Low-cost Wireless Condition Monitoring for an Ultracold Atom Machine
MMAY 2020 1
Low-cost Wireless Condition Monitoring for anUltracold Atom Machine
Matthew Chilcott and Niels Kjærgaard
Abstract —We present a flexible wireless monitoring systemfor condition-based maintenance and diagnostics tailored fordynamic and complex experimental setups encountered in mod-ern research laboratories. Our platform leverages an Internet-of-Things approach to monitor a wide range of physical pa-rameters via customized sensor modules that broadcast to anetworked computer. We give a specific demonstration for aso-called ultracold atom machine, which is the workhorse ofmany emerging quantum technologies and marries a broadspectrum of equipment and instrumentation into its setup. Wesupply prescriptions for the implementations of a range of sensormodules and describe the services we use to store, process, andvisualize the data collected.
Index Terms —Condition monitoring, fault detection and iso-lation, process monitoring, sensor networks, ultracold atommachine, wireless instrument control, wireless sensors.
I. I
NTRODUCTION T HE advent of laser cooling opened up a way to producetrapped atomic gases at temperatures only a few mi-crokelvin above absolute zero, initiating the field of ultracoldatomic physics [1]. Augmenting laser cooling with evaporativecooling, a push was made to the nanokelvin domain whereBose-Einstein condensates and degenerate Fermi gases wereattained. These ultracold atomic systems have remained at thecutting-edge of research because they offer a pristine envi-ronment to explore fundamental quantum phenomena, and theexperimental platforms used in their production can accuratelybe referred to as ‘ultracold atom machines’ [2], [3]. Suchmachines are extremely complex, hybridising many differenttechnologies from optical, microwave, vacuum, electronic, andmechanical engineering. They typically incorporate a rangeof both commercially available equipment and custom-builthardware, patching a variety of systems together to get afunctional ultracold atom machine and an optimised processsequence. The end product of running a process cycle – asmall gas cloud a few billionths of a degree above absolutezero temperature, levitated inside a vacuum chamber by elec-tromagnetic fields – is extremely sensitive to changes in theprocess parameters and the environment. A reproducible resultis hence strongly tied to these conditions remaining stable. Ac-cordingly, variations in the end product may be associated withvariations in specific monitored process parameters throughsuitable analysis. For example, Ref. [4] employed commercial
Manuscript received May 5, 2020; revised NN. (corresponding authors:Matthew Chilcott; Niels Kjærgaard.)
The authors are with the Department of Physics, QSO—Centre forQuantum Science, and Dodd-Walls Centre for Photonic and Quan-tum Technologies, University of Otago, Dunedin, New Zealand (e-mail:[email protected]; [email protected]). data acquisition hardware to monitor 50 parameters of anultracold atom machine and tied fluctuations in the final atomnumber to the temperature of a magnetic field coil.Condition monitoring [5] is a well known paradigm inconventional industrial plants, as monitoring equipment pa-rameters allows for condition-based or predictive maintenance,leading to increased equipment up-time and smaller mainte-nance and operating costs. These benefits could also be reapedin a research laboratory setting, where however particularchallenges emerge when dealing with research equipment. Pastgenerations of, for example, microwave frequency synthesiz-ers, current supplies and laser controllers will not nativelyinterface with a condition monitoring system. Such equipmentmight however represent a considerable investment and fromperformance point of view remain completely adequate: a highend Hewlett Packard 26 GHz YIG based synthesizer fromthe 1980s remains a fine instrument today. Incorporation ofsuch instruments into a monitoring system along with rangeof sensor modules can be achieved using ‘Internet-of-Things’(IoT) technologies and ideas – an approach which is formedaround having large numbers of devices (‘things’) connectedto the Internet [6], and has grown in popularity with