pmm223 - eCommons@Cornell
Contents
1. When the overall height and the height at which the system ceases accelerating are known the difference can be found A2 which is the height traversed by the system over the time that it sinks h h h 4 This then leads to the sinking time 2 LE 5 Meanwhile the xy plane or drift velocity v 1 is found by accounting for the drag coefficients of both the anchor and float parts of the system en TC Pepa EEE ELLA 6 Ga Now it s possible to find the displacement from the drop point in the xy plane 1 Fi liu Tj 7 D Phase 2 Rising Next the terminal velocity of the float alone must be found for its journey to the surface Ee eae Broo fake 8 ou CaFsrrF Now it s possible to find the drift velocity v 2 of the float alone mum Cre gt EEE 9 Goole Then the characteristic time of the float alone t2 can be calculated Ty zn 10 a p Next the height at which acceleration stops is calculated 15 22 Vig Talo 2 11 And now the height that it travels at its terminal velocity can be calculated h4 h h h 12 The amount of time that this takes 4 can then be calculated TII 13 Finally the displacement from the anchor on the bottom f2 can be calculated Pa Prev ir Tu 14 This means that the total displacement from the initial drop point has a magnitude of P ur acing em Taxy Frar 15 E2. NY pmm223 cornell edu Abstract The dependence of a tuning fork quartz crystal oscillator s frequency f on temperature T is observed over the temperature range 5 to 20 C From this a parabolic f T function is fit to the crystals data and used to compensate for sampling period drift in an Analog to Digital Converter ADC system based around this crystal at various temperatures Resolution and uncertainty of this method are discussed INTRODUCTION A tuning fork quartz crystal oscillator 1s one of a family of devices that vibrate at a given frequency when invested with energy by way of an electric field However internal mechanical stresses coupled with the device s thermal expansion and contraction cause this frequency to vary with temperature The variation can be roughly characterized by a parabolic function f T such 1 that f T k T k2T k3 Each crystal is designed with a stability temperature Ty and corresponding frequency fo near which small changes in temperature result in small changes in frequency Most commercially available crystals have a 70 at around room temperature 20 2 C Operation at temperatures far from results in increasing deviations from f When such a crystal is used to generate the sampling clock for an ADC at temperatures far from the sampling rate will be inaccurate by the same factor Existing methods used to minimize the temperature dependence of the oscillator include a
3. and C Univerrsity System and 45 26 27 28 29 30 31 32 33 34 35 methods for correcting clock synchronization errors no 13 Apr 2013 Martin J Peppin National Voluntary Laboratory Accreditation Program U S Acoustical Testing Services 1994 ANSI ASA S1 15 1997 Part 1 R 2011 Measurement Microphones Part 1 Specifications for Laboratory Standard Microphones pp 1 8 ANSI ASA 51 20 2012 Procedures for Calibration of Underwater Electroacoustic Transducers pp 1 53 S J Cooke S Hinch M Wikelski and D Andrews Biotelemetry a mechanistic approach to ecology in Ecology amp Evolution 2004 Y Ropert Coudert and Wilson Trends and perspectives animal attached remote sensing Attp dx doi org 10 1890 1540 9295 2005 003 0437 TAPIAR 2 0 CO 2 Sep 2008 C Rutz and C Hays New frontiers in biologging science Biology Letters vol 5 no 3 pp 289 292 Jun 2009 M Fedak P Lovell B McConnell and C Hunter Overcoming the Constraints of Long Range Radio Telemetry from Animals Getting More Useful Data from Smaller Packages Integr Comp Biol vol 42 no 1 pp 3 10 Feb 2002 SPOT LLCGlobalStar Inc SPOT Connect Datasheet findmespot com Online Available http www findmespot com en docs SPOT Connect Sheet REV3 July2011 pdf Accessed 28 Nov 2013 Prying into the intimate de
4. and the data from them used with this algorithm to interpret the recorders sinking and rising characteristics If this is successful then an inverse of this method might also be used with IMUS to easily determine current profiles Xl CONCLUSION This algorithm has a number of oversimplifications and perhaps unsupportable assumptions but is designed as a first step towards understanding these dynamics Further work is needed but for the range of instrumentation that currently exists this algorithm should be able to help bring placement uncertainties back under control 25 Characterization of marine autonomous recording units Peter Marchetto Bioacoustics Research Program Cornell Lab of Ornithology and Department of Biological and Environmental Engineering Cornell University Ithaca NY 14850 pmm223 cornell edu Abstract The Marine Autonomous Recording Unit MARU is a common tool used in many underwater passive acoustic surveys As such its output data is used for many research questions from detection and localization of animal calls to measurement of anthropogenic background noise in the ocean In this paper the means of characterizing this system to give the most useful traceable data and its uncertainties is described XII INTRODUCTION The MARU has been used for over fifteen years to collect recordings of marine mammals fish and anthropogenic noise at sea along with many other sounds Lately it has becom
5. hg This means that the first part will be expressed by h4 h where the flow layer is thicker than the terminal velocity distance From here it s a matter of simple vector addition to get ies rete Cmt cose mat cos b rt sing ryt sin dy 21 T _ tan Fala sin irn lr y lay Eyed bos gtep g cua g 22 As can be seen above the addition of flow layers can be accomplished fairly easily and so a multi layer flow either on the sinking or rising part of the model 15 simple IX PROBABILITY MAPPING By multiplying the above displacement vectors by a standard Gaussian probability distribution a 24 mapping of the Probability Density Function PDF of the position of the system can be generated for any time during the modeled interaction An example can be seen below in Fig 2 Mean Displacement Mean Search Area c470 45 v 2 m sec Ca 0 45 v 2 m sec 250 140000 120000 100000 80000 Area m 60000 Displacement m 40000 20000 0 20 40 60 80 100 120 140 160 180 200 t sec Fig 2 Mean displacement left and mean search area right graphs assuming a c of 0 45 and a V Of 2 m sec The error bars show the 10 points in the gaussian distribution X FUTURE WORK At this point in time Inertial Measurement Units IMUs have been deployed in two underwater recording units in Cape Cod Bay MA USA These will be recovered soon
6. the subsystems mentioned above characterizations were developed Some of these such as the acoustic characterizations were based on known protocols while others such as the oscillator characterizations had to be created from whole cloth 1 Crystal f T Characterization The frequency of the crystal oscillator 1s recorded using a 34410A multimeter Agilent Englewood CO USA in frequency mode while the temperature 1s recorded using a K type thermocouple connected to a 1048 0 thermocouple input Phidgets Calgary AB CA The parabolic regression 1s done a custom LabView National Instruments Austin TX USA interface program and its output is stored along with the 8 serial number in a database locally The Device Under Test DUT is placed into a freezer allowed to drop to 5 2 then raised back to 20 C Again this method 15 the same as in 9 and 25 2 Signal Chain Characterization The main board of the MARU is placed in a freezer kept at 0 2 C and connected as seen in fig 3 30 Nexus Preamp Ch1 Ch1 Out APx 525 In Ch2 Ch2 Hydrophone Ek Hydrophone Assembly d Freezer Fig 3 Test system connection diagram The output of the SCCB is split and routed into the 8 for encoding as well as into the second input channel of an APx525 audio analyzer Audio Precision Beaverton OR USA Stimulus sounds are played from the analyzer through a 55 W amplifier XL 55 RadioShack Fort W
7. to the strange anti resonant behaviors of the Metglas family of magnetostrictive amorphous metal alloys under Dr Philip Anderson at RCNJ After graduating with his BS he moved on to work for a calibration company in State College PA writing and performing calibration procedures then a genomics software company in the same town implementing a QA QC program for them for the first time From here he moved on to studying ferroelectric polymers and their piezoelectric and dielectric properties at the Penn State Materials Research Lab as a visiting scientist Peter has been at Cornell University first as a staff member since 2009 then also as a graduate student through the Employee Degree Program since 2010 in the Bioacoustics Research Program at the Cornell Lab of Ornithology Since 2014 he has also been a member of the Soil and Water Lab doing instrumentation design for their experiments Since coming to Cornell he has been focused on design challenges relating to the placement of recording and logging instruments in harsh environments the very work that this dissertation hopes to address 1 For Katie ACKNOWLEDGMENTS Those to whom I owe a great debt of gratitude for their help during the last several years would make a list greater than the length of this entire document so I ll try to be concise First of all like to thank my committee for encouraging me and helping me to bring this work to fruition I d like to thank Harold
