A next generation upgraded observing platform for the automated Birmingham Solar Oscillations Network (BiSON)
AA next generation upgraded observing platform for theautomated Birmingham Solar Oscillations Network (BiSON)
S. J. Hale a,b , W. J. Chaplin a,b , G. R. Davies a,b , and Y. P. Elsworth a,ba
School of Physics and Astronomy, University of Birmingham, Edgbaston, Birmingham B152TT, United Kingdom b Stellar Astrophysics Centre, Department of Physics and Astronomy, Aarhus University, NyMunkegade 120, DK-8000 Aarhus C, Denmark
ABSTRACT
The Birmingham Solar Oscillations Network (BiSON) is a collection of ground-based automated telescopesobserving oscillations of the Sun. The network has been operating since the early 1990s. We present developmentwork on a prototype next generation observation platform, BiSON:NG, based almost entirely on inexpensive off-the-shelf components, and where the footprint is reduced to a size that can be inexpensively installed on theroof of an existing building. Continuous development is essential in ensuring that automated networks such asBiSON are well placed to observe the next solar cycle and beyond.
Keywords: robotic telescope, instrumentation, microelectromechanical systems (MEMS), Internet of Things(IoT), Industry 4.0 protocols, helioseismology, solar oscillations
1. INTRODUCTION
Figure 1. The six station Birmingham Solar Oscillations Network (BiSON). Image credit: Google Maps
The Birmingham Solar Oscillations Network (BiSON) is a collection of ground-based automated solar tele-scopes observing helioseismic oscillations of the Sun-as-a-star. The network as it stands today was completedin 1992, and consists of six remote sites shown in Fig. 1. A detailed history of the network is available inRef. 1 and 2. Here, we present development work on a prototype next generation small observation platform,
Further author information:Send correspondence to S. J. Hale
2. AUTOMATION OF A CONSUMER-GRADE TELESCOPE MOUNT
There are two key issues when attempting to automate a small commercial off-the-shelf (COTS) telescope mount.The first is access to a communications protocol with a published application programming interface (API) inorder to allow full computer control of the mount. Many COTS mounts do not have facility for computer controlat all, and rely only on the supplied proprietary control handset. In cases where the handset offers an ASCOMcompatible control interface this typically does not work without first calibrating the mount position within thehandset software, and they offer no facility to store a permanent calibration since they are expected to be packedaway at the end of each observing session. Typically movement is tracked by counting stepper-motor steps onceinitial alignment has been completed, and so the calibration is lost as soon as the mount is moved by hand orthe power is interrupted. Larger, more expensive, research-grade mounts employ absolute position encoders toavoid this problem. A fully automated telescope must be able to recover after a power failure without manualrecalibration.The communication and control problem can be solved by simply choosing only manufacturers that open-source and publish their control API. The issue of alignment and recovery from power failure without userintervention is rather more tricky. Potentially a mount could be retro-fitted with absolute position encoders, orlimit switches to indicate a home position, but this moves away from the intention of simple off-the-shelf use.Instead, it is possible to make use of microelectromechanical systems (MEMS) sensors to directly detect theattitude of the telescope. Such devices are most commonly used in smartphones and tablets, with accelerometersused to control screen orientation, and magnetometers used to determine heading when navigating. Theseinexpensive sensors can be easily mounted on the telescope itself and avoid costly modifications to the mount.A consumer-grade equatorial mount was trialled at the Mount Wilson (Hale) Observatory 60 foot solar tower,an existing BiSON site, in order to determine the possible accuracy of MEMS attitude sensors, and the precisionof solar autoguiding. Figure 2, left panel, shows the temporary binocular configuration of fibre collection opticsfor solar data acquisition, and a solar-filtered CCD camera for guiding. Figure 2, right panel, shows a prototypeautomated weather-proof housing with a rolling-roof suitable for a small mount and optics. The final design andconstruction of an enclosure is subject to further work.In subsection 2.1 below we demonstrate determining telescope attitude from an ADXL345 three-axis ac-celerometer, in subsection 2.2 acquiring magnetic heading from a HMC5883L three-axis magnetometer, and insubsection 2.3 guiding the mount on the centroid of the solar disc estimated using computer vision techniques.
