measure acceleration in the the threespatial directions and 3 gyroscopes to mea-sure the rocket’s rotation rate about threeaxes. We used an IMU to track the trajec-tory and altitude over the rocket’s flight.

Calibration: We used the same method ofcalibration as we did in the Accel & Gyrosrotation lab. First, we calibrated a turntableto verify that we could control it’s RPM withprecision. Then using 3 different IMU ori-entations, we experimentally determined therelationship between the acceleration experi-

1

enced by the accelerometers and the voltageoutput, as well as the rotational velocity ex-perienced by the gyroscope and its outputvoltage. Table 1 shows a selection of thecalibration values we found, with the otherlow-G accelerometers and the other gyroshaving similar calibration constants to thedisplayed X Accel and Z gyro respectively.The slopes match the datasheet closely, al-

though the

Having determined the relationship be-tween the voltage output of the IMU andthe acceleration and rotation it experiences,we could place the IMU on the rocket andnumerically integrate the data to obtain therocket’s position and velocity over time. [1]

Sensor Literature Calibration Curve Measured Calibration Curve

X Accel V = 0.031 V/(m/s2) + 1.5 V V = 0.03 V/(m/s2) + 1.641 V

High G Accel V = 0.00388 V/(m/s2) + 1.5 V V = 0.002094 V/(m/s2) + 1.639 V

Z Gyro V = 0.1146 V/(rad/s) + 1.5 V V = 0.119 V/(m/s2) + 1.513 V

Table 1: Representative selection of calibration data for accelerometers and gyros on IMU

Pressure Sensor

We also used a pressure sensor to measurethe altitude of the rocket through its flight.The relationship between pressure and alti-tude is given by the expression:

p = 101325(1 − 2.25577 · 10−5h)5.25588 (1)

where p is the air pressure and h is the al-titude above sea level[3]. This allows usto convert pressure data collected from therocket flight to altitude. We can also com-pare the altitude data from the pressure sen-sor to that of the IMU to confirm accuracy

in our measurements.

Calibration: The pressure sensor outputsvoltage that varies linearly with absolutepressure. To calibrate the pressure sensor,we used a hand vacuum pump and a high-accuracy pressure sensor to check the con-stants of the linear variation. We found thatthe voltage output by the pressure sensorvaried with pressure P as V = 0.044 V/kPa ·P − 0.415 with R2 = 0.999, compared to theliterature sensitivity of 0.045 V/kPa and off-set of 0.02 V. Figure 1 shows the pressuresensor calibration apparatus.

Figure 1: Pressure Sensor Calibration Apparatus with two sensors to compare readings.The pressure sensor read 91.0 kPa on the ground, compared to the 91.5 ± 0.2 kPa readingfrom the weather station.

2

Dust Sensor

After some research to identify the productsin rocket exhaust and the variety of sen-sors available, we concluded that it would beinteresting to study how dust concentrationchanged as a result of rocket emissions. Con-ventional dust sensors use an infrared (IR)light and phototransistor unit to measurelight scattering as a result of dust, but arehighly sensitive to vibration and were not ap-propriate for use on our rocket. Thus, in theinterest of measuring dust in a high vibra-tion environment, we made our own infrareddust sensor using the same science as a con-ventional sensor with our own low vibrationmount. Figure 2 shows a cross-sectional viewof a PVC tube through which dust travels,with the IR diode mounted perpendicularlyto the IR photo-transistor in a black PVCtube to minimize light reflection.

The LED used was powered from theoutput of a unity-gain buffered op amp us-ing a 100 Ω resistor to provide a consistent

50 mA current. The IR phototransistor out-puts current as a function of light intensity(W/m2), so we convert this into voltage us-ing a transimpedance amplifier with an op-amp. Transimpedance amplifiers work as anideal current-to-voltage converter since theyhave the low loading effects and low out-put impedance necessary for our datalogger.Once assembled, we calibrated the restingvoltage reading to be in the 1.5 V rangewhen there was no dust to allow for maxi-mal gain while still avoiding the saturationat 3.3 V.