8. which to derive critical information relating an individual s position and the environmental conditions to which it is exposed The SPLOT is still being developed and the next step in development 15 to reduce the power budget by powering sensors only as needed and enabling a larger amount of acoustic data 4 to be gathered by adding a comb filter to see what frequency bands have the highest amount of acoustic energy A datalogging version of the device 1s also under development which would allow for logging various parameters to on board memory to be downloaded upon retrieval Development also continues on making a single board solution using a newer smaller satellite modem which would allow for use on many more species having a mass of 0 5 kg VII ACKNOWLEDGEMENTS The authors would like to thank David Winkler for his advice about environmentally hardened biological sensing systems and putting them on animals and Al Molnar for his advice and encouragement Thanks also to Michael Thonney and his staff at the Cornell Teaching and Research Farm for their help with field testing Thanks most of all to the ewes who helped by wearing the SPLOT prototype their part in this work was conducted under Cornell IACUC Protocol 2008 0111 Amendment 0001 42 1 2 3 4 5 6 7 8 REFERENCES A compact digital recording system for autonomous bioacoustic monitoring The Journal of the Acoustical Socie
9. 3 Preliminary stationary data Color indicates incident SPL in dBA as indicated in the legend Fig 4 Track of a sheep from SPLOT waypoints taken every ten minutes The starting points are in red and then progress through yellow green blue and purple During this time the sheep and its flock were driven from the barn at the bottom of the map to the field across the road where they spent most of their time grazing and where the highest concentration of points is The end of the track is on the barn side of the road where the sheep went after it jumped the fence and crossed the road 40 Comparison of Trends 160 4 m 140 3 5 a 120 3 gz 8 100 4 2 5 SPL dB A 50 BO 9 0 lt 60 1 5 kPa sel Z 40 1 C 2 20 0 5 0 2 T 12 6 12 16 04 12 6 12 20 52 12 7 12 1 40 t M D Y H M Fig 5 Comparison of data trends over a reporting period VI CONCLUSION AND FUTURE DIRECTIONS The SPLOT is an effective portable tracking platform that may be used virtually anywhere on the globe to measure environmental parameters Furthermore its eight ADC inputs make it expandable to additional measurements such as system power status light intensity or other environmental parameters While 41 characters is not a lot of space the size and rate of these data tweets are sufficient to provide a robust suite of data by
10. A Anchor m xz yz Plane Cross Sectional Area of Az4 Anchor m Volume of Float L Vr Volume of Anchor L V4 Release to Recovery Time sec toat Acceleration Due to Gravity m sec 0 9 81 Terminal Velocity m sec T Bearing 0 B Outputs Variable Symbol Phase 1 xy Plane Displacement Phase 2 xy Plane Displacement P P Phase 3 xy Plane Displacement VII ALGORITHM This algorithm addresses three phases of xy plane drift while the object 1s in motion in the z axis as seen in Fig 1 20 T1 t T2 t Fig 1 The three phases of the model from left to right sinking with an anchor float system detachment and rising of the float from the anchor and surface drifting of the float by itself C Phase I Sinking First the terminal velocity of the system is calculated Since the maximum terminal velocity 1s what s needed in this model we treat this as being only the anchor Thus we find the first terminal velocity based on the above parameters lijcl cca Same Ppa Te fa a 22 1 rr lI 1 E y BO tua xA 1 Once the terminal velocity is known the characteristic time can be calculated This 15 the time over which the system experiences a logarithmic acceleration ending when it reaches vz Ti zi 2 This 1s then multiplied by the average velocity over the acceleration period v to give the height of the system at the end of its acceleration ty ta Ti 12 3 21
11. ADDRESSING ENGINEERING CHALLENGES IN BIOACOUSTIC RECORDING Dissertation Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy by Peter M Marchetto May 2015 O 2015 Peter M Marchetto ADDRESSING ENGINEERING CHALLENGES IN BIOACOUSTIC RECORDING Peter M Marchetto Ph D Cornell University 2015 Bioacoustics is a very challenging field due to the necessity of putting fragile power hungry pieces of instrumentation out in 1solated hostile environments for long periods of time Making these instruments as durable as possible while also considering weight power consumption size and affordability 1s a constant struggle In this dissertation multiple engineering challenges associated with these environments are addressed BIOGRAPHICAL SKETCH Peter Marchetto earned his Bachelor of Science degree from Ramapo College of New Jersey 2007 and his Master of Science degree upon his successful completion of the A exam at Cornell University 1n 2015 Before any of that he began his life 1n research by working in the applied physics and biomedical engineering lab of Dr Gordon Thomas at NJIT on biomedical instrumentation calibration From there he worked under Dr Robert Fechtner as a research assistant in the ophthalmology department at UMDNJ helping with tonometry and laser tomography studies After that he turned his attention
12. Cheyne for being a fantastic mentor for the last nearly six years Chris Clark for being an absolute inspiration about what it means to be an engineer working in the biosciences David Wink Winkler for his wonderfully encouraging talks with me and of course my advisor Todd Walter for taking me on as a grad student kind of at the last minute and pushing me to finish with a gentle but firm and humorous hand Thanks also to the Cornell Employee Degree Program for funding my tuition My thanks also to the two people responsible for getting me started in the graduate program Wolfgang Sachse and Dan Aneshansley I d also thank my advisors in my previous institutions Gordon Thomas at NJIT 1s single handedly responsible for getting me started physics and instrumentation research not to mention introducing me to liquid nitrogen Similarly Phil Anderson at RCNJ 1s responsible for helping me to become a researcher and for guiding my progress in experimental physics My co workers at the Lab of Ornithology also deserve a great deal of credit especially my compatriots in engineering and deployment Rob Koch Adam Strickhart Karl Fitzke Rob MacCurdy Sherwood Snyder Ed Moore and Rich Gabrielson to name a few The two that stand out the most are Chris Tessaglia Hymes and Ray Mack without their encouragement support and open doors ears and minds this work would not have taken form Finally I d like to thank my family my parents Nancy an
13. Fig 3 Finer resolution view of two candidate events aligned from synchronization 16 B I L H I Uncorrected Clock Drift 700 600 UI c c Drift sec NJ i 100 0 50 100 150 200 t Days Fig 4 Typical clock drift from an average MARU sample clock at f 64 kHz As candidate events are recorded on multiple units throughout an array they create an ensemble of correlations By using an autocorrelation function with settings on the order of the clock drift expected at a date the deployment a series of correlated events can be found and these then used as anchor points for a stretch shrink interpolation method By doing this the true duration of recordings within the array and synchrony of the units clocks might be recovered Detections and Offsets 700000 600000 y 1 000x 2 476 995 500000 _yl 1l000x 2 478 036 400000 Start ti sec X End 300000 Linear Start Linear End 200000 100000 0 100000 200000 300000 400000 500000 600000 700000 t5 sec Fig 5 Detections within the last two weeks of the test Trendlines from the candidate events show a possible drift of up to 1 sec over the course of one week The y intercept is equal to the start delay between the two units adjusted for clock drift 17 Another illustration of the capabilities of this method is in seen Fig 5 In this graph th
14. Phase 3 Floating The time that the float spends on the surface can then be calculated from the overall deployment time and its difference with the rise time tearrace risas Oty T Ta 16 The velocity on the surface 15 then calculated by taking a portion of the drift velocity in the water and of the drift velocity in the air v Lee Do nou sin Boy sin 830 ee Da Pea cas de COS Bey 17 And its bearing from the vector addition of the two velocities TIR TAS sin Pe D rre tan EE Del 18 hye ipa Pea TES cos This can then be used to find the final maximum displacement 3 Puy B Urysariaze ear duce 19 23 And so the final displacement from the initial drop point can be given by Fine Pore ny 20 MULTI LAYER FLOWS In the case of multi layer flows the model assumes that there are sharp transitions from one layer to the next In these cases the post terminal velocity term will be taken and vector addition will be necessary to determine the eventual bearing of the sinking or rising device An example follows below F Two Layer Flow Sinking In this scenario unit is sinking through two flow layers one with a flow velocity of v4 04 and the other with a flow velocity of vg The first layer 15 h4 deep while the second is hg deep where h h