theavailability and declining cost of hardware with networkingcapability. IoT ideas are also finding use in home automation[7], wearable technology, healthcare [8], and are moving intowireless sensor network technologies in conventional process-ing plants [9], [10]. Wireless sensor networks themselves areused in numerous settings including sailboats [11], launchvehicles [12], heritage buildings [13], agriculture [14], andenvironmental [15] and land slide monitoring [16].In this paper, we focus on a wireless low-cost, long-termmonitoring system, with an emphasis on experimental appa-ratus, which entails interfacing with commercial laboratoryequipment (for example via IEEE-488 or RS-232) along withhomebuilt and customized sensors using IEEE 802.11b/g/nWiFi as the communication backbone. Our monitoring systemaugments a wireless sensor network with services for datacollection and analysis as illustrated in Fig. 1. We havetested our monitoring system over a period of 2 years inconnection to an ultracold atom machine at the Universityof Otago which over the past decade has developed intoan apparatus of significant complexity [17]–[22]. We havefound the operation of the monitoring system robust, reliable,and useful in detecting equipment failures in our laboratory,especially those of a chiller used for cooling water, andthe lab air conditioning – saving us from further equipmentfailure, and answering the questions “does it feel warm inhere?” and “why is the experiment performing badly” with a r X i v : . [ phy s i c s . i n s - d e t ] M a y MAY 2020
Fig. 1. The architecture and data-flow pipeline of the monitoring system. Some examples of the sensors we use are shown. logged data. More recently, we have extended our data loggingframework to contact tracing of lab user during the COVID19pandemic. In the following we will describe key features ofour monitoring system, including the construction of wirelesssensor modules, examples of their usage and their integrationinto a flexible architecture that can log data from a diverserange of sources. II. S
ENSOR M ODULE
Our sensor modules are based on either the ESP8266 orESP32 micro-controllers by Espressif Systems which are self-contained System-on-Chips with built-in WiFi. These modulesalso contain analog-to-digital converters (ADCs), digital serialcapabilities (including UART, I2C, SDIO, SPI), and a numberof general-purpose input-output (GPIO) pins. We specificallyuse the NodeMCU V1.0 [24] and the DevKit C developmentboards for the ESP8266 and ESP32 respectively, which addUSB serial interfaces for convenient programming as well asa supply voltage regulator.Each of our sensor modules is programmed with firmwareaccording to the application and containing drivers for theconnected sensor. Table I shows the cost of the developmentboards, and the extra hardware for three typical example nodes- a simple temperature sensor, an RS-232 interface and aGeneral Purpose Interface Bus (GPIB) device.Our code [23] provides a ready-to-use framework wherefirmware is uploaded to the ESP development board via itsUSB connection (we use the PlatformIO tools [25] for thispurpose). Once deployed, the device can be reprogrammed viaWiFi from a web-browser. Our code library contains detailed Reference [23] provides ready-to-use resources (sensor module firmware,service automation, and PCB designs for manufacture) and gives detaileddocumentation on how to replicate our monitoring system. information and instructions on configuring and deployingthe modules. Essentially, the steps to be followed can besummarized as: • Set up the data collection services on a networked com-puter in the laboratory (only needs to be done once). • Connect sensors to a the ESP development board. • Specify the connected sensors in the firmware. • Build firmware on computer (via PlatformIO) and uploadit to the development board via USB (the sensor modulewill then broadcast a WiFi network to configure itsconnection). • Deploy sensor module in lab. • Connect to WiFi broadcast by the sensor module toconnect to WiFi in the lab and publish measurementsto the data collection server.Below we present some of the sensors that we have de-ployed in our laboratory and for which we provide code andhardware examples.