A MEMS accelerometer measures linear acceleration, and this includes the acceleration due to the Earth’sgravitational field vector. Inside the device, a micro proof-mass is suspended by restoring springs, and deflectionof the proof-mass due to acceleration is detected by measuring changes in capacitance between the proof-massand sensing plates. An embedded micro-controller digitises the signal and allows the data to be output over adigital serial bus. A total of three measurement axes are required in order to sense acceleration in all directions.A thorough discussion of calibration of a three-axis accelerometer and extracting pitch and roll angles is given igure 2. Left: Sunrise at the Mount Wilson (Hale) Observatory. A small consumer-grade equatorial mount was config-ured with a binocular arrangement of a CCD camera for guiding, and optical fibre collection optics for data acquisitiontrial. Photo credit: S. J. Hale. Right: A small telescope enclosure and weather-proofing prototype with rolling roof. Photocredit: Ash Manufacturing Company. in Ref. 6 and 7. We will now summarise the calibration process and go on to apply the techniques to anAnalog Devices ADXL345 accelerometer fixed to the optical tube of an equatorially mounted telescope.The Analog Devices ADXL345 accelerometer is factory calibrated to output values in units of g , with aprecision of 4 milli- g per least-significant bit and a dynamic range of ± g . The accelerometer measures boththe gravitational field vector, and linear acceleration due to motion. Measurement accuracy of the gravitationalfield is reduced when the device is subject to additional external accelerations, however on a quasi-static designsuch as a telescope mount this is not a concern. The factory calibration is a 6-term model providing gain and offsetparameters for each of the three channels, and this generally provides adequate results in the typical use case oforienting a smartphone. Re-calibrating after the device is installed allows for greater precision to be achievedby correcting for thermal stresses introduced during soldering, and the convenience of allowing installation at anarbitrary angle – the device axes do not need to align with the system axes. A general 12-parameter calibratedoutput G can be defined in terms of the factory calibration G f by [7, eq. 35], G = G x G y G z = WG f + V = W xx W xy W xz W yx W yy W yz W zx W zy W zz G fx G fy G fz + V x V y V z , (1)where the gain matrix W includes any rotation of the integrated circuit package and also corrects for all possiblecross-talk interactions, and V are the channel offsets.If M measurements are used for the 12-parameter calibration, then the i -th measurement at pitch angle θ [ i ]and roll angle φ [ i ] becomes [7, eq. 36], G x [ i ] G y [ i ] G z [ i ] = W xx W xy W xz W yx W yy W yz W zx W zy W zz G fx [ i ] G fy [ i ] G fz [ i ] + V x V y V z = − sin( θ [ i ])cos( θ [ i ]) sin( φ [ i ])cos( θ [ i ]) cos( φ [ i ]) , (2)which can be decomposed into three equations, W xx G fx [ i ] + W xy G fy [ i ] + W xz G fz [ i ] + V x = − sin( θ [ i ]) W yx G fx [ i ] + W yy G fy [ i ] + W yz G fz [ i ] + V y = cos( θ [ i ]) sin( φ [ i ]) W zx G fx [ i ] + W zy G fy [ i ] + W zz G fz [ i ] + V z = cos( θ [ i ]) cos( φ [ i ]) , (3) cce l e r a t i o n X a x i s − . − . − . − . .
00 0 .
25 0 .
50 0 .
75 1 . A cce l e r a t i o n Y a x i s − . − . − . − . . . . . . A cce l e r a t i o n Z a x i s − . − . − . − . . . . . . A cce l e r a t i o n X a x i s − . − . − . − . .
00 0 .
25 0 .
50 0 .
75 1 . A cce l e r a t i o n Y a x i s − . − . − . − . . . . . . A cce l e r a t i o n Z a x i s − . − . − . − . . . . . . Figure 3. Data from an Analog Devices ADXL345 accelerometer fixed to the optical tube of an equatorially mountedtelescope, slewed through its full range of motion. Left: The raw data with standard factory calibration. Right: The dataafter 12-parameter calibration. All points now sit on the surface of a sphere of radius 1 g . with residuals, r x [ i ] = − sin( θ [ i ]) − W xx G fx [ i ] − W xy G fy [ i ] − W xz G fz [ i ] − V x r y [ i ] = cos( θ [ i ]) sin( φ [ i ]) − W yx G fx [ i ] − W yy G fy [ i ] − W yz G fz [ i ] − V y r z [ i ] = cos( θ [ i ]) cos( φ [ i ]) − W zx G fx [ i ] − W zy G fy [ i ] − W zz G fz [ i ] − V z . (4)If we now consider only the x -component, equation 4 can be simplified to [7, eq. 41], r x = Y x − X β x , (5)where r x is the M -length array of residuals to the calibration fit, Y x is the array of x -components of the gravitationfield for the true measured angle, X is the matrix of accelerometer measurements, and finally β x is the solutionvector for four of the