15. al pdf Accessed 20 Feb 201 5 P Ailliot E Fr nod and V Monbet Long Term Object Drift Forecast in the Ocean with Tide and Wind http dx doi org 10 1137 050639727 Aug 2006 N Tsai Analysis of a Free Fall Anchoring System Offshore Technology Conference 197 R Lampitt and M Burnham free fall time lapse camera and current meter system Deep Sea Research Part A Oceanographic 1983 Rascle F Ardhuin and E A Terray Drift and mixing under the ocean surface coherent one dimensional description with application to unstratified conditions Journal of Geophysical Research Oceans 1978 2012 vol 111 no 3 p C03016 Mar 2006 E Bjerregaard and E Sorensen Forces and Moments on Under Water Bodies Obtailled From Wind Tunnel Tests Offshore Technology Conference 1982 B Hackett Breivik and C Wettre Forecasting the Drift of Objects and Substances in the Ocean in Ocean Weather Forecasting Springer Netherlands 2006 pp 507 523 V Chandrasekhar W Seah and Y Choo Localization 1n underwater sensor networks survey and challenges ACM International Workshop on UnderWater Networks 2006 P Marchetto Strickhart R Mack and H Cheyne Temperature compensation of a quartz tuning fork clock crystal via post processing presented at the 2012 IEEE International Frequency Control Symposium FCS pp 1 4 H Cheyne Strickhart P Marchetto R Mack
16. amp it would cause a conductive path to be formed within the bulb allowing a small spike the current through the resistor This would then be recorded by the unit s on board Analog to Digital Converter ADC based on the TattleTale Model 8 Onset Computer Bourne MA USA 15 This data was stored on a CompactFlash card and retrieved after the deployment ended The two units were first exposed to variable temperature environments for a week such that their clocks would drift differently and induce errors in their timing 14 to ADC V Fig l Schematic of the detector circuit is on the order of 1 while V is below the threshold voltage of the neon bulb which is or the order of 90 VDC The two units used were then set next to one another with an interstitial distance of 1 m The recording delay between the two start times was approximately 45 minutes The temperature instability in the sample clocks in both encoders was on the order of 2100 ppm C and the skew before drift was on the order of 173 megasamples After the test both units were unsealed and their data offloaded The data was converted from a raw bitstream to AIFF files at 16 bits and 64 kHz sampling frequency These sound files were then analyzed in Raven 1 5 Bioacoustics Research Program Cornell Lab of Ornithology Ithaca NY USA 16 IV RESULTS AND ANALYSIS A batch detector was run in Raven to find any instances of candidate events Candid
17. ange must include any temperature to which the crystal might be exposed during its use The clock crystals 1n our testing were part of ADC boards Tattletale Model 8 Onset Computer Corp Pocasset MA The crystals were used along with low power oscillator HA7210 Intersil Corp Milpitas with a temperature stability of 0 1 ppm C A 14 VDC rail and a ground rail were attached to eight of the ADC boards on a mounting fixture The temperature of each crystal was taken by a thermocouple input to a USA thermometer board Model 1048 Phidgets Inc Calgary AB The 40 kHz clock output of the ADC boards was input to a DMM in frequency counting mode 34410A Agilent Technologies Inc Canta Clara CA through a signal relay board Model 1048 Phidgets Inc Calgary AB A virtual instrument created in LabView National Instruments Austin TX was used to log data from the f T curve and was also used to control the system using the relay board The ADCs were attached to the mounting fixture and the program started The first few readings were taken at room temperature 20 C then the fixture was lowered into a freezer Model FFFCOSM2KW Frigidaire Martinez GA and remained there until its temperature stabilized around 10 2 C At this point it was removed and allowed to return to room temperature in order to characterize system hysteresis The second phase of the process requires measurement and recording of the crys
18. are two things that one can be certain of when placing recording devices in the outdoors not all of them will come back unscathed and not all of them will function as planned In between the times when they re placed and retrieved these can go through some strange conditions which is especially true of marine recording devices During placement they can drift with currents While sitting at depth they can chill down to temperatures that affect the uniformity of their sample clocks Over time their components can age at different rates and there s always the chance of variability from unit to unit of sensitivity There s also a need to measure environmental variables in the context of noise level so as to find out what the properties of the medium through which sound is traveling are Another problem 15 that of localization when trying to use Time Delay of Arrival TDoA methods to ascertain the origin of a signal across an array of receivers several sources of uncertainty crop up The first and largest of these 15 that of clock drift As stated above recorders are usually in variable temperature environments and so wind up with a large temperature induced shift in clock rate This can be on the order of several Hz which seems small compared to clock frequencies in the tens of kHz but when the recorder is shy a few Hz for several months temporal drifts in the recording on the order of minutes may occur and these drifts will not be uniform a
19. artz tuning fork clock crystal via post processing presented at the 2012 IEEE International Frequency Control Symposium FCS 2012 pp 1 4 H Cheyne A Strickhart P Marchetto and R Mack System and methods for correcting clock synchronization errors M S Gordon P Goldhagen K P Rodbell T H Zabel H H K Tang J M Clem and P Bailey Measurement of the flux and energy spectrum of cosmic ray induced neutrons on the ground Nuclear Science IEEE Transactions on vol 51 no 6 pp 3427 3434 2004 D K Bailey Abnormal Ionization in the Lower Ionosphere Associated with Cosmic Ray Flux Enhancements Proceedings of the IRE vol 47 no 2 pp 255 266 1959 M Friedlander and F Jones A thin cosmic rain particles from outer space American Journal of Physics 2001 Visual Communications Company LLC Neon Indicator Lamps 7 amp 8 veelite com Online Available http vcclite com pdf Neon 20Indicator 20Lamps 7 amp 8 pdf Accessed 20 Feb 2015 Onset Computer Inc Tattletale Model 8 Installation and Operation Manual onsetcomp com Online Available http www onsetcomp com files manual pdfs TT8C Man pdf Accessed 03 Nov 2011 R A Charif and L M Strickman Raven 1 4 User s Manual birds cornell edu 2010 44 17 18 19 20 21 22 23 24 25 Online Available http www birds cornell edu brp raven Raven14UsersManu
20. asured Rising Hz f Fitted Hz 9996 8 39996 6 10 0 10 20 30 T C Fig 5 Sample graph of the f T characterization The frequency of the clock while cooling can be seen in the blue graph while rising can be seen in the red The fitted curve of f T 0 001521T 0 0626697 39997 516522 at R 0 974 can be seen as the green line between them 2 Signal Chain Characterization Each unit also had a file generated with its Signal to Noise Ratio SNR frequency response dynamic response noise floor and DC offset recorded These files may be used to inverse filter the data taken from each unit back to the initial input signal and also can give an idea of what the subsequent ambient and signal sound levels are Samples of these characterizations are seen in Fig 6 32 Unit 4129 03 04 2014 0 01732 stdev 0 003015 0 05 0 04 0 03 0 02 0 01 150 100 Probability density 50 Sensitivity V Pa 2 20 200 2000 f Hz 0 010 0 015 0 020 0 025 Sensitivity V Pa Fig 6 Sample frequency left and dynamic right response plots from a unit The frequency response shows the transfer function at given amplitudes while the dynamic response shows a histogram of device sensitivity at a 250 2 Hz with probability in ppt XVI CONCLUSION For the first time true NIST traceable characterizations of passive acoustic monitoring devices have been created These profil
21. ate events were defined as being 72100 LSBs 20 8 V above the background noise and less than 0 001 sec in duration A sample candidate event as seen in Raven is seen in Fig 2 The same candidate event as a voltage time series 1s seen Fig 3 The synchrony error of about 68 usec 15 commensurate with the error expected from clock drift early on in a deployment as seen in Fig 4 Two sets of points from the last two weeks of the deployment are seen in Fig 5 where the clock drift of the two units has brought them to within 41 minutes of each other 15 QO 0 800 0 600 0 400 0 200 0 000 aah aAma aA aA AAA AA AAA aa A A AA UA AUN edite As AA 0 200 0 400 0 600 0 800 ku 53 32 225 3 32 226 3 32 227 3 32 228 3 32 229 0 800 0 600 0 400 0 200 0 000 T 0 200 0 400 0 600 0 800 ku m 548 07 58 48 07 581 48 07 581 48 07 582 48 07 582 48 07 583 B Sound 1 CRSync Aurora 2 3 32 23 3 32 231 3 32 232 3 32 233 3 32 234 48 07 583 48 07 584 48 07 584 48 07 585 Fie 2 Temporal offset of two candidate events Detector Potential mV die OO m m an m m mu m EL WE LA de iilii RAAM NINNA 2 www mamie HE PET TE LIE Clock Time msec 48 07 585 3 32 235 48 07 586 3 32 236 D 48 07 586