Calibration: To test the operation of thedust sensor, we blew large particles of dustthrough the sensor to observe the increasein voltage corresponding to a particle pass-ing through. As expected, we observed thatlarger dust particles resulted in larger voltageblips, and it was also observed that the tran-simpedance amplifier added no additionaltime constant to the negligible time constantof the phototransistor on the order of 100 ns[4].

Figure 2: Our Dust Sensor Design (left) Compared with a Conventional Miniature DustSensor (right) [2]

Humidity Sensor

We use a humidity sensor to measure the Rel-ative humidity (RH), the ratio between the

actual vapor density and the saturation va-por density [5]. Because water is modeled as

3

Figure 3: Picture of humidity sensor calibration apparatus

an ideal gas, we expect the relative humidityto change as a function of altitude becauseof the temperature, pressure, and the pres-ence of cloud or fog. When considering ahumidity sensor to use, we took into accountboth the sensitivity and the cost. The sensorwe chose was the Honeywell HIH-5030-001Humidity sensor. Although it had a slow re-sponse time of 5s in slow moving air, it hada near linear output voltage, an accuracy of±3% RH, low current draw, and a low-costof about $8, making it the best choice givenour budget.

Calibration: To calibrate the humidity sen-sor, we used the Crane Drop Cool MistHumidifier along with ann Omega humidityprobe. By recording the RH read by the hu-midity probe and the voltage output fromthe HIH-5030-001 sensor, we were able tomake a linear voltage calibration curve forthe sensor. The setup for the humidity cali-bration is shown below in Figure ZZ. Thisresulted in a linear relationship of RH =40.35(Vsupply) − 10.12 with R2 = 0.99. Thesensor also accurately read 31.56% RH onground compared with the 32 ± 0.5% RHreading from the weather station brought.

Figure 4: Our Rocket next to a yard stick for size Comparison

Hardware Design

Rocket

We built our rocket using the Aerotech Ar-reaux rocket kit, and modified it to suit ourdesign requirements. The rocket body (con-

sisting of the payload tube and the bodytube) was extended so that the PC boardand motor could fit since the generic rocketbody was too short. After measuring the di-mensions of the PC board, we constructedthe rocket body with a 38.5 cm payload sec-

4

tion and a 57.9 cm body tube. Both rocketbody sections were fiberglassed by rollingwith epoxied fiberglass sheet. Once the fiber-glass had dried, the rough areas on the fiber-glass were sanded using damp sandpaper tocreate a smooth surface. The holes for thefins and launch lugs were cut out using x-acto knives and sandpaper. The fins andlaunch lugs were attached with superglueand then any gaps between the fin base andbody tube surface were sealed with a cementmixture. Once all sealants and adhesiveshad dried completely, the rocket was spray-painted with multiple layers for even color-ing. Finally, the shock cord was connectedto the motor tube within the rocket, andthe parachute was tied onto the shock cordto complete the general assembly. Figure 4shows an external view of our fully assembledrocket.

Dust Sensor Apparatus

A proof-of-concept infrared dust sensor wasbuilt using an infrared LED mounted nor-mal to an infrared photo-diode (which gener-ates current proportional to IR light power)across a rigid black PVC tube. When dustor any small particle obstructs the beam oflight from the LED, light is scattered ontothe photodiode causing a small voltage spikeproportional to the size of the dust and thedensity of dust in the air. Since the orienta-tion of the LED has a great impact upon thereading of the photodiode, both componentswere held in place with a generous coat ofhot glue to minimize position shifts as a re-sult of rocket vibration. This tube was thenthreaded through a set of semi-circular rigidfoam cut-outs that was glued to our acrylicrocket divider. To get continuous airflow,the dust sensor tube was snaked through ournose-cone with holes cut on both sides to anoutput hole near the bottom of the payloadtube facilitated by a bend of flexible siliconetubing.