calibration parameters. The optimum least squares solution for β x can be found by makinguse of the Normal Equations for least squares optimisation, β = ( X T X ) − X T Y , (6)such that [7, eq. 46], β x = W xx W xy W xz V x = ( X T X ) − X T Y x = ( X T X ) − X T − sin θ [0] − sin θ [1] . . . − sin θ [ M − , (7)and similarly for the remaining two axes. In order to solve for 12-parameters the minimum number of measure-ments M is four, and these should be well distributed in the measurement space. More measurements give amore robust calibration. The calibrated accelerations are subsequently determined using equation 1.Figure 3 shows a range of measurements made whilst the mount was slewed slowly through its full range ofmotion. Both factory calibration and improved 12-parameter calibration are shown. The roll and pitch anglescan be computed from the calibrated measurement matrix G using [6, eq. 25–26], φ xyz = tan − (cid:18) G y G z (cid:19) , (8)nd, θ xyz = tan − − G x q G y + G z , (9)where φ and θ are the measured roll and pitch angles respectively, and the subscript xyz notes that the anglesare according to the aerospace rotation sequence R xyz where rotation is first in yaw, then pitch, then roll.Equation 8 becomes unstable when the telescope is pointing near the zenith, since the x -axis becomes alignedwith the gravitational field vector and enters a condition known as Gimbal Lock. Any rotation in roll can no longerbe detected and roll becomes undefined. A common work-around for this problem when using aerospace rotationsequence R xyz is to modify equation 8 to include in the denominator a fraction µ of the x -axis measurementwhilst remembering to maintain the sign of G z after taking the square root [6, eq. 38], φ xyz = tan − G y ± p G z + µG x ! , (10)such that φ is slowly driven to zero as the telescope approaches a vertical orientation. An additional ambiguityis caused while the telescope is vertical, since with an equatorial mount it becomes impossible to determine ifthe telescope is on the east or west side of the pier. This ambiguity can be resolved by the addition of a secondaccelerometer installed directly on the polar axis of the mount.In testing, the accelerometer achieved an accuracy of approximately ±
4° in pitch and ±
6° in roll. The reducedperformance in roll is due to the above work-around at high pitch angles. Performance from both axes is similar ifonly moderate pitch angles are considered. Mészáros et. al. trialled the two-accelerometer technique by mountingtwo Freescale MMA8453Q accelerometers on their telescope at Konkoly Observatory located at the PiszkéstetöMountain station, and achieved accuracy to better than a degree with more sophisticated calibration, by makinguse of full temperature compensation, and taking good care of power supply stability via a custom printed circuitboard. The simpler treatment shown here is adequate for a solar autoguider with a capture angle of a few degrees.Accelerometers are capable of completely resolving telescope attitude only with an equatorial telescope mount.With an altitude-azimuth mount, only changes in altitude (pitch) can be measured. Rotations in azimuth (yaw)are aligned with the gravitational field vector and cannot be detected, and so a different sensor is required.Next, we look at using a three-axis magnetometer to detect the orientation of Earth’s magnetic field and directlymeasure the telescope “heading” angle. In order to detect rotation about a vector parallel to Earth’s gravitational field vector (i.e., yaw, heading, azimuth)we need to be able to detect rotation within Earth’s magnetic field, otherwise known as an electronic compass.We used a Honeywell HMC5883L three-axis magnetometer, which uses magneto-resistive sensors to measureboth the direction and the magnitude of Earth’s magnetic field. The Honeywell HMC5883L magnetometer isfactory calibrated to output values in units of µT, with a dynamic range of 800 µT, and an embedded 12-bitADC providing a documented 1° to 2° heading accuracy. As with the accelerometer, a magnetometer requiresa final in-situ calibration in order to achieve the best precision. A thorough discussion of calibration of athree-axis magnetometer is given in Refs. 9–12. We will now summarise the calibration process and go on toapply the techniques to the Honeywell HMC5883L, including tilt-compensation based on earlier results from theAnalog Devices ADXL345 accelerometer.A magnetic vector measured at the device B d can be defined after arbitrary device rotation in terms of thelocal geomagnetic field vector B by [9, eq. 5], B d = WR x ( φ ) R y ( θ ) R z ( ψ ) k B k cos δ δ + V , (11)where