22. bt CX 8 2 aa N Where kg 1s Boltzmann s constant T is the temperature in Kelvins e is the electric permittivity V t 15 the voltage at time t r t 1s the distance between the two tines of the tuning fork at time 1 Y 1s the Young s modulus of the material 1s the thickness of the tines the flexural plane 1 e f b is the thickness of the tines through the flexural plane i e L and Zis the length of the tines i e L and 1 b 2 However while the temperature dependence in this expression may seem small in magnitude it is key to remember that r Y a and are all temperature dependent and thus should be stated as Y T a T b T 1 and r t T respectively Thus these could be stated by the following Yo YpT Te T ao aAT 1 bo aAT 1 o aAT 1 ro aAT 1 2 Where is the Young s modulus at absolute zero Yr 15 the Young s modulus at high temperatures i e room temperature of 20 2 C e 7 is the Boltzmann factor o is the linear thermal expansion coefficient a0 b0 and o are the theoretical lengths in the three dimensions KT of the object at absolute zero 1 is the reduced frequency such that e 27 ata given temperature and 1s the theoretical separation between the tines at absolute zero 2 3 Furthermore the permittivity e 1s affected by temperature and 1s anisotropic within the oscillator Given the complexity of this fa
23. cross an array of recorders This induces a TDoA error on the order of km The second order error is that of acoustic multipath error This is usually on the order of a few wavelengths and occurs when reflection or refraction occurs in the signal path and can lead to several tens or hundreds of meters Finally at the third order mechanical drifting during sinking from the deployment point of a recorder due to an in water current 15 responsible for errors on the order of several tens or hundreds of meters but in a specific direction xiii In the papers that follow all of the aforementioned ideas are addressed Under the overarching theme stated in the title of engineering challenges in this area these five papers begin to give some idea of how to address some of the largest problems in this field today In the first paper Temperature compensation of a quartz tuning fork clock crystal via post processing a compensation algorithm addresses temperature induced frequency drift in a sample clock of a recording device This algorithm takes as its input the fitted curve of the frequency temperature dependence function This function is defined by cooling a clock while recording the frequency that it s operating at then warming it back up and finally fitting a quadratic curve to the resultant data This paper has been published in the Proceedings of the IEEE IFCS conference a peer reviewed proceedings journal The second paper Use
24. ctor it shall be expressed merely as 7 This affects the total energy thus e T V t 2 2 sin w T t lo aAT 1 ro aAT 1 Yo Yr T 1 7 12 kpT r amp aAT 1 v12 sn w T o aAT 1 ro AT 1 Rs SCHAT 1 P 3 All three terms have a temperature dependence At 70 the three terms are balanced such that the mechanical term dominates At lower temperatures where there 1s less energy overall the electrical term dominates and at higher temperatures where Y T 1s minimized and the material becomes more mechanically compliant the thermal and electrical terms dominate Since the mechanical resonance of the tuning fork is most efficient at 70 the surprising result from this theoretical analysis 1s that indeed the system will follow a parabolic temperature response METHODS Implementing temperature correction of frequency based on this model takes place 1n four phases The first phase 15 the characterization the function for the crystal oscillator The second is datalogging the temperature of the oscillator during the period requiring correction The third 1s the combination of the temperature log and the function into a frequency drift profile for that period The final stage 1s the application of the frequency drift profile to the data recorded in that period The first phase requires measurement of the crystal s oscillation rate over a range of temperatures This r
25. d Pete Marchetto my in laws John Caroline and Matt Myers and Dan and April Bupp for understanding when deadlines loomed and for always being interested in my progress Most VI importantly though I d like to thank my wife Katie without whom I never would ve moved to Ithaca started the PhD program or undertaken many of the other prerequisites to this document s existence truly she 15 the unseen hand holding mine throughout all of my work the best biologist ever collaborated with and the love of my life Thank you all so much vli TABLE OF CONTENTS Biographical Sketch Acknowledgements List of Figures List of Tables List of Abbreviations List of Symbols Preface Temperature Compensation of a Quartz Tuning Fork Clock Crystal via Post Processing Use of Cosmic Ray Air Shower Products for Synchronization of Underwater Recording Units Motion in the Ocean The dynamics of sinking and rising objects through current levels Characterization of marine autonomous recording units The Sound Pressure Level Observing Transponder SPLOT a satellite enabled sensor package for near real time monitoring References viii 1 vi IX xii xii 12 19 26 34 43 LIST OF FIGURES 1 1 Localization Comparison 1 2 Fitted Curve 2 Detector Circuit 2 2 Candidate Events 2 3 Fine Resolution Candidate Events 2 4 Uncorrected Clock Drift 2 5 Detections and Offsets 3 Drift P
26. e each of these three constants has some amount of uncertainty induced by hysteresis This means that the sampling frequency of a given recorder varies as a function of temperature and therefore 1f temperature varies with time then the sampling frequency does too as f f T t where Tt is the record of temperature at a given time For further information previous work can be found in 24 and 25 2 Amplitude Uncertainty The sources of amplitude uncertainty occur at multiple points in the signal chain of MARU 21 The analog signal chain 1s outlined in fig 1 I I Hydrophone 1 E a a PP Ilse Fig 1 MARU analog signal chain including 2 4 pole adjustable band pass filter two separate reference voltage sources The crystal on the TTS is the aforementioned source of temporal uncertainty Beginning from the input the hydrophones all of model 94 55 HTI Long Beach MS USA have individual sensitivities M on the order of 160 dB re 1 V uPa These sensitivities are themselves based on input values of supply voltage to the preamplifier frequency and temperature giving a multivariate function M V f T where Vin is the input voltage to the preamplifier f 1s the frequency of the incoming signal giving the frequency response and 7 as above 1s the operating temperatu
27. e delay between individual detections 1s shown and the independent time scales for each unit are shown on the x and y axes respectively By showing the delay which will vary from clock to clock and unit to unit along with the time pairings of candidate events with similar nearest neighbor times can be found In this particular case the candidate events on the first unit are coincident with the first and last candidate events from the second unit This means that the probability of those being the same event 15 quite high and one or both of these pairings could be used for temporal alignment of the recordings V CONCLUSION The method described above has applications in any scenario where the addition of a clock signal post processing 1s useful Furthermore the synchronization signals 1f used with a sufficiently precise clock can be used to trace back the angle of incidence of cosmic rays in this energy band Finally the correlation between multiple units across an array and across multiple candidate events is more than sufficient to support synchronization for localization using multilateration on sound data ACKNOWLEDGEMENTS The authors would like to thank Robert Koch for his help in deployment Christopher Tessaglia Hymes and Raymond Mack for their help constructing the test units and Rebecca Ruggles for the use of her dock This work was internally supported by the Bioacoustics Research Program of the Cornell Lab of Ornit
28. e more apparent that the MARU s weak point is in the uncertainty of the measurements made from its data This paper is intended to address those uncertainties and to put forth how each unit is now characterized and its data calibrated 26 XIII TYPES OF UNCERTAINTY There are two major types of uncertainty to be addressed by this paper temporal and amplitude The temporal uncertainty evolves from the method of creating the recording namely the sample clock The amplitude uncertainty 15 generated by several different sources at different points along the signal chain 1 Temporal Uncertainty Temporal uncertainty is generated by the behavior of the crystal oscillator that drives the sample clock In any digital recording system a sample clock 1s needed to read out the level received by the Analog to Digital Converter ADC This clock needs to run at twice the peak frequency that is to be recorded known as the Nyquist frequency or 2 In most cases this clock is a quartz crystal that resonates at the frequency desired However quartz clock crystals have a parabolic relationship between their frequency and temperature They are stable at their quoted frequency fo only when at the temperature at which they were cut To At any other temperature they diverge as in Eq 1 fT k 2 T k 1 Where kj kz and k are constants T 15 the crystal s temperature and k3 fo for the particular crystal such that f T5 fo Furthermor