Humidity Sensor Apparatus

The HIH-5030-001 humidity sensor, like thedust sensor needed to be open to air-flow.Unlike the dust sensor however, the humiditysensor is not sensitive to light and could bemounted on the outside of the payload tube.To securely attach the HIH-5030-001, theleads of the sensor were soldered to a lasercut PCB mount. This was then cementedto a small, rectangular cardboard box whichallowed the hole drilled to be minimized at3/16” for the wires to pass through the pay-load section. The flexible wires soldered ontothe mount were threaded through this holeand the cardboard box was cemented to thepayload section and reinforced with hot glue.

PC Board and Electronics

To take measurements while our rocket flew,we used a measurement payload mountedon a printed circuit (PC) board as shown inFigure 5 which recorded each sensor’s elec-trical signals using a low-power MuddLoggv4 datalogger.

Figure 6 gives the circuit diagram webuilt on the PCB. Note that each sensor out-puts to a MCP6004 quad op-amp using unitygain buffers to impedance match the sensorswith the low input impedance of the Mud-dLogg. For power, a lithium ion 9V batteryis used to accommodate the 150 - 200 mAhcurrent draw of the circuit which mostly re-sults from the two voltage regulators. Forthe sensors not on the PC board, (such asthe dust sensor apparatus) flexible strandedwires were soldered from the PC board to arow of header pins and connectors (from thesensors themselves) and given a coat of hotglue to avoid shorting. When mounted in therocket, these connectors were given a wrap oftape to avoid accidental disconnects. We alsoattached electrical tape to the bottom of thePC board to facilitate smooth sliding acrossour acrylic rocket divider.

5

Figure 5: PC Board Rocket Sensor Payload

Figure 6: PC Board Rocket Sensor Payload

6

Modeling and Simulation

OpenRocket Simulation

OpenRocket was used to predict the flightpath and characteristics such as apogee andburn time. Prior to the actual launchdays, simulations were run with the genericweather inputs to give a general idea of thelaunch. After the actual launches, the ac-tual weather conditions could be used to cre-ate a more accurate simulation of the flight.This is used to confirm that data collectedduring flight follows what is expected. Sinceno meaningful data was collected on the firstlaunch day, we only needed to modify theG79 simulation to compare with data col-lected on the second launch day as shownin Figure 7.

Pressure

On earth, air pressure results from the grav-ity of the column of air above it. Sincethe column of air above changes size as al-titude changes, pressure is a function of al-

titude except for sections of the sky (mostlyin the stratosphere) where pressure is con-stant. This relationship is further elaboratedby Figure 8. To make this conversion frompressure to altitude, we then simply use aderivative of the ideal gas law know as thehypsometric equation:

h2 − h1 =RdT

gln

(p1p2

)(2)

using Rd = 287.04 J/(kg K) and g = 9.8m/s2 (with T being temperature) which re-lates the thickness of a layer of air to the dif-ference in pressure across hypsometric equa-tion. This model assumes constant tempera-ture which we know is incorrect, but since ab-solute temperature (from Kelvin) varies lessthan 7 C or 2.5% below 2000 ft, we can sim-ply use the temperature at the mean altitudeof flight from Open Rocket simulation. Thistemperature is found using the lapse rate onthe day of launch shown in Figure 9 with anoffset adjusted for the ground temperatureat the time of launch.

Figure 7: Launch simulation using OpenRocket

7

Figure 8: Standard Atmospheric Variation of Pressure with Altitude [6]

Figure 9: Temperature vs. altitude on launch day from 4 thermocouples [7]

8

Launch

Motors

The G79W, H135 and G125 motors were al-ready mostly assembled. Our main task forassembly was determining what our optimalburn time was and adjusting the motor tofit this constraint, and we found this usingOpenRocket simulations. We found the op-timal time to apogee for the H135W, G125T,and G79W to be 12.1, 10, and 8.69 secondsrespectively. The initial burn time for eachof the motors was 14 s, 14 s, and 10 s forthe H135W, G125T, and G79W. We were re-stricted to subtracting motor burn time byone second intervals. Thus we set our burntimes to be 12 seconds for the H135, 10 sec-onds for the G125, and 8 seconds for the G79.

Video

A video camera was mounted along the sideof our rocket for the two launches on the first

day, and the final launch on the second day.For the first two launches, the camera didnot record video, thus we spent time trou-bleshooting the issue and concluded that weshould try out another team’s camera for thelast launch and were able to collect footageof our rocket’s flight.