φ , θ , and ψ are roll, pitch, and yaw angles as previously, δ is the magnetic inclination at the measurementlocation, V is the “hard-iron” offset vector, and W is the “soft-iron” gain matrix. So-called hard-iron offsets areagnetic fields generated by nearby permanent magnets, such as other components on the PCB, and motors inthe telescope mount. These components are generally in fixed positions and rotate with the device, and so theyappear as an additive magnetic field vector within the reference frame of the magnetometer. So-called soft-ironinterference is due to temporary induction of magnetic fields in otherwise normally unmagnetised components,such as the sheet steel of the telescope mount and housing, caused by the geomagnetic field itself. Soft-iron effectsare much more complicated to model since they depend on the orientation of the device within the geomagneticfield, and typically suffer magnetic hysteresis effects as the device rotates. The soft-iron matrix W is a 9-elementmatrix similar to that used during calibration of the accelerometer in equation 1, and in addition to calibratingsoft-iron effects also calibrates rotation of the integrated circuit package and corrects for all possible cross-talkinteractions and gain variations between channels. In addition to removing offsets, the measurements from amagnetometer need to be derotated back to the flat plane where φ = θ = 0, since as with a typical analoguecompass it works only when held level. Rearranging equation 11 for ψ we get [9, eq. 6], R z ( ψ ) k B k cos δ δ = cos ψ sin ψ − sin ψ cos( ψ ) 00 0 1 k B k cos δ δ = R y ( − θ ) R x ( − φ ) W − ( B d − V ) , (12)and so [9, eq. 9], B fx B fy B fz = cos ψ cos δ k B k− sin ψ cos δ k B k sin δ k B k = R y ( − θ ) R x ( − φ ) W − ( B d − V ) , (13)where B f is equal to the magnetometer measurements with both soft-iron and hard-iron effects removed, andderotated to a flat plane where the z -component B fz is equal to k B k sin δ . The yaw angle, or compass heading,is found from [9, eq. 10], ψ = tan − (cid:18) − B fy B fx (cid:19) , (14)and this simply requires the addition of the known local magnetic declination to convert from magnetic north totrue north, and so obtain the azimuthal angle.In many cases the soft-iron effects are insignificant, and only the hard-iron offsets dominate. This allows asimplification during calibration since only four parameters need to be determined, and these are the magnitudeof the geomagnetic field strength k B k , and the three components of the hard-iron vector V . The soft-iron matrix W becomes the identity matrix. With these assumptions we can follow the same calibration procedure as forthe accelerometer, by developing a performance function to be minimised by optimising the calibration fit andagain using equation 6 to solve the fit through matrix algebra.Figure 4 shows a range of measurements made by a Honeywell HMC5883L magnetometer slewed randomlythrough the full range of motion. It is clear from the left panel of Figure 4 that the data do indeed suffer from ahard-iron offset, which moves the data away from the origin. After applying the 4-parameter calibration, shownby the blue dots in the right panel of Figure 4, the data now sit on the surface of a sphere centred at the origin,with radius equal to the local geomagnetic field strength in Birmingham calibrated at 54 . φ = θ = 0and the z -component B fz is equal to k B k sin δ . These data are now calibrated and ready for the heading tobe extracted using equation 14. The resulting accuracy after calibration is approximately ± ag n e t i c X a x i s − −
20 0 20 40 M ag n e t i c Y a x i s − −
20 0 20 40 M ag n e t i c Z a x i s − − M ag n e t i c X a x i s − −
20 0 20 40 M ag n e t i c Y a x i s − −
20 0 20 40 M ag n e t i c Z a x i s − − Figure 4. Data from a Honeywell HMC5883L magnetometer slewed randomly through the full range of motion. Left: Theraw data with a hard-iron offset. Right: The blue dots show the data after 4-parameter calibration removing the hard-ironoffset. All points now sit on the surface of a sphere centred at the origin, with radius equal to the local geomagneticfield strength in Birmingham calibrated at 54 . φ = θ = 0 and the z -component B fz is equal to k B k sin δ . After the Sun is brought within the capture angle of the autoguider camera by the coarse MEMS-based pointing,mount control is handled purely via imaged-based guiding much like any other telescope. The camera trialled hasa 6 . .
75 mm CCD with 4 .
65 µm square pixels in a 1392 by 1040 array. When coupled with an 80 mm focallength objective lens this produces an approximate field of view of 4 .