29. elow and k ko and k3 frequency drift model in Hz C Hz C and Hz 1000 2000 3000 4000 5000 6000 7000 8000 respectively In general N was not an integer value so to 10 000 correct for this N was rounded to the nearest integer and running tally of the error that this introduced was kept Whenever the absolute value of the tally exceeded one that sample was removed from the tally and added to or Fig 1 Example sound localization by compensated and uncompensated data Symbol key X sound source O acoustic receivers triangles location estimates from uncompensated data dots location estimates from temperature compensated data Note that the uncompensated data have a higher maximum error and variance than the temperature compensated data subtracted from the current interval s sample count This correction factor kept the sub sample error over the entire temperature profile less than one Both the rounding and the sample accumulation were accounted for in the C term of Eq 4 The fourth and final phase corrects the recorded data using the error information involving increasing or decreasing the amount of data in each temperature log interval proportionally to its frequency error There are several methods for changing the number of samples in each interval from simple duplication or deletion of select samples to filter based operations that minimally affect frequency content The fourth phase wa
30. ere the focus of this project The sensor platform described herein could be expanded to sense any number of different environmental variables and could easily be used almost anywhere on the Earth s landmass The instrument s novelty mainly lies in its inexpensive satellite communications and the addition of sound level sensing to a terrestrial animal tag both of which would allow for quick and easy portability to an agricultural animal monitoring system Keywords acoustic animal welfare environmental impact noise noise abatement noise pollution remote sensing sensors 34 INTRODUCTION Until recently in situ remote sensing has mostly been characterized by a need to retrieve data manually from data loggers or by in person checking of environmental variables at given times 29 32 Some instruments have been developed to report from their remote locations using wired or wireless telemetry but these have been limited by wire length non omnipresent telephone service and wireless link constraints such as FCC regulations on transmission energy or line of sight constraints for FM signals not to mention power budgets of portable electronics So far satellite data transfer has remained slightly over the horizon from most in the environmental sciences just due to cost size of equipment and complexity of satellite communications This 1s no longer the case with the advent of the SPOT Connect Personal Locator Beacon SPOT LLC Coving
31. es can now be used to confirm and adjust output data and other measurements and for uncertainty calculations for both localization and sound levels XVII ACKNOWLEDGEMENTS The author would like to thank his wife and R guru Katherine Marchetto for her help in creating figures for this manuscript The author also wishes to thank Chris Tessaglia Hymes Ray Mack and Harold Cheyne for their help in ironing out the kinks in the calibration systems 33 The Sound Pressure Level Observing Transponder SPLOT a satellite enabled sensor package for near real time monitoring P M Marchetto H A Cheyne C W Clark and D J Aneshansley The authors are Peter M Marchetto ASABE Member Graduate Student Harold A Cheyne Research Associate Christopher W Clark Senior Scientist of the Cornell Lab of Ornithology Bioacoustics Research Program Ithaca NY USA and Daniel J Aneshansley professor Department of Biological and Environmental Engineering Cornell University Ithaca NY USA where Mr Marchetto is also a student Corresponding author Peter Marchetto B62 Riley Robb Hall 111 Wing Dr Ithaca NY 14853 phone 201 403 5470 e mail pmm223 cornell edu Abstract Many in situ sensing applications for the environmental sciences have suffered from a lack of means to communicate information in near real time The monitoring of incidental noise exposure on an individual animal and feasibility of use of a new instrument for such w
32. g would never deviate by more than one interval Because the thermal mass of the MARU is high even 20 C changes in air to water temperature at deployment took over 8 hours to propagate throughout the MARU This ensured that temperature values taken every 15 minutes adequately described the temperature changes experienced by the crystal At deployment the MARU itself was activated and configured to record audio and placed in the water At recovery the MARU was pulled from the water and deactivated The temperature datalogger continued to sample the temperature in the MARU throughout the entire recording period After the MARU was returned to the lab it was unsealed and the audio recording and temperature datalogger were removed for processing The third phase generates a frequency drift profile over the correction period Giving the temperature time series as input to the function gives the modeled frequencies vs time These be compared to f to calculate the error in each interval and over the whole correction period Depending on the nature of the data recorded during this period an alternate method of expressing the effect of drift may be appropriate The third phase was implemented for our case as follows In the lab the temperature record was extracted from the datalogger and saved in a generic Comma Separated Value format CSV The temperature data were then copied into and processed using a Microsoft Excel 2007 macro enabled s
33. haracterized well leaving the post processing synchronization using the pinger signals error prone Out of system synchronization methods have been tried with the MARUS including the end sync method and in water tone playback both of which rely upon an in situ injected stimulus and post processing temperature compensation 9 10 All of these methods still leave something to be desired terms of uncertainty as the best of them can still only achieve a clock linked uncertainty of 3 ppm Most of the electromagnetic spectrum cannot penetrate the ocean RF above several kHz has a skin depth measured in single meters and anything above that up to gamma radiation has a relatively short propagation length Gamma rays can travel some distance in water but they re easily confused with natural background radiation owing to the presence of naturally radioactive salts seawater This leaves the upper end of the energy spectrum leaving anthropogenic photon sources behind and delving instead into the realm of cosmic rays II THEORETICAL BACKGROUND Cosmic rays are actually comprised of a bevy of different elementary particles that are incident on the atmosphere Most of these are protons have been accelerated at up to about 0 99 c by movement through galactic and stellar magnetic fields Their interaction with the atmosphere creates a cascade of less massive particles that radiates outward from the initial interaction site while propagating do
34. hase Diagram 3 2 Displacement and Search Area 4 Analog Signal Chain 4 2 Rail Voltage 4 3 Test System Diagram 4 4 Tests System State Diagram 4 5 Sample f T Curve 4 6 Sample Response Curves 5 SPLOT Block Diagram 5 2 SPLOT State Diagram 5 3 GPS Uncertainty 5 4 Sheep Track 5 5 Comparison of Trends IX 15 16 16 17 17 21 25 28 29 31 31 32 33 37 38 40 40 4l LIST OF TABLES 1 1 Average Absolute Sync 1 2 Average Relative Sync 3 1 Input Parameters 3 2 Output Variables 5 Measurement Uncertainties 5 2 Measurement Units and Precisions 20 20 39 39 MARU SPLOT SPL CLO BRP TDoA LIST OF ABBREVIATIONS Marine Autonomous Recording Unit Sound Pressure Level Observing Transponder Sound Pressure Level dB re 20 uPa in air or 1 Pa in water Relative Humidity Cornell Lab of Ornithology Bioacoustics Research Program Time Delay of Arrival Xl M r t a b LIST OF SYMBOLS Frequency Hz Temperature C Coefficient units variable Energy J Radius or linear distance m Electrical potential V or Volume L Young s modulus Pa Angular velocity sec or rad sec Permittivity F m Population integer number Carryover constant real number Pressure Pa Sensitivity V Pa Drag coefficient unitless Density kg L Time sec Depth m z axis only Velocity m sec Area m Time constant sec Gravity 9 81 m sec Angle or rad xli PREFACE There