GPS Validation

On launch day, we recorded the starting andending points of the rocket flight to com-pare the actual rocket path to those of ourIMU simulations. As can be seen in Fig-ures 10 and 11, both rocket flights with theG79W motor moved from the point at lefton the launch pad a considerable distance,with the windy first flight resulting in a muchgreater distance traveled than the calmer sec-ond flight. This makes sense since the greaterwind should cause the rocket and parachuteto drift farther on the way down.

Figure 10: Flight 1 traveled 240± 2 m when flown under 10± 4 mph winds from the WSW[8]

Figure 11: Flight 2 traveled 882 m when flown under 62 mph winds from the W [8]

9

Results

We launched our rocket a total of four times,twice on two days. Our rocket launched andlanded safely each time. However, we onlygot meaningful results on the Day 2 Flight 1.On Day 1, our datalogger stopped recordingduring the first flight because the SD cardpopped out. We swapped it out a differentdatalogger which used a tray to store the SDcard, and we were able to collect data forflight 2 on Day 1. However, our humiditysensor was not ready, our pressure sensormade erroneous measurements because itspositioning too close to the vent hole causedit to measure much lower pressures thanreality giving unrealistically high altitudereadings. We did obtain good measurementson Day 2 flight 1, which is analyzed below.Unfortunately, on Day 2 flight 2, there wasan unusual reversal of acceleration halfwaythrough the burn which we attribute to theIMU sliding insiding the payload section, thedust sensor got disconnected so it read zerovolts during the entire flight, and the humid-ity sensor probably got damaged from the

previous flight on landing, because it did notgive any meaningful results. All our analy-sis below is done on data obtained on Day2 flight 1 of rocket launches, unless statedotherwise.

Pressure Sensor

With the absolute pressure data, we con-verted the voltages to pressures using thecalibration curves and then used the hypso-metric equation described in the Modelingsection to convert these pressures into alti-tudes above ground level to get the flight pro-file. For both flights that our rocket madewith the G79W motors, the converted alti-tude profiles are shown in Figures 12, withboth rockets reaching apogee in ∼7.4 secondsand landing after ∼45 seconds. Though bothlaunches reach apogee faster than the pre-dicted 10 seconds from OpenRocket, we findthat the predicted apogee of 510 meters isclose to the experimental apogees of 506 mfor Flight 1 and 480 m for Flight 2.

Figure 12: Flight 1 (left) traveled 240 ± 2 m when flown under 10 ± 4 mph winds from theWSW, Flight 2 (right) traveled 88±2 m when flown under 6±2 mph winds from the W [8]

10

IMU

With the IMU data, we assume that therocket is absolutely vertical while sitting onthe launch pad since the IMU has drifted sev-eral millivolts in the offset after calibration.Knowing this, a noise-free second of on-the-pad data is averaged (with standard devia-tion) for each sensor to make new offsets foreach launch, taking into account 9.8 m/s2 ofgravity for the two downward sensors. Usingthese offsets, we can easily convert voltagesto rotation rates and accelerations through-out the flight times found from the pressuresensor in Matlab. For the upward direc-tion, our IMU has two accelerometers of dif-ferent sensitivities, and we chose to use themore sensitive one whenever the accelerationwas below the maximum (3.3V) reading ofthe datalogger for more accurate readings.Then, we rectify the data into the groundframe by multiplying by an orientation ma-trix C(t) made from the rotation measure-ments, where we assume that the startingorientation is upright, making C(0) = I orthe identity matrix. Then, we find the valueof this rotation matrix through time step bystep using C(t+ t) = C(t) +C(t)∗, where wecan use the small angle approximation usingthe rotation rates ωx, ωy, and ωz to make δψfollowing the methods of Oliver Woodman inAn introduction to inertial navigation [11].Essentially, we use:

Ω(t) =

0 −ωbz(t) ωby(t)ωbz(t) 0 −ωbx(t)−ωby(t) ωbx(t) 0

(3)

to get the speed of transformation, andmultiply this by the time step to get thetransformation angle where δψ = δt · Ω(t).Then, we multiply our accelerations in thebody frame by this transformation to get ac-

celeration in the ground frame, so aground =abody · C(t) and subtract gravity (∼9.8 m/s2

at 855 m ground altitude) from the down-ward direction. Once this is done, we checkto ensure all of the on-the-pad accelera-tions now average about zero, since offseterrors propagate quickly through integra-tion. Then, to get from acceleration to po-sition, we numerically integrate the accelera-tion matrix down the time direction using aSimpson’s rule cumulative integration func-tion found online, which we used because itshowed smaller random drifts than triangularintegration in testing, and is known to givebetter approximations for noisy data since ituses a quadratic method [12]. Finally, thisposition data is plotted in 3D to visualizethe resulting flight path, as can be seen inFigure 13.

To understand the error in these positionestimates, we used the methods from Ray-mond Chow’s paper [13] to propagate theerrors for both the accelerometer and gyro-scope caused by offset bias, scaling coeffi-cient bias and random noise. The noise esti-mates for the IMU sensors were taken fromthe sensor datasheets, since those were mostconservative, and once calculated these po-sition errors were added in quadrature foreach point in timeSensorDatasheets. Simi-lar to the findings from O. J. Woodman’spaper [11], the most significant error con-tribution was found to be gyroscope noise,which added up to a ∼1000 meter positionaluncertainty after only ten seconds. Usingthis method, both G77W flights were ana-lyzed, which produced a reasonable apogeeestimate of 436 ± 1000 m after 10 s of flightfor Flight 1. This flight’s IMU altitude witherror is compared with the pressure altitudeto show their reasonable agreement in Figure14.

11

Figure 13: 3D plot of path taken by rocket in Flight 1 from IMU

Figure 14: Processed IMU data with Err. vs. Absolute Pressure Altitude vs. Time forFlight 1

12

Filtering the position data

For the position data from the pressure sen-sor and the IMU, it was predicted that me-dian or mean filtering would lower noise andhave a significant impact on position esti-mates. Nonetheless, after using 5-wide me-dian filters which choose the moving medianof 5 points and varying the number of pointschosen greatly (up to 131-wide), very lit-

tle change in the position estimates was ob-served (only 4 meters after the 10 secondflight). As is shown in Figure XX, this fil-tering seemed to not be tremendously im-pactful. Low-pass butterworth filters werealso examined with little success, showingthat the noise was actually not the most sig-nificant contributor to flight error since theflights were so short.

Figure 15: For both types of position data, 5-wide median filtering caused little change

Dust Sensor

After flight, we found reasonable readingsfrom our dust sensor corresponding to par-ticulates observed in the air. The resultsfrom this reading for Flight 1 can be found inFigure 16. To further understand this datain the spatial dimension, this data was plot-ted against altitude as can be seen in Figure17. One way to quantify this measurementwould be to take the area under the curve.The presence of dust will only ever increasethe voltage from the baseline offset voltageof 1.55 V, so the integration will always givea positive result. The ascent gives an area

of 9.8777 Vs and the descent gives a voltageof 15.6661 Vs. This shows us that the sen-sor read more counts of dust on the descentthan the ascent. We believe this may havehappened for one of two reasons; either thespeed of the dust sensor on the ascent causedthe readings to be more general showing lay-ers of fog or dust at different heights, or thedescent blew the rocket into the path of pre-vious rocket emissions where dust was mostconcentrated. Unfortunately, our clips forconnecting the dust sensor to the PC boardwere not nearly reliable enough for rocketflight and came unconnected for the secondflight giving us only this holistic comparison.