6° by 3 . field of view. The Sun has an extent of about 32 and so produces an image about 160 pixels in diameter onthe sensor. Two filters were used to bring the image within the dynamic range of the CCD chip. These were aneutral density filter with optical density of 5 (i.e, transmission of approximately 10 − ), and a 10 nm bandpassfilter centered on 780 nm. Images were read from the camera as frequently as possible, approximately once persecond. The image exposure time was 79 ms, but the cadence was restricted by the CCD read-out time and USBtransfer rate.Images from the camera were processed to determine the position of the Sun to sub-pixel accuracy usingthe OpenCV (Open-source Computer-Vision) library, running on an inexpensive single-board-computer. Thecentroid position of the solar disc was found using by applying a black and white threshold, finding the contoursin the image, and subsequently reading out the contour centroid position. The guiding error was determinedby comparing the current solar centroid position with a desired target value. The position error for each axiswas then passed through a proportional–integral–derivative (PID) control loop feedback mechanism in order todetermine the correct mount drive rate, and the mount motors updated with the new drive rate using the motorAPI published by the mount manufacturer. The PID control algorithm is a servo feedback system that can bedefined simply by, Output = K P e ( t ) + K I Z e ( t ) dt + K D d e ( t )d t , (15)where e is the control error defined as the desired setpoint minus the current value, and K P , K I , and K D are theproportional, integral, and derivative coefficients. The solar guiding algorithm made use of only the proportionaland integral parameters. −
20 0 20 40RA guider error (arcseconds)0 . . . . . P r o b a b ili t y d e n s i t y µ = 0 . x = 0 . σ = 6 . − −
20 0 20 40Dec guider error (arcseconds) µ = − . x = − . σ = 7 . Figure 5. Guider performance histogram. The dashed red line indicates the equivalent Gaussian for the measured medianand standard deviation.
The distribution of position-error for both axes, calibrated in arcseconds, is shown in Figure 5. The controlalgorithm achieves performance on both axes to better than ± when logged at an approximate 1 Hz cadenceover an entire day. This is more than sufficient for Sun-as-a-star observations when considering the Sun isover 1800 in observed diameter.By making use of these techniques is it possible to achieve full automation of consumer-grade telescopemounts. In the next section we will discuss prototype updates to the BiSON control systems.
3. SUPERVISORY CONTROL AND DATA ACQUISITION
In the early 1990s, the original network control system was based around a standard desktop PC runningMicrosoft DOS, and a Keithley System 570 digital input/output interface for connection to dome, telescope,and instrumentation control hardware. The computer acted as a centralised controller, handling all systemsfrom dome and mount pointing, to temperature stabilisation and data readout, using a Supervisory Control andData Acquisition (SCADA) monolithic control system architecture. The data would be retrieved each day over along-distance telephone call via dial-up modem, and later over the internet. The Keithley interface devices wereused until 2002, when the DOS PCs were replaced with the current systems running GNU/Linux with hardwareinterconnects predominately via RS-232, although a PCI general purpose digital-IO card is still used to interfaceto some legacy equipment.The so-called “Industry 4.0”, the fourth industrial revolution, has since brought about inter-connected ma-chines, devices, and sensors – the “Internet of Things” (IoT) – that have allowed control systems to becomedecentralised. By migrating from a monolithic to a microservice based architecture (MSA) in a DistributedControl System (DCS) it is possible to further reduce both the physical footprint, complexity, and deploymentcost. A microservice follows the Unix philosophy of “Do One Thing and Do It Well”, where each service is asingle purpose process capable of handling a request independently. Such decoupled processes allow a modularapproach to system design, reducing development into more manageable components and removes the need tounderstand the entire monolithic system. Independent modules also mean they can be independently upgradedor replaced, removing the common risk of being forced to redesign an entire system simply because one compo-nent has become obsolete. When all devices and sensors are connected to every other service, it can also becomepossible trigger maintenance processes autonomously by monitoring data from all points in the overall systemand predicting potential failures. Such flexibility is essential in the operation of a worldwide constellation ofsmall telescopes, that inevitably become heterogeneous due to gradual roll-out of repairs and upgrades. Figure 6shows a block diagram of the proposed MSA for BiSON:NG, and also for implementation at the existing networktelescopes.Communication between microservices is typically handled using lightweight protocols with well-defined in-puts and outputs. Most IoT devices communicate over internet protocol networks making use of industry eatherSensingSecure Edge DeviceEnclosureControlTimeseriesDatabaseAttitudeControl InstrumentControlWeatherSensingSecure Edge DeviceEnclosureControlTimeseriesDatabaseAttitudeControl InstrumentControl
WeatherSensingSecure Edge DeviceEnclosureControlTimeseriesDatabaseAttitudeControl InstrumentControl
Telescope 1 Telescope 2