35. have been observed to have slightly different gain and filter characteristics despite them supposedly being discretized into certain groups of presets The gain and filter characteristics too are determined by the rail voltage available and the temperature of operation The offset voltage a virtual ground of sorts to reference the signal to is also generated on the SCCB and added to the AC component of the output from the filters Since the rail voltage on the SCCB side of the signal chain is 5 VDC then to preserve the dynamic range the offset voltage is half that at 2 5 VDC This 15 created by a single regulator whose output has also been found to be temperature dependent such that it is Vopersccr Vin Further downstream this signal 1s piped into the analog input of the TattleTale Model 8 datalogger Onset Pocasset MA USA This datalogger 1s designed to have a 12 bit ADC and a sensitivity of 1 mV LSB This means that its reference voltage of 4 096 VDC is generated by an on board voltage regulator which is also temperature dependent giving V rrs Vin T 20 In all of the above cases the unregulated voltage rails will trend towards zero over time given that the entire system 1s powered by alkaline batteries on two rails 7 5 VDC for the SCCB and 13 5 VDC for the TT8 Tmperature is not moderated for the MARU anything except for a very large heat sink called the ocean XIV CHARACTERIZATION METHODS For each of
36. heating it to 75 known as furnacing the crystal b cutting the crystal such that its 7 1s at the target temperature for operation or c cutting the crystal to create a flatter function that 15 minimizing k and kp Disadvantages of these methods are that furnacing 15 power intensive while custom crystal cutting for flat f T or target To 1s cost prohibitive for most applications and does not compensate for additional temperature variations A data acquisition platform used by our group the Marine Autonomous Recording Unit or MARU 1 typically operates near 0 C for months at a time resulting in the accumulation of several minutes of sampling period drift Additionally its power and cost budgets are limited making furnaced or custom cut crystals infeasible Using characteristic f T curves for the MARU S crystal in conjunction with data on its temperature over time we developed a method to minimize the frequency error and were able to demonstrate reduction sampling period drift as it pertains to acoustically derived location estimates IL THEORETICAL BACKGROUND The mode of the oscillator question 1s flexural the xy plane The energy of the system 1s equivalent to the sum of the thermal energy the electrical energy Ez and the mechanical energy of the oscillator The last term is also temperature dependent such that V t 2 Y 2 Ep kBT 2 50 a
37. hology 18 Motion in the Ocean The dynamics of sinking and rising objects through current levels Peter Marchetto Bioacoustics Research Program Cornell Lab of Ornithology and Department of Biological and Environmental Engineering Cornell University Ithaca NY USA ABSTRACT This paper proposes a predictive algorithm for determining landing and surfacing radii of dropped and floated instruments in water Particular attention is paid to examples of traversal of multiple current layers I INTRODUCTION Many methods exist for finding where a drifting object may be on the surface of the ocean or in 1t 17 23 and many methods seem to also exist to explain how the relative forces of water and wind may affect that object s motion while floating However when deploying objects that sink to the ocean floor or rise to its surface 1t would be helpful to have a means of predicting the landing and surfacing radii The algorithm described in this paper aims to fix the current dearth of such methods 19 VI DEFINITION OF VARIABLES Inputs Parameter Symbol Float Drag Coefficient Float Density kg L PsF Anchor Drag Coefficient CdA Anchor Density kg L PsA Delay Time sec Loo Deployment Depth m h Water Current Velocity m sec Dow Wind Speed m sec n Medium Density kg L p xy Plane Cross Sectional Area of Float AxyF m xz yz Plane Cross Sectional Area of Float Ar m xy Plane Cross Sectional Area of Axy
38. kes suggestions on how to use the characterization data to calibrate the acoustic data returned from the recorder This paper has been submitted to the Journal of the Acoustical Society of America Express Letters The final paper of this group Sound Pressure Level Observing Transponder A Satellite Enabled Sensor Package For Near Real Time Monitoring is a description of a proof of concept experiment in getting ambient noise levels and all of the parameters needed to calculate acoustic impedance back from the field via satellite modem The experiment was a success as the paper shows and the design itself is viable and useful This paper has been submitted to the Transactions of the American Society of Agricultural and Biological Engineers By looking deeply into the engineering challenges facing bioacousticians and anyone else who would record data in these harsh environments solutions can be found to improve recorders and to create smaller uncertainties in the resultant data Furthermore careful design based on the results of these studies will allow for more devices to come back unscathed functioning as planned with more data that can be of use to those doing the recording XV Temperature Compensation of a Quartz Tuning Fork Clock Crystal via Post Processing Peter Marchetto Adam Strickhart Raymond Mack and Harold Cheyne Bioacoustics Research Program Cornell Lab of Ornithology Cornell University Ithaca
39. n the MARUS for temperature is 0 2 C 7 However these were not all tested together at one point in time and so may be offset from one another Further characterization against known physical standards is called for in this case The correction factor makes the time scale slide back and forth by 0 5 samples and thus contributes 1 uncertainty of 2 5 10 VI CONCLUSION Using this method the timing uncertainty resulting from thermally induced drift in quartz crystals can be compensated This compensation method is robust enough to be used for low f applications such as for acoustic recordings and localization and is an attractive alternative when power or cost constraints preclude the use of furnaced or cut crystals Finally the method described gives similar uncertainties to a TCXO or OCXO at lower cost ACKNOWLEDGEMENTS The authors would like to thank Chris Clark and the administrative staff of the Bioacoustics Research Program at the Cornell Lab of Ornithology for supporting this physical research in furtherance of their stated biological goals 11 Use of Cosmic Ray Air Shower Products for Synchronization of Underwater Recording Units Peter Marchetto and Harold Cheyne Bioacoustics Research Program Cornell Lab of Ornithology Cornell University Ithaca NY USA Department of Biological and Environmental Engineering Cornell University Ithaca NY USA ABSTRACT This work explores the potential of usi
40. ng cosmic ray air shower products to synchronize an array of autonomous underwater recorders in post processing Because the electromagnetic spectrum below 1 GeV is blocked by the Faraday cage behavior of bodies of water and ELF length antennae are not feasible in this application air shower products offer an attractive alternative for an in deployment synchronization signal A test was performed using low cost ionization detectors in two sealed Marine Autonomous Recording Units MARUS The data collected were then analyzed and candidate events were matched to demonstrate the feasibility of using these signals to time synchronize multiple independent devices I INTRODUCTION Most underwater recording and sensing platforms including the Marine Autonomous Recording Unit MARU 8 lack the capability of synchronizing themselves with other units across an array Most current methods of underwater sensor array synchronization rely either on GPS clocks through a surface expression with a dedicated antenna or on carefully orchestrated acoustic pings which are extremely power intensive Power limitations and or deployment 12 logistics e g extreme depth usually preclude having a surface expression for the recorders for GPS Using an acoustic pinger to generate known acoustic events for synchronization has the disadvantage that neither the underwater recorder s precise position nor the speed of sound between the pinger and the recorder can be c
41. of Cosmic Ray Air Shower Products for Synchronization of Underwater Recording Units makes use of the ground penetrating nature of cosmic rays to propose a means of synchronizing an array of recorders sitting in an RF inaccessible location like the bottom of the ocean These cosmic rays along with the fast subatomic particles that they create during interactions in the atmosphere ground and water create a signal that can be seen easily by two recorders in an array and which propagates at the same rate as the sample rate of the recorders Thus a signal picked up by two recorders can be used to anchor them in time to one another and signals picked up across an array can be used through cross correlation to align the timing of recordings from the entire array This paper is being submitted to the journal Review of Scientific Instruments The third paper Motion in the Ocean The dynamics of sinking and rising objects through current levels addresses another kind of drift In this case this 1s the mechanical drift caused by oceanic sub surface currents which induce a significant XIV uncertainty into the latitude and longitude of the location of a recorder on the seafloor This paper has been submitted to the Ocean Modeling Fourth in the list of papers 1s Characterization of Marine Autonomous Recording Units MARUS a characterization protocol paper This paper describes exactly how to characterize an underwater recording unit and ma