A. Temperature Sensor
For contact temperature measurement we primarily use theDS18B20 from Maxim Integrated, which we fit into variousplaces in our machine. Each sensor has an on-board digitalinterface, and is capable of reporting the temperature to aresolution as good as . [26]. We find the TO-92package (small “transistor” shape) of the DS18B20 to be aconvenient size to locate throughout our apparatus. We alsoemploy variants enclosed in a water-proof housing.To construct a wireless sensor module from the DS18B20,one can simply connect it to an ESP8266 as shown inFig. 2(a), then configure the firmware specifying a DS18B20 isconnected, and a handle to identify the measurement. MultipleDS18B20 devices can be connected to the same GPIO pin andread out separately as they have unique IDs. HILCOTT et al. : LOW-COST WIRELESS CONDITION MONITORING FOR AN ULTRACOLD ATOM MACHINE 3 (b)(c)
GPIO2GND
ESP8266
DATAGNDVCC
DS18B20 (a)(d)
A0GND
ESP8266NodeMCUTOUTGND GPIO21GPIO22GNDSDASCLGNDA0A1A2A3GND ESP32ADS1115VCCADDR 3V3
GPIO4GND
ESP8266
SDAGNDVCC
MLX90393
ESP8266
SDAGNDVCC
BME280 % kPa GPIO21GPIO22GNDSDASCLGND ESP32MCP9600 VCCADDR 3V3(e)(f)
Fig. 2. Connection diagrams for wireless sensor modules. (a) Temperaturesensor module; the connection of VCC is optional, as the 1-Wire data interfaceis capable of powering the sensor. (b) thermocouple module (c) Environmentalsensor module (temperature, pressure, and humidity). (d) Voltage measure-ment; the onboard ADC on the ESP8266 the signal is connected to the A0pin on the NodeMCU board. The signal is referenced to the device ground.(e) Voltage measurement using the ADS1115 external ADC. The ADS1115can also make differential measurements by measuring the voltage betweentwo of the analog input pins. (f) 3-axis magnetometer. Parts c) and d) arerepresentative of I2C sensors, which require power lines and two connectionsfor the serial interface. The modules (b), (c), (e), and (f) integrates I2C sensors,which require power lines and two connections for the serial interface. Manyof the sensors we use additionally require configuration by setting the voltageof another pin on the sensor, e.g. CS, CSB, SDO and ADDR above. TABLE IT
YPICAL B ILL OF M ATERIALS FOR MONITORING NODES .Part Price (USD)Controller Development BoardsESP32 DevKit C $7.49ESP8266 (NodeMCU V2) $5.99GPIB MonitorGPIB Connector (e.g. DigiKey 1024RMA-ND) $5.53Arduino Nano $4.66
Resistor $0.01PCB Manufacture $2.00RS232 MonitorDB9 Connector $0.77MAX202 Transceiver $1.03
Resistor $0.01 ×
100 nF
Capacitors $0.01PCB Manufacture $2.00Temperature MonitorDS18B20 Thermometer $1.69
For an even smaller sensor head or to measure temperaturesexceeding ◦ C , we use K-type thermocouples, with theMCP9600 16-bit thermocouple-to-digital converter from Mi-crochip (see Fig. 2(b)). This device features all required signalconditioning, support for multiple types of thermocouples,and built-in compensation for the cold junction formed byconnecting the thermocouple to the converter. B. Atmospheric Sensors
To measure atmospheric signals of interest, such as ambienttemperature and humidity, we use the AM2302 from AosongElectronics – a small digital sensor which can measure theambient temperature and relative humidity. Another sensor weuse for this purpose is the BME280 from Bosch Sensortec,which additionally measures the atmospheric pressure (seeFig. 2(c)).From these signals one can, for example, detect changesin the lab environmental control or the function of a positivepressure environment. In particular, we have found it usefulto monitor the outside temperature with an atmospheric sensormodule mounted on the side of the building to examine theimpact of weather on our cooling water system’s performance.Monitoring the performance of the lab air conditioning hasalso proven useful.
C. Analog Voltage Measurements
Virtually any physical quantity we want to measure can beconverted to an analog voltage by some electrical transducer,for example: currents (using a sense resistor, or hall-effectsensor), resistance (using a bridge circuit), temperatures (usinga thermistor), forces/pressures (with a strain gauge), andmagnetic fields (with hall-effect sensors). Furthermore, manyindustrial sensors produce an analog signal of 4–20 mA , MAY 2020 and many research instruments provide a voltage output tomonitor signals of interest. The variety of applications formonitoring analog signals makes this an important feature ofour framework. Fig. 2 (d) and (e) shows two configurationsfor this task.
1) Internal ADC:
The ESP8266 features an onboard 10-bit ADC with a native to range, which on theNodeMCU boards has been extended to a to . rangewith a 220:100 kΩ resistive voltage divider. This range canbe adjusted by suitable replacement of the resistors of theNodeMCU board or by adding a resistor inline with the signal.The ESP32 includes two 12-bit ADCs, which can be config-ured to take maximum voltages of up to . . These ADCscan be configured to measure on a number of different GPIOpins, meaning that one can monitor multiple signals in parallel.However, the ESP32’s ADCs suffer from non-linearity in theirreading, so we recommend using the ADS1115 external ADC,which we describe below.The onboard ADC is referenced to the chip ground, andhence the ESP controller must share its ground with the signal.The ground does not have to be at the earth potential, asthe controller can be powered from an isolated/floating powersupply or battery. This allows for performing measurementsin galvanic isolation from earthed equipment.Our testing has demonstrated that the analog input pinsof the ESP controllers are not tolerant of negative voltages(less than − . ), or voltages exceeding . . It should benoted that the controller should be powered whenever it isconnected to a signal source, as the input’s behaviour andtolerance changes when not powered.The onboard ADC can be sampled at a rate of about ,but for most of our monitoring applications, we sample atmuch lower rates, e.g., .