13

Figure 16: Dust Sensor Voltage as a function of time over entire flight (top), ascent (middle),and descent (bottom)

Figure 17: Dust Sensor Voltage as a function of Altitude over ascent (top) and descent(bottom)

Humidity Sensor

For the humidity data, we imported thehumidity sensor voltage values versus timedata files into Matlab and converted thevoltages into humidity values using the ex-perimental sensor calibration described inthe Ground-Based testing section, ensur-ing that the ground humidity readings werewithin ±0.5% of the weather station read-ings. When converting the voltages to hu-midity readings, we had to consider thatthe HIH-5030-001 sensor output is depen-dent on the temperature such that True RH= Sensor RH

1.0546−0.00216T [9]. The temperature as afunction of altitude was given above in Fig-ure WW. By performing a linear regressionon this data from 0-2000m we were able toobtain a linear relationship of T = −0.003 ·Altitude + 17.88 with R2 = 0.999. We thenimported the altitude values versus time ob-tained from the processing of the absolute

pressure sensor voltage readings. Taking intoaccount the effect of temperature on the sen-sor reading, the humidity as a function ofboth altitude and time are shown in Figure18 for both flights that our rocket made withthe G77R motors.

When analyzing the data from the firstflight, the humidity over time increases.However, the rocket flight had already endedat about 50s where the humidity starts de-creasing. We expect the humidity to startdecreasing after the rocket reaches apogee ataround 10s and starts descending because thehumidity increases as the rocket ascends. Al-though the humidity does eventually returnto the ground humidity at about 32%RH,this doesn’t occur until well after the rockethas landed. This indicates a slow responsetime of the sensor. Assessing the humidityover altitude confirms this intuition. Duringthe ascent of the rocket, represented by thered part of the curve, the humidity is increas-

14

Figure 18: Data from flight 1 shows the slow time constant of the humidity sensor

Figure 19: Data from flight 1 shows the slow time constant of the humidity sensor

ing as expected from the graph of humidityover time. However, it continues to increaseas the rocket descends, represented by thedark blue section of the curve. The verticaldrop in the graph at the end of the descentwhen the altitude is 0m indicates that the

humidity sensor was reacting to a drop in hu-midity even as the rocket was on the ground.This confirms that the sensor?s slow responsetime caused the late reading in the drop inhumidity which occurred during descent.

15

The reason for this long response timeis that the air velocity is much higher thanthe sensor is sensitive to. The quoted re-sponse time of 5s is only true in slow-movingair, which Honeywell defines as 5m/s or less[9]. The air velocities that the sensor is re-ceiving during the flight are approximately100-200m/s. We experimentally determinedan estimate of about 70s for the responsetime by finding the time for the humiditysensor to reach 90% of the actual humid-ity at a specific time. This confirms resultsfound by Dooley and O’Neal when experi-mentally determining the effect of airspeedon RH reading in capacitive sensors. Doo-ley and O’Neal found that for a RH sensorthat had a response time of 15s for ambientair, the response time in air moving at 6m/shad increased to 47s [10]. This is a 35.5%increase in response time per m/s increase inair velocity. If we use this relationship forairspeeds between 100-200 m/s we expect aresponse time between 160-350s for the sen-sor to read 90% of the actual humidity. Al-though this relationship predicts a slower re-sponse time than we experimentally found,it confirms the result that higher air velocitycauses an increased response time. Our find-ings that humidity increases as altitude in-creases confirms our expectations. Althoughthe decrease in temperature as a result of in-creasing altitude would suggest that humid-ity should decrease, low-level clouds and val-ley fog were present the day of the launch.This would cause the humidity to increaseas the rocket reaches altitudes where theseclouds and fog were present. The humidityreadings with respect to time for the secondflight suggest that the humidity data fromthe second flight was corrupted in some way.The humidity at time=0s shows a negativevalue which is impossible especially whenthe humidity reading from the weather sta-tion brought was around 24% at the ground.This indicates that the sensor may have beendamaged upon impact from the landing of

the first flight, since the sensor was mountedon the exterior of the rocket.

Conclusion

In conclusion, the engineering objective oftracking the flight path of the rocket throughspace was achieved through the use of anIMU and absolute pressure sensor. The ab-solute pressure sensor measured an apogee of506 ± 5 m and 480 ± 5 m for flight 1 and 2respectively. The IMU measured an apogeeof 436 ± 1000 m and using a 6-DOF modelplotted the 3-D flight path of the rocket. Thepressure sensor was found to be a more accu-rate method of predicting apogee when boththe IMU and pressure sensor measurementswere compared to the OpenRocket predic-tion of apogee at 510m.