42. orth TX USA and into TEBM36S12 8 A balanced mode radiator Tectonic Elements St Neots Cambridgeshire UK mounted in a case The intensity of sound from the radiator is measured by Type 8103 reference hydrophone Bruel amp Kj r N rum DK collocated with the DUT s which is interfaced to the first input channel of the audio analyzer through a Nexus series preamplifier Br el amp Kjaer N rum DK Fig 4 Test system state diagram The recording is usually done at the default settings for MARU f 2 kHz gain 23 5 dB 10 Hz and LPF 800 Hz A suite of tests is done as seen in the state diagram fig 4 The recording is extracted from the on board memory and compared to the expected template The extracted recording and all relevant test data are saved on the test system computer for later analysis This method was developed from parts of 26 27 and 28 31 XV RESULTS 1 Crystal f T Characterization For each unit a temperature compensation curve consisting of a quadratic fit curve from the frequency and temperature data was composed The three characteristic coefficients ko and k were exported from this as per the method used in 9 These were found to have a deviation on the order of 2 Hz from their center frequency of 39999 5 kHz A sample is shown in Fig 5 Unit 9766422 01 27 2011 3999872 39998 39997 8 39997 f Measured Falling Hz f Hz f Me
43. preadsheet to generate a temperature correction profile over time for the recording The spreadsheet tool used information from the deployment configuration to calculate the correction factors primarily the temperature model function f T for the ADC board used in the MARU The tool used the sampling rate of the audio recording to properly scale the frequency drift of the crystal It used the time zone information for the temperature datalogger and audio recording start times to correct for deployments that occur around the world or that span daylight savings changes The output of the tool was a CSV file that contains the start time of the temperature time series in the time zone of the audio recording and a listing of time intervals and number of missing or extra audio samples in that interval The corrective sample counts were calculated as follows x cau BEES LC fs fs kiT ka fe 4 1 Where N is the sample count t was the interval duration in seconds f was the desired sampling rate of the 9000 audio recording in Hz was the measured temperature for Receivers 7000T the interval in degrees Celsius fe was the theoretical id Uncompensated R 5000 Sound Source were the quadratic coefficients of the a m Temperature Compensated ___ 909 ay 2604 m frequency of the crystal at 7 in Hz C was an accumulated ji correction factor in samples which is discussed b
44. re The hydrophone 15 comprised of at least two different types of plastic surrounding a piezoceramic transducer Navy type IV lead zirconium titanate Each of these has its own thermal expansion function and so all three translate to the actual bias pressure upon the transducer Ppc T which controls its sensitivity Another non temperature related effect 1s that of having a nearly constant 1 mA current draw from a bank of alkaline batteries with no voltage regulator in line This means that the potential across the rails of the preamplifier will drop with respect to time and that the hydrophone will slowly grow deaf over the course of a recording as M Vin trends towards zero Given that there are four 20 Ah battery packs in parallel feeding this circuit this will give a very slow decline but it s still fast enough to be worth mentioning as seen in fig 2 28 Rail Voltage for SCCB Vrail VDC A 0 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 t hours Fig 2 Rail voltage over time for a standard deployment The MARU s Signal Conditioning and Control Board SCCB also has a voltage follower that acts to buffer the incoming signal to an acoustic modem while preserving the signal on the other side of the fork in the signal path and a large band pass filter array This array 15 comprised of a 4 pole high pass followed by a 6 pole low pass with a preamplifier on either end of the units in the current fleet
45. s are inputs into four of the eight analog inputs on an Arduino microcontroller Sparkfun Electronics Boulder CO USA The data from these are collated into a string of less than 41 characters and are then sent out every ten minutes after a GPS lock 15 achieved The GPS interaction and data transmission are moderated by a SatUplink Shield board Sparkfun Electronics Boulder CO USA 35 and both GPS and satellite uplink capabilities are ultimately provided by the GPS SATCOM modem board from a SPOT Connect PLB GlobalStar SPOT LLC Covington LA USA Figure 1 shows a block diagram of the system while Figure 2 shows its state diagram 36 Phidgets Sound Sensor 1133 2 Ah LiPo Battery Phidgets Pressure Phidgets T amp RH Sensor 1115 Sensor 1125 Analog Inputs Sparkfun LiPower Boost UART Converter Arduino Pro Mini Microcontroller Sparkfun SatUplink Shield GlobalStar SPOT Tracker GPS SATCOM Modem Board Fig 1 Block diagram of the SPLOT For a test of the system it was first set outside stationary at a known location This allowed for an evaluation of its location uncertainty This can be seen in Fig 3 and is about 5 m The next test of the system was to put it on an animal to be tracked In this case sheep Sr were chosen both due to their availability and the need for a test subject large enough to carry the 2 kg mass of the development board based prototype device and its protective housing Several te
46. s implemented in our case as follows The temperature correction profile s format was designed to compare the actual and desired positions of two samples calculating the difference between their actual interval and desired interval and then evenly duplicating or omitting samples throughout the interval This process was applied to the data 15 minute intervals using the calculated sample count for each temperature log interval On this scale the assumption that frequency drift 1s constant over the interval 15 reasonable given the large thermal mass of the MARU In cases where alignment of the data from multiple MARUs was necessary improved results were obtained by applying the temperature correction of frequency to an acoustic synchronization that uses a GPS timed audio To evaluate this procedure three MARUS were assembled and prepared as in the first two phases above Their recordings were started and the units were placed in a large chest freezer at 2 C 0 5 C They were left for 18 days Periodically the freezer was opened and a sound was made to provide a distinct simultaneous event on each recording After 18 days the recordings were terminated Phases three and four of the above procedure were applied to the recordings and the results analyzed IV RESULTS AND ANALYSIS Two synchronization metrics were measured First the absolute error between the recorded time of arrival and the expected time of arrival was calcula
47. sts were conducted at the Cornell Teaching and Research Farm in Harford NY with animals wearing the device for multiple days In these tests the animal was allowed to roam either in a barn or in a pasture depending on the disposition of the flock during those days of the animals used in the testing were monitored to see that they stayed with the flock and didn t stay off on their own due to the added mass of the tag The tag itself was affixed using a large dog harness in such a way that it would stay the middle of the animal s back thus not causing undue stress This observation of behavior seemed sufficient to make certain that the sheep was not overly stressed as in Vandenabeele et al 2014 Initialize SPOT Daughterboard Check for ACK Get Sensor Values Wait for delay Concatenate to Char Array Send Char Array Fig 2 State diagram of SPLOT v 1 firmware IV UNCERTAINTY The SPLOT produces a multi dimensional array of measurement outputs meaning that there are 38 several precisions and uncertainties The precisions of the measurements are shown in Table 2 The uncertainties of each measurement were determined either from manufacturer s specifications by calibration with a known standard or by post measurement comparison The uncertainties assessed by these methods are shown in Table 1 The SPL was done by way of a transfer calibration with a NIST traceable Type 4189 microphone Br el amp Kjaer Naer
48. t by 368 meters on average with a standard deviation of 301 meters Raw Data and Fitted Curve 39999 18 39999 00 39998 80 39998 60 39998 40 39998 20 Frequency Hz 39998 00 39997 89 10 0 5 0 0 0 5 0 10 0 15 0 20 0 22 5 Degrees Ee LA RKR Degrees C rar Raw Data NAM TEN Frequency Hz 1 Fitted Data Fig 2 Example of the recorded system hysteresis of a crystal oscillator for the given temperature range and the fitted curve used in the creation of the frequency drift profile V UNCERTAINTY The uncertainty in this experiment can be expressed in three parts the uncertainty of the characterization system the uncertainty of the temperature recording during deployment and the residual correction factor for the interpolation phase the characterization system the thermocouple thermometers are the largest source of uncertainty Second to that is the DMM uncertainty in frequency mode A minimal amount of uncertainty 1s introduced by the relay box and the system noise is relatively low as the 1 f flicker noise is lower at its maximum amplitude than the signal at 40 kHz Also as the temperature decreases so does the Johnson Nyquist noise so the thermal noise in the system for the region of interest will be low The uncertainty of the thermocouple thermometers is 0 5 C 5 and that of the DMM is 600 uHz 6 The rated uncertainty of the dataloggers in use i