Telescope n Remote Orchestration and TelemetryData Archive Platform HeathMonitoring
Public Outreach NotificationsOpen Data API
Figure 6. A block diagram of the proposed microservice architecture for BiSON:NG, that is easily scalable to many smalltelescopes in a globally distributed network. standard Ethernet or WiFi. This has a significant advantage in allowing many devices to be easily connectedusing standard network switches, rather than relying on the limited connectivity options available on a singlePC. Components such as sensors and controllers from different manufacturers need to communicate using stan-dardised communication protocols. As is common with standards there are several from which to choose, suchas OPC UA (Open Platform Communications Unified Architecture), DDS (Data Distribution Service), MQTT(Message Queuing Telemetry Transport), CoAP (Constrained Application Protocol), Extensible Messaging andPresence Protocol (XMPP), Advanced Message Queuing Protocol (AMQP), and others. Whilst there is noone best protocol, we have chosen to use MQTT due to it being designed for low-bandwidth connections fromremote locations. It is extremely lightweight, with open source examples of implementation on micro-controllerswhere a small code footprint is essential, and single-board computers, both of which we use extensively.MQTT is an ISO standard client/server publish/subscribe protocol developed by IBM. The protocol re-quires a message broker through which clients communicate, with messages organised into topics. Messages arebroadcast on a one-to-many distribution basis. The protocol can scale easily up to hundreds or even thousandsof devices, allowing decoupling of applications since any service with an interest in a particular data feed cansimply subscribe to the relevant topic. Several quality-of-service (QoS) options are available, ensuring messagesre delivered either at most once, at least once, or exactly once, providing robust communication even overhigh-latency or unreliable networks. Typically only a single broker is required, although multiple brokers can belinked for either redundancy, load sharing, or connecting between local and remote sites. There are also securityadvantages to using MQTT, since communication can be easily authenticated over a Transport Layer Security(TLS) encrypted channel. The broker manages all security credentials and certificates, and also tracks the clientconnection states allowing rapid notification of a disconnected service.A number of free and open-source packages are used where possible. Data archival is handled both locally ateach telescope, in order to survive network outages, and remotely for aggregation. A timeseries-specific databaseis used, such as Timescale and InfluxDB, to ensure performance at scale. Platform monitoring and telemetryvisualisation is produced by Grafana, with system log messages written locally and, where necessary, publishedto a Slack channel for urgent notification. Public outreach is possible through automatic status updates viasocial media such as Twitter. Through the upgrades discussed here, we move from a full rack of electronics to physically small servicesclosely coupled near their respective areas of instrumentation.
4. CONCLUSION
The Birmingham Solar Oscillations Network has achieved an average annual duty cycle of around 82% sincecommissioning in 1992, providing an unparalleled baseline of unresolved-Sun helioseismic observations. Theaim of observing potential solar gravity-modes ( g -modes) requires much lower noise levels over long time periodsthan currently achieved by any Sun-as-a-star observations, since they are expected to have very low amplitudesand low frequencies. The instrumental noise level is dominated by atmospheric scintillation and that of solarorigin, but both can be beaten down by combining multiple incoherent measurements from many simultaneousobservations,
23, 24 and so there is a need to considerably increase the number of BiSON observing sites.We have shown here that it is possible to achieve full automation of an inexpensive consumer-grade tele-scope mount through the addition of MEMS inertial sensors, within a small physical package requiring onlya basic weather-proof automated enclosure. A microservices control architecture allows the control systems torun entirely on inexpensive single-board computers and micro-controllers, removing the need for a full rack ofelectronics and again reducing cost. Initial trials of these small form factor techniques at three existing BiSONsites have shown an improvement in performance through the reduction of certain noise sources such as guidererrors.
2, 25–28
In order to ensure that BiSON is well placed to observe the next solar cycle and beyond, continuousdevelopment and improvement is essential.The impact of BiSON:NG extends far wider than the field of solar astronomy. A large network of smallinexpensive robotic telescopes can easily be made dual purpose, for both solar and stellar astronomy. Telescopesas small as 6 inches have research applications where, perhaps operated by schools and Citizen Scientists, theyare capable of follow-up observations of exoplanet transits.
29, 30
Dramatically increasing the number of telescopesavailable to students for education, outreach, and public engagement makes it possible for young people to accessand participate in research, and can help to encourage people from all backgrounds into a science career.
BiSON:NG offers an unprecedented opportunity for multifaceted science, engagement, and collaboration.