2) External ADC:
For higher resolution measurements,we use the ADS1115 analog-digital converter from TexasInstruments, which provides higher-resolution (16 bit) conver-sion, on four input channels, and a programmable-gain input.This allows a maximum measurement range of ± .
144 V ,though the input pins can only tolerate − . to . whenrun off a supply. Negative signals are only possiblewith differential measurements between two positive voltages.Similarly to the ESP’s internal ADC, one can make floatingmeasurements with care. D. Magnetic-field Sensors
Ultracold atoms are particularly sensitive to magnetic fields.Hall-effect sensors are one way of measuring these fieldsand present an analog signal which can be monitored aspreviously described. We use a digital magnetic field sen-sor, the MLX90393 from Melexis. This is a 16-bit, 3-axismagnetometer capable of measuring fields of up to ±
500 G ,which we find suitable for monitoring our magnetic trapsand background fields. A variety of alternative digital 3-axismagnetometers exist, but many are manufactured as digitalcompasses and saturate at fields of as little as , whichrenders them useless near our electromagnets. These sensorscan be connected as depicted in Fig. 2(f).
E. Digital Signal Monitoring
Commercial instruments embedded in our ultracold atommachine often have TTL outputs, e.g. “PLL locked”, “Laserlocked” or “Error”, which we track via sensor modules. Otherinteresting digital signals include laser interlock status, com-parator outputs, or timing signals, so the monitoring system isaware of the state of a process cycle.The GPIO pins of the ESP boards can be used to monitor thestate of digital signals. The ESP controllers are . devices.The ESP32’s GPIO pins will not tolerate higher voltages, butthe ESP8266’s GPIO pins are tolerant and can acceptstandard TTL signals. For higher voltages, a resistorshould be put in series with the signal to limit the currentto less than
12 mA , preventing damage of the GPIO pins.
1) Time-sensitive Digital Signals:
Our flow meters forcooling water produce pulsed signals with the pulse ratebeing a measure of the flow rate. The GPIO pins of the ESPcontroller could be used for these tasks, but the processor inthe ESP controller is busy managing WiFi tasks, and runninga real-time operating system, which is unideal for monitoringtime-sensitive signals. The ESP32, however, has an onboardpulse counter peripheral which can be used. For the ESP8266we have used an external processor (an ATmega328P fromMicrochip on an Arduino Nano 3 board [27]) to perform time-critical functions. The ATmega328P has a conventient timer-counter peripheral for this purpose.To measure the pulse rate of our flow meters we have oneof these external boards programmed as a frequency counterand we read the frequency out with the ESP controller via adigital serial connection (I2C). The external processor can beused to make pulse width measurements (e.g., monitor a laserbeam shutter’s opening time to detect if it is stuck) or performa Fourier transform on analog signals to examine the spectrumof an incoming signal without loading ESP controller.