49. tails of animal lives use of a daily diary on animals 2008 Sparkfun Electronics Inc SatUplink Shield sparkfun com Online Available https www sparkfun com products retired 1 1088 Accessed 28 Nov 2013 46 36 37 38 Phidgets Inc 1133 User Guide phidgets com Online Available http www phidgets com docs 1133 User Guide Accessed 28 Nov 2013 Phidgets Inc 1125 User Guide phidgets com Online Available http www phidgets com docs 1125 User Guide Accessed 28 Nov 2013 Phidgets Inc 1115 User Guide phidgets com Online Available http www phidgets com products php category 3 amp product 1d 1115 0 Accessed 28 Nov 2013 47
50. tal s temperature over the course of the correction period The measurement interval should be based on the expected rate of temperature change of the crystal and on any implementation specifics of the temperature datalogger including the temperature datalogger s own temperature and timing uncertainties The second phase was executed as follows in our implementation The characterized ADC board was installed in a MARU A digital temperature logger Hobo U23 Pro V2 external Temperature Data Logger Onset Computer Corp Pocasset MA was installed next to 1t Directly attaching the temperature probe to the crystal was not feasible given the electronics layout but the thermal mass of the system was great enough to eliminate the risk of persistent or transient temperature gradients within the MARU Additionally the electronics surrounding the crystal are low enough power that self heating was not significant The temperature datalogger was set to a sampling interval of 15 minutes and activated just prior to the sealing of the MARU This interval took into account the datalogger s battery capacity and 5 storage space granting three years of battery life and fourteen months of storage life This covered the majority of use cases of the MARU The 15 minute interval was at least twice as long as the expected accumulated frequency drift of the clock crystal over a standard deployment 90 days ensuring that the correction factor applied to the recordin
51. ted to evaluate the change accuracy 8 of the synchronization Second the relative error in time of arrival between units was calculated to evaluate the change in precision of the synchronization In Table 1 the average absolute error of each unit is given In Table 2 the average relative error between units is given This improvement is equivalent to using a TCXO on the ADC board as the uncertainties are about equivalent 4 Average Absolute Synchronization Uncompensted Compensated Std Dev s Std Dev s 0 615 0 453 0 426 0 262 0 619 0 429 0 508 0 226 3 0 688 0 447 0 396 0 223 Table 1 Average absolute error per channel for all test sounds Average Relative Synchronization Uncompensated Compensated Std Dev s Std Dev s 0 331 0 207 0 136 0 145 0 970 0 944 0 316 0 229 Table 2 Average relative error between channels for all test sounds In Fig 1 the above relative synchronization data were applied to a theoretical typical arrayed deployment in which the produced sounds would reach each unit at the same time localization algorithm was run on the data to evaluate the practical effect of the synchronization improvement As can be seen the temperature compensation algorithm improves the accuracy of the localization by a factor greater than two The uncompensated locations are incorrect by 981 meters on average with a standard deviation of 931 meters The compensated locations incorrec
52. ton LA USA which is capable of sending up to 41 characters of text back through GlobalStar s satellite network once every ten minutes 33 Other systems such as the Icarus Initiative have yet to come fully online and logging tags such as the Daily Diary 34 must be retrieved to have data downloaded Thus a commercially available system that can be used today 1s to be sought II OBJECTIVES The objective of this study 1s to test the capabilities of an animal tag to measure at least 4 environmental variable together with location via GPS and determine the methodology for when and how often to transmit this data to a satellite Sheep will be used as a test animal and sound pressure level temperature pressure and relative humidity will be the environmental variable collected Temperature relative humidity and barometric pressure are all used to calculate the acoustic impedance of the air around the instrument and the animal under study 35 Ill MATERIALS AND METHODS The SPLOT is designed to send back four different parameters Sound Pressure Level SPL temperature relative humidity and atmospheric pressure The SPL 1s measured using a model 1133 sound sensor Phidgets Calgary AB CA the temperature and relative humidity are measured using a model 1125 T RH sensor Phidgets Calgary AB CA and atmospheric pressure 15 measured using a model 1115 pressure sensor Phidgets Calgary CA These four analog measurement
53. ty of America vol 108 no 5 pp 2582 2582 J Friedt and Carry Introduction to the quartz tuning fork American Journal of Physics 2007 J Vig Quartz Crystal Resonators and Oscillators US Army Electronics Technology and Devices Dallas Semiconductor Inc Maxim Semiconductor Inc DS32kHz 32 768kHz Temperature Compensated Crystal Oscillator datasheets maxim ic com Online Available http datasheets maxim ic com en ds DS32kHz DS32KHZS pdf Accessed 11 Jun 2012 Phidgets Inc 1048 PhidgetTemperatureSensor 4 Input Product Manual phidgets com Online Available http www phidgets com documentation Phidgets 1048 0 Product Manual pdf Accessed 11 Jun 2012 Agilent Inc 34410A 11A 6 1 2 Digit Multimeter cp literature agilent com Online Available http cp literature agilent com litweb pdf 34410 90001 pdf Accessed 11 Jun 2012 Onset Computer Inc HOBO Pro v2 user s Manual onsetcomp com Online Available http www onsetcomp com files manual pdfs 10694 I MAN U23 pdf Accessed 11 Jun 2012 T A Calupca K M Fristrup and C W Clark A compact digital recording system for 43 9 10 11 12 13 14 15 16 autonomous bioacoustic monitoring The Journal of the Acoustical Society of America vol 108 no 5 pp 2582 2582 Nov 2000 P Marchetto A Strickhart R Mack and H Cheyne Temperature compensation of a qu
54. um Denmark and a loudspeaker on an APx 525 audio analyzer Audio Precision Beaverton OR USA in a hemianechoic chamber The microphone Device Under Test DUT and the speaker were placed at the corners of a 1 m equilateral triangle and a series of tones at different frequencies and intensities were played Measurement Uncertainty Method 0 00006 Post Comp SEE Ee Transfer Cal 2 Table 1 Uncertainties of measurements Measurement Precision Units 0 00001 33 SPL 0 01 1 36 SPL P 25 Table 2 Precisions and units of reported measurements V RESULTS AND ANALYSIS The initial results of this work are very encouraging as they demonstrate that the environmental parameters in question can indeed be tracked via the SPLOT Figure 3 shows the data from the stationary test which indicates that the positional uncertainty 1s 5 m in either axis and the SPL can be measured to a resolution of 0 1 dBA Tracking with the device 1s possible as shown in Figure 4 which means that latitude 39 longitude and time are all reported back in an effective manner Moreover the data trends over a given period of time as seen in Figure 5 are not very covariant with each other most correlations were below 2 0 10 thus confirming that they must be tracked independently SPL dB A 9 0 to 16 667 9 16 667 to 33 333 33 333 to 50 50 to 66 667 66 667 to 83 333 83 333 to 100 Fig
55. wnwards towards the surface of the earth known as an air shower Most cosmic rays in the 100 GeV energy range occur with a frequency of about 1 day have an air shower surface expression on earth of about 10 14 km depending on incident particle speed 13 and a shower interaction depth of up to 15 km below the planetary surface 11 13 This gives extraordinarily good odds that candidate events will happen often enough during a deployment to make them useful for synchronization In a worst case scenario consider a particle with an energy of 100 GeV traveling at 75 c incident at a very shallow angle through the atmosphere 11 Given an array nearest neighbor distance of 10 km the transit time of the array segment interacting with this air shower will be 44 usec which 1s approximately 0 067 m with a sound speed of 1500 m sec or three samples at a sampling frequency of 64 kHz This gives an average synchronization error for the event of 2 samples at 64 kHz IIl METHOD In this experiment two MARUS were deployed with an ionization detector in place of their standard hydrophones The detectors had an 81 VDC nominal power supply comprised of nine 9 volt batteries in series connected to a 1 resistor in series with a neon indicator lamp model 2ML Visual Communications Co Poway CA USA 14 A standard monaural 1 8 tip sleeve jack was added to measure the voltage across the resistor Whenever ionizing radiation struck the l
Download Pdf Manuals
Related Search