ACKNOWLEDGMENTS
We would like to thank all those who have been associated with BiSON over the years. We particularly ac-knowledge the technical assistance at our remote network sites, with sincere apologies to anyone inadvertentlymissed: At Mount Wilson: Ed J. Rhodes, Jr., Stephen Pinkerton, the team of USC undergraduate observingassistants, former USC staff members Maynard Clark, Perry Rose, Natasha Johnson, Steve Padilla, and ShawnIrish, and former UCLA staff members Larry Webster and John Boyden. At Las Campanas: Patricio Pinto,Andres Fuentevilla, Emilio Cerda, Frank Perez, Marc Hellebaut, Patricio Jones, Gastón Gutierrez, Juan Navarro,Francesco Di Mille, Roberto Bermudez, and the staff of LCO. At Izaña: Pere Pallé, Teo Roca Cortés, AntonioPimienta, and the team of operators who have contributed to running the Mark I instrument over many years. AtSutherland: Pieter Fourie, Willie Koorts, Jaci Cloete, Reginald Klein, John Stoffels, Brendt Christian, and thestaff of SAAO. At Carnarvon: Les Bateman, Les Schultz, Sabrina Dowling-Giudici, Inge Lauw of Williams andughes Lawyers, and NBN Co. Ltd. At Narrabri: Mike Hill and the staff of CSIRO. The authors are grateful forthe financial support of the Science and Technology Facilities Council (STFC), grant reference ST/R000417/1.Additional funding was secured via the STFC Impact Accelerator account, with the assistance of Alan Tibbattsfrom University of Birmingham Enterprise. Funding for the Stellar Astrophysics Centre (SAC) is provided byThe Danish National Research Foundation, grant reference DNRF106.
REFERENCES [1] Hale, S. J., Howe, R., Chaplin, W. J., Davies, G. R., and Elsworth, Y. P., “Performance of the BirminghamSolar-Oscillations Network (BiSON),” Sol. Phys. , 1–28 (Jan. 2016).[2] Hale, S. J.,
Birmingham Solar Oscillations Network: The Next Generation , PhD thesis, School of Physicsand Astronomy, University of Birmingham, UK (July 2019).[3] Hale, S. J., “Fibre-feed tests at Mount Wilson in 2016 September,” BiSON Technical Report Series, Number382, High-Resolution Optical-Spectroscopy Group, University of Birmingham, UK (Dec. 2016).[4] Hale, S. J., “Sunrise on Research.” Gatsby Charitable Foundation Trust (May 2017).[5] Ash Manufacturing Company, “Observatories and tracking instrument shelters,” (Sept. 2020).[6] Pedley, M., “Tilt Sensing Using a Three-Axis Accelerometer,” Application Note 3461, Freescale Semicon-ductor, Inc. (Mar. 2013).[7] Pedley, M., “High-Precision Calibration of a Three-Axis Accelerometer,” Application Note 4399, FreescaleSemiconductor, Inc. (Oct. 2015).[8] Mészáros, L., Jaskó, A., Pál, A., and Csépány, G., “Accurate Telescope Mount Positioning with MEMSAccelerometers,” PASP , 769–782 (Aug. 2014).[9] Ozyagcilar, T., “Calibrating an eCompass in the Presence of Hard- and Soft-Iron Interference,” ApplicationNote 4246, Freescale Semiconductor, Inc. (Nov. 2015).[10] Ozyagcilar, T., “Layout Recommendations for PCBs Using a Magnetometer Sensor,” Application Note 4247,Freescale Semiconductor, Inc. (Nov. 2015).[11] Ozyagcilar, T., “Implementing a Tilt-Compensated eCompass using Accelerometer and Magnetometer Sen-sors,” Application Note 4248, Freescale Semiconductor, Inc. (Nov. 2015).[12] Ozyagcilar, T., “Accuracy of Angle Estimation in eCompass and 3D Pointer Applications,” ApplicationNote 4249, Freescale Semiconductor, Inc. (Nov. 2015).[13] Bradski, G., “The OpenCV Library,”