F. Serial Interface Readout
Commercial instruments are often outfitted with a digitalserial interface for remote control and diagnostics. Theseinterfaces, e.g. RS232, RS485 or GPIB, generally provide theability to query the status of the device (e.g. via SCPI). Despitethe obvious usefulness of wireless GPIB/serial interfaces [28],[29], curiously, commercial vendors do not provide solutionswith integrated wireless capabilities, but invariably require awireless router on top of an already costly GPIB-to-Ethernetinterface [30], [31]. The cost of the latter can range from $200to $1600 depending on supplier. As an attractive and morecompact alternative, we provide a solution which integratesGPIB-to-WiFi on a single printed circuit board (PCB), with atotal component cost that is less than $20.To connect the ESP controller to the communication port ofa commercial instrument, we need a protocol translator capableof handling and generating the voltage range of the interfacebus. GPIB or IEEE-488 is a parallel bus interface, which onlyrequires TTL signal levels. To manage this bus, we add anATmega328P on an Arduino Nano 3 [27] board, following adesign provided by Ref. [28]. This extra component acts asa level-shifter, and a serial-to-parallel converter, making the
HILCOTT et al. : LOW-COST WIRELESS CONDITION MONITORING FOR AN ULTRACOLD ATOM MACHINE 5
Fig. 3. Left: A wireless GPIB interface which, for example, we use to monitorthe power supplies of our electromagnets. The ESP32 uses an Arduino Nanoto interface with GPIB, where 3 separate power supplies are queried over thebus. Right: A wireless RS232 interface which, for example, we use to monitora commercial laser controller. Our RS232 module acts as Data TerminalEquipment (DTE), but many scientific instruments are wired as DTE, so italso features a Data Communications Equipment (DCE) wired connector toavoid the need for a null-modem cable. The MAX202 used to drive theseserial interfaces is on the reverse side of the board. A postage stamp is shownfor size reference. system GPIB compatible. We avoid using more powerful linedrivers, which would be required for driving a GPIB bus withmany instruments – our largest single bus has 3 instruments– and makes this setup GPIB compatible, rather than GPIBcompliant. The ESP controller and Arduino are assembled ona custom PCB with the appropriate connector to plug straightinto an instrument (see Table I). The board can be poweredfrom a supply via the exposed header, or the USB portson the Arduino or ESP development board.For RS232, we use the MAX202 [32], which is a TTL-to-RS232 adapter. RS232 devices must be able to withstandshort-circuit to ground, and incoming voltages of up to ±
25 V ,which are constraints the ESP controller cannot satisfy, but theMAX202 can. The MAX202 additionally contains a chargepump doubler and inverter to produce signals of ±
10 V fromthe power supplied. We also use a custom PCB for thismodule. IEEE-488 and RS-232 are industry standards thathave been used in test instrumentation since the 1960s forautomation and control. Our system therefore opens up theattractive possibility of retrofitting old high-end equipmentwith wireless monitoring capabilities.III. P
UBLISHING D ATA TO S ERVER
The sensor modules connect to a message queuing service[33], [34] to which they transmit data (see Fig. 1). This serviceis run on a networked computer in the lab, which forms adata collection server. Data messages are sent as strings fromeach sensor module to the service with the value and units ofeach measurement. The messages are ‘published’ to different‘topics’ on the server, which differentiates each signal beingmonitored. Other clients are then able to ‘subscribe’ to the‘topics’, and receive the measurements. Reference [23] includes production ready computer-aided manufacturing(CAM) files.
When first deployed, a sensor module initially starts upits own WiFi network and awaits configuration from a web-browser (a mobile phone can be used for this purpose). Thisconfiguration specifies the SSID and password for the WiFinetwork to which the data collection server is connected, aswell as the address of the server. Once configured, the device’snetwork disappears, and the module starts transmitting data.In order for the data to be useful for analysis, we must havea way of storing it. We use the “agent” program Telegraf [35]to collect the data messages and store them in a time-seriesdatabase, InfluxDB [36]. The data flow is shown in Fig. 1.
A. Other Data Sources
It is straightforward to send measurements to the messagequeuing service from sources other than our sensor modules,as the messages are transported in a simple string format,and the message queue interface is an open standard withmany implementations available. Instruments that have USB,FireWire or network interfaces do not lend themselves easilyto interfacing with the ESP controllers, but using suitabledrivers for these instruments connected to a computer, onecan upload measurements to the message queue componentof the monitoring system to take advantage of the analysistools described in the following sections.With an appropriate front-end, events can also be publishedto the server from direct user input via a web browser, e.g.from a cell phone. Examples of relevant events include peri-odic servicing maintenance tasks such as filter and lubricantreplacement. Recently we have included the logging of sharedequipment sterilization prompted by regulations put in placedue to the COVID-19 pandemic. COVID-19 also introducedan acute requirement for contact tracing of laboratory per-sonnel. We therefore currently log the user occupation of thelaboratory, with a sign-in board built using NodeRED, a toolwe also use for real-time analysis as discussed below. IV. F
RONT E ND / A CTION A NALYSIS
The setup we describe uses separate tools for real-timeanalysis, and for interacting with historical data (i.e., datastored in the InfluxDB database). Details on the services andtheir configuration can be found with our code [23].Here we explain the user-facing tools that we use.