The scientific objectives of assessing theeffects of a rocket launch on dust concentra-tions and measuring humidity against alti-tude were achieved with less than ideal re-sults. The dust sensor used successfully mea-sured counts of dust on ascent and descentbut without further data points to compareto our readings stand alone. For the humid-ity sensor used, it was confirmed that humid-ity rose going through low level clouds but wealso validated an increase in time constantproportional to rocket speed, making our re-sults not representative of the particular airat any given altitude.

For further work, we would suggest im-proving upon our PC board mount and con-nections to sensors so launches would bemore reliable in producing readings. To pro-duce more representative humidity readings,the humidity sensor should be placed in thestill air inside the rocket to allow for optimaloperation of the capacitive humidity sensor.Finally, the dust sensor should be tested mul-tiple times to see if we are able to reproducethe results, and to determine whether rocketlaunches produce noticeable particulate pol-lution.

16

References

[1] Accelerometer and Gyroscope Calibration Lab.<http://www.eng.hmc.edu/NewE80/AccelGyroLab.html>

[2] Application Note of Sharp dust sensor GP2Y1010AU0F (Sharp Corp., 2005).(http://www.sharp-world.com/products/device/lineup/data/pdf/datasheet/gp2y1010au appl e.pdf).

[3] “http://www.engineeringtoolbox.com/air-altitude-pressure-d 462.html”

[4] 4.8mm Semi-Lens Silicon PIN Photodiode PD438B (Everlight, 2012).(http://www.everlight.com/file/ProductFile/PD438B.pdf).

[5] R. Nave. Relative Humidity (Georgia State University Hyperphysics Database, 2012)(http://hyperphysics.phy-astr.gsu.edu/hbase/kinetic/relhum.html)

[6] Altitude above sea level and air pressure. The Engineering Toolbox. (2014).(http://www.engineeringtoolbox.com/air-altitude-pressure-d 462.html).

[7] E. Spjut, Lucerne Valley Temperature Data. Harvey Mudd College Engineering Dept., (April23, 2016). (http://www.eng.hmc.edu/NewE80/TemperatureLab.html).

[8] Weather and Hazards Data Viewer (NOAA, Salt Lake, 2016).(http://www.wrh.noaa.gov/map).

[9] HIH-5030/5031 Series Low Voltage Humidity Sensors.(Honeywell International Inc., Golden Valley, 2010).(http://sensing.honeywell.com/index.php/ci id/49692/la id/1/document/1/re id)

[10] J.Dooley, D.O?Neal. The Transient Response of Capacitive Thin-Film Polymer HumiditySensors. (HVAC&R Research, 2009). ( http://esl.tamu.edu/docs/terp/2008/ESL-PA-08-09-01.pdf)

[11] O. J.. Woodman, An introduction to inertial navigation University of Cambridge ComputerLaboratory, (2007). (https://www.cl.cam.ac.uk/techreports/UCAM-CL-TR-696.html).

[12] D. Garcia, Cumulative Simpson?s Integration Function (BiomCardio, Montreal, 2006). Web.(http://www.biomecardio.com/matlab/index.html)

[13] R. Chow, Evaluating inertial measurement units. Test and Measurement World, pp. 34 (2011).(http://m.eet.com/media/1166674/26170-tmw 1111 f4 imu.pdf).

[14] BMA145 datasheet tri-axial analog acceleration sensor. Bosch Sensortec, Reutlingen, Germany,(2010). (http://www.eng.hmc.edu/NewE80/PDFs/BMA145(accel).pdf).

[15] 3LPR450AL dual-axis pitch and roll analog gyroscope. STMicroelectronics, (2014).(http://www.eng.hmc.edu/NewE80/PDFs/LPR450AL.pdf).

[16] 4LPY450AL dual-axis pitch and roll analog gyroscope. STMicroelectronics, (2009).(http://www.eng.hmc.edu/NewE80/PDFs/LPY450ALTR.pdf).

17