Dr. Dobb’s Journal of Software Tools (2000).[14] Profanter, S., Tekat, A., Dorofeev, K., Rickert, M., and Knoll, A., “OPC UA versus ROS, DDS, and MQTT:Performance Evaluation of Industry 4.0 Protocols,” in [ ], 955–962 (2019).[15] International Organization for Standardization, “Information technology — Message Queuing TelemetryTransport (MQTT) v3.1.1,” (June 2016).[16] Timescale, “PostgreSQL for time-series,” (Oct. 2018).[17] InfluxData, “InfluxDB: An open-source time series database,” (Sept. 2013).[18] Grafana Labs, “Grafana: The open observability platform,” (2014).[19] Slack Technologies, “Slack: Brings all your communication together,” (Aug. 2013).[20] Twitter, Inc., “Twitter: A microblogging and social networking service,” (Mar. 2006).[21] Appourchaux, T., Belkacem, K., Broomhall, A. M., Chaplin, W. J., Gough, D. O., Houdek, G., Provost, J.,Baudin, F., Boumier, P., Elsworth, Y., García, R. A., Andersen, B. N., Finsterle, W., Fröhlich, C., Gabriel,A., Grec, G., Jiménez, A., Kosovichev, A., Sekii, T., Toutain, T., and Turck-Chièze, S., “The quest for thesolar g modes,” A&A Rev. , 197–277 (Feb. 2010).[22] Hale, S. J., Chaplin, W. J., Davies, G. R., Elsworth, Y. P., Howe, R., and Pallé, P. L., “Measurementof Atmospheric Scintillation during a Period of Saharan Dust (Calima) at Observatorio del Teide, Izaña,Tenerife, and the Impact on Photometric Exposure Times,” PASP , 034501 (Jan. 2020).[23] Davies, G. R., Chaplin, W. J., Elsworth, Y., and Hale, S. J., “BiSON data preparation: a correction fordifferential extinction and the weighted averaging of contemporaneous data,” MNRAS , 3009–3017 (July2014).24] Lund, M. N., Chaplin, W. J., Hale, S. J., Davies, G. R., Elsworth, Y. P., and Howe, R., “Spatial incoherenceof solar granulation: a global analysis using BiSON 2B data,” MNRAS , 3256–3263 (Dec. 2017).[25] Hale, S. J., “Fibre-feed tests at Izaña in 2017 May,” BiSON Technical Report Series, Number 384, High-Resolution Optical-Spectroscopy Group, University of Birmingham, UK (Sept. 2017).[26] Hale, S. J., “Fibre-feed tests at Izaña in 2017 September,” BiSON Technical Report Series, Number 385,High-Resolution Optical-Spectroscopy Group, University of Birmingham, UK (Sept. 2017).[27] Hale, S. J. and Ross, E., “Jabba fibre conversion at Carnarvon in 2018 April,” BiSON Technical ReportSeries, Number 389, High-Resolution Optical-Spectroscopy Group, University of Birmingham, UK (June2018).[28] Hale, S. J., “Klaus fibre conversion at Mount Wilson in 2018 June,” BiSON Technical Report Series, Number390, High-Resolution Optical-Spectroscopy Group, University of Birmingham, UK (Aug. 2018).[29] Dragomir, D., Harris, M., Pepper, J., Barclay, T., Villanueva, S., Ricker, G. R., Vanderspek, R., Latham,D. W., Seager, S., Winn, J. N., Jenkins, J. M., Ciardi, D. R., Furesz, G., Henze, C. E., Mireles, I., Morgan,E. H., Quintana, E. V., Ting, E. B., and Yahalomi, D., “Securing the Legacy of TESS through the Careand Maintenance of TESS Planet Ephemerides,” The Astronomical Journal , 219 (Apr. 2020).[30] Zellem, R. T., Pearson, K. A., Blaser, E., Fowler, M., Ciardi, D. R., Biferno, A., Massey, B., Marchis, F.,Baer, R., Ball, C., Chasin, M., Conley, M., Dixon, S., Fletcher, E., Hernandez, S., Nair, S., Perian, Q.,Sienkiewicz, F., Tock, K., Vijayakumar, V., Swain, M. R., Roudier, G. M., Bryden, G., Conti, D. M., Hill,D. H., Hergenrother, C. W., Dussault, M., Kane, S. R., Fitzgerald, M., Boyce, P., Peticolas, L., Gee, W.,Cominsky, L., Zimmerman-Brachman, R., Smith, D., Creech-Eakman, M. J., Engelke, J., Iturralde, A. r.,Dragomir, D., Jovanovic, N., Lawton, B., Arbouch, E., Kuchner, M., and Malvache, A., “Utilizing SmallTelescopes Operated by Citizen Scientists for Transiting Exoplanet Follow-up,” PASP , 054401 (May2020).[31] Sadler, P. M., Gould, R. R., Leiker, P. S., Antonucci, P. R. A., Kimberk, R., Deutsch, F. S., Hoffman, B.,Dussault, M., Contos, A., Brecher, K., and French, L., “MicroObservatory Net: A Network of AutomatedRemote Telescopes Dedicated to Educational Use,”
Journal of Science Education and Technology , 39–55(Mar 2001).[32] Gould, R. R., Sunbury, S., Sienkiewicz, F., Deutsch, F., Ibrahim, A., Dussault, M., and Krumhansl, R.,“The Other Worlds Project: Students Detect Exoplanets Using the MicroObservatory Online Telescopes,” in[ Earth and Space Science: Making Connections in Education and Public Outreach ], Jensen, J. B., Manning,J. G., and Gibbs, M. G., eds.,
Astronomical Society of the Pacific Conference Series , 162 (Sept. 2011).[33] Gomez, E. L. and Fitzgerald, M. T., “Robotic telescopes in education,”
The Astronomical Review13