A. Visualisation
To visualise the data stored in the InfluxDB database weuse Grafana [38] which offers “drag and drop” constructionof monitoring dashboards and database queries. Grafana isaccessed via a web-browser, allowing users to view the dataremotely and from multiple computers simultaneously. It canalso be made publicly and globally accessible via the internet,as we have done [37].Grafana provides a number of ways of visualising data as‘panels’ on a dashboard. Fig. 4 shows a simple example using The registration of users entering and leaving lab could also be conve-niently achieved with an RFID proximity reader as a node in our monitoringsystem.
MAY 2020
Fig. 4. Screenshot of a monitoring dashboard built using Grafana. The top graph shows the temperature of our cooling water in a storage tank, and comingout of a electromagnet coil. The middle plot shows the temperature of MOSFETs used for controlling current. The bottom row has several gauge panels, withthe green colour indicating normal operating points. The lower plot is the cooling water supply flow-rate and pressure. Reference [37] provides a link to alive instance of our monitoring system. graph and gauge panels. Grafana includes other panel typesincluding bar gauges, tables, and heat maps. With these tools,one can create “control panel” type displays, or even “digitalchecklists”, where the operating state is indicated by colour,quickly highlighting anything that is not operating normally.Grafana supports multiple dashboards, and one can set thedisplay to automatically change through them to get a liveoverview of the system. Grafana also offers a simple way forusers to query and export data, avoiding having to interfacewith the database directly.
B. Real-time Analysis
Our real-time analysis operates on data received from themessage queue server directly, so detects if there is a faultwith the experiment as data comes in. We may either wishto avoid wasting time collecting faulty experimental data, orwe may try to act to correct the failure or minimise damagecaused by the failure. The last point is the realm of safetydevices (e.g. thermal fuses), and while it is not recommendedto displace hardware interlocks and failsafes, real-time analysisopens up interesting ways to augment these. An advantageof using the monitoring system to perform these features isthe large amount of data available. For example, one mayuse data from multiple sensors simultaneously, e.g. differentialflow rates between source and return lines can detect a leak.For real-time analysis we use NodeRED [39] which runson the data collection server and provides a web-based graph-ical and JavaScript programming environment for reacting tomessages from the message queuing server.We use NodeRED to perform some preventative actions,such as shutting off our coolant pump if the system detects aleak or if the coolant tank is low. We can also disable thepower supplies if the electromagnets being cooled get toowarm or if the flow rate drops off. It is worth again noting that, especially the last two of these, should have dedicatedhardware interlocks to act as a fail-safe. Real-time analysisis not meant as a replacement for proper safety design, butcan take action to avoid the need to replace thermal fuses.Additionally, it can notify the user that there has been a fault,and where the fault occured.To take preventative actions we must have some devicescapable of effecting control over parts of the experiment.While we have focused on the monitoring uses of our sensormodules, they are equally capable of producing a controlsignal. Typical examples include controlling relays, as shownin Fig. 1, or triggering interlock circuits. Hence the modulesprovide a way to react to changes in the experiment, such asturn off a pump if a leak is detected.The above control examples produce digital signals usingthe GPIO pins. The ESP32 includes an onboard digital-to-analog converter to produce analog signals. The ESP8266 doesnot have an onboard converter, but one can be formed withpulse-width modulation and an output filter.V. C
OMPARISON OF D ATA S OURCES
As a demonstration of the capabilities of this system, Fig. 5shows the voltage of a commercial power supply (Keysight6690A) as monitored by 4 different data sources, all connectedto sensor modules. One channel of the ADS1115 externalADC, a Keysight 34401A 6.5-digit multi-meter, and the powersupply itself are used to monitor the supply voltage to ourquadrupole trapping coils, as shown in Fig. 6, and the tem-perature of our cooling water is monitored with a DS18B20.Both the multi-meter and the power supply are monitored via aGPIB connection from sensor modules (as pictured in Fig. 3).The external ADC is connected directly to the coil (withoutany conditioning), but is over-sampled by a factor of 10 andthen averaged. Both the external ADC and the multi-meterare sensitive enough to pick up the drift in the power supply
HILCOTT et al. : LOW-COST WIRELESS CONDITION MONITORING FOR AN ULTRACOLD ATOM MACHINE 7 S u pp l y V o l t a g e [ V ] External ADCMultimeterPower Supply0 500 1000 1500 2000 2500 3000 3500Time [s]910 W a t e r [ C ] Fig. 5. The voltage of the power supply for our quadrupole trap as measuredfrom different sources. The power supply readout is not sensitive enough todetect the drifts which the external ADC and the 6.5 digit multi-meter areable to detect. The signal from external ADC has much more variation. Thetemperature of the water in a storage tank in our cooling system is alsoshown. The power supply is operating in a constant current mode, so thevoltage fluctuates with the temperature induced resistance change of the coil.
External ADCMultimeterPower Supply
ADS1115
Fig. 6. A test configuration for our system. A power supply sources acurrent to our quadrupole electromagnets, trapping a cloud of cold atoms.The coil voltage is monitored by the power supply itself, a GPIB compatiblemulti-meter, and an ADS1115 ADC. The electromagnets are cooled bya recirculating water system (chiller and pumps not pictured), with thetemperature of the coolant tank monitored. voltage as it attempts to source a constant
35 A current. Thevoltage fluctuations correlate highly with the temperature ofthe water cooling the coils, suggesting that we are observingthe change of resistance due to temperature variation with acoefficient of . µ Ω K − .VI. C ONCLUSION
We have described a condition monitoring system targetedat the research laboratory. Our platform operates with a num-ber of wireless sensor modules distributed around the lab. Themodules are capable of monitoring with a variety of sensors,and also commercial (or home-built) instruments with analogor digital interfaces. The present work provides a vehicle fordetecting equipment failures and for providing informationthat quickly highlights the point of failure. It also providesa framework for logging process parameters and operating points of the day-to-day running of the experiments in whichdrifts can predict the failure of some components.We have considered the specific example of the ultracoldatom experiment, but this system is suitable for a range ofsimilarly complex experimental setups. Our system is easilydeployed and detailed instructions are provided to do this [23],making collecting information about an experiment’s healtha small investment. The code for this project [23] is open-source, and uses third-party open-source libraries and tools.We hope that work this will stimulate further development ofthis project. A
CKNOWLEDGMENT
The authors would like to thank Susanne Otto and RyanThomas for their testing and feedback on this system.R
EFERENCES[1] L. Fallani and A. Kastberg, “Cold atoms: A field enabled by light,”
EPL(Europhysics Letters) , vol. 110, no. 5, p. 53001, jun 2015.[2] E. W. Streed, A. P. Chikkatur, T. L. Gustavson, M. Boyd, Y. Torii,D. Schneble, G. K. Campbell, D. E. Pritchard, and W. Ketterle, “Largeatom number bose-einstein condensate machines,”
Review of ScientificInstruments , vol. 77, no. 2, p. 023106, feb 2006.[3] H. J. Lewandowski, D. M. Harber, D. L. Whitaker, and E. A. Cornell,“Simplified system for creating a boseeinstein condensate,”
Journal ofLow Temperature Physics , vol. 132, no. 5/6, pp. 309–367, 2003.[4] O. Krarup, “Imaging single atoms in an optical lattice,” Master’s thesis,Aarhaus University, 2018.[5] G. Vachtsevanos, F. Lewis, M. Roemer, A. Hess, and B. Wu,
IntelligentFault Diagnostrics and Prognosis for Engineering Systems . John Wileyand Sons, Inc, 2006.[6] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami, “Internet of things(iot): A vision, architectural elements, and future directions,”
FutureGeneration Computer Systems , vol. 29, no. 7, pp. 1645 – 1660, 2013.[7] W.-T. Sung and S.-J. Hsiao, “The application of thermal comfort controlbased on smart house system of iot,”
Measurement , vol. 149, p. 106997,2020.[8] L. Piwek, D. A. Ellis, S. Andrews, and A. Joinson, “The rise of consumerhealth wearables: Promises and barriers,”
PLOS Medicine , vol. 13, no. 2,pp. 1–9, 02 2016.[9] F. Shrouf, J. Ordieres, and G. Miragliotta, “Smart factories in industry4.0: A review of the concept and of energy management approached inproduction based on the internet of things paradigm,” in , Dec 2014, pp. 697–701.[10] L. Hou and N. W. Bergmann, “Novel industrial wireless sensor networksfor machine condition monitoring and fault diagnosis,”
IEEE Transac-tions on Instrumentation and Measurement , vol. 61, no. 10, pp. 2787–2798, 2012.[11] A. Bergeron and N. Baddour, “Design and development of a low-cost multisensor inertial data acquisition system for sailing,”
IEEETransactions on Instrumentation and Measurement , vol. 63, no. 2, pp.441–449, 2014.[12] M. Razfar, J. Castro, L. Labonte, R. Rezaei, F. Ghabrial, P. Shankar,E. Besnard, and A. Abedi, “Wireless network design and analysis forreal time control of launch vehicles,” in
IEEE International Conferenceon Wireless for Space and Extreme Environments , 2013, pp. 1–2.[13] L. Lombardo, S. Corbellini, M. Parvis, A. Elsayed, E. Angelini, andS. Grassini, “Wireless sensor network for distributed environmentalmonitoring,”
IEEE Transactions on Instrumentation and Measurement ,vol. 67, no. 5, pp. 1214–1222, 2018.[14] P. T. Lam, T. Q. Le, N. N. Le, and S. D. Nguyen, “Wireless sensingmodules for rural monitoring and precision agriculture applications,”
Journal of Information and Telecommunication , vol. 2, no. 1, pp.107–123, 2018. [Online]. Available: https://doi.org/10.1080/24751839.2017.1390653[15] N. Harris, A. Cranny, M. Rivers, K. Smettem, and E. G. Barrett-Lennard,“Application of distributed wireless chloride sensors to environmentalmonitoring: Initial results,”
IEEE Transactions on Instrumentation andMeasurement , vol. 65, no. 4, pp. 736–743, 2016.
MAY 2020 [16] P. Giri, K. Ng, and W. Phillips, “Wireless sensor network system forlandslide monitoring and warning,”
IEEE Transactions on Instrumenta-tion and Measurement , vol. 68, no. 4, pp. 1210–1220, 2019.[17] A. Rakonjac, A. B. Deb, S. Hoinka, D. Hudson, B. J. Sawyer, andN. Kjærgaard, “Laser based accelerator for ultracold atoms,”
OpticsLetters , vol. 37, no. 6, p. 1085, mar 2012.[18] A. B. Deb, B. J. Sawyer, and N. Kjærgaard, “Dispersive probing ofdriven pseudospin dynamics in a gradient field,”
Physical Review A ,vol. 88, no. 6, dec 2013.[19] A. B. Deb, T. McKellar, and N. Kjærgaard, “Optical runaway evapora-tion for the parallel production of multiple bose-einstein condensates,”
Physical Review A , vol. 90, no. 5, nov 2014.[20] C. S. Chisholm, R. Thomas, A. B. Deb, and N. Kjærgaard, “A three-dimensional steerable optical tweezer system for ultracold atoms,”
Review of Scientific Instruments , vol. 89, no. 10, p. 103105, oct 2018.[21] R. Thomas, M. Chilcott, E. Tiesinga, A. B. Deb, and N. Kjærgaard,“Observation of bound state self-interaction in a nano-eV atom collider,”
Nature Communications , vol. 9, no. 1, nov 2018.[22] B. J. Sawyer, M. Chilcott, R. Thomas, A. B. Deb, and N. Kjærgaard,“Deterministic quantum state transfer of atoms in a random magneticfield,”
The European Physical Journal D
AR488 Arduino GPIB Interface digit multimeter. [Online]. Available:https://literature.cdn.keysight.com/litweb/pdf/34450-90032.pdf[32] Texas Instruments, “MAX202 5-V dual RS-232 line driverand receiverwith ± Matthew Chilcott received the B.A. degree in math-ematics and computer science in 2016, and the B.Sc.degree in physics (with honours) and electronicsin 2017 from University of Otago, Dunedin, NewZealand.He has been a Software and Electronics Engineerfor a number of companies in New Zealand, andremains a freelance Technical Consultant. He iscurrently a doctoral researcher in the Department ofPhysics, University of Otago.Mr. Chilcott is a student member of SPIE and TheOptical Society.