(This online version of the project report is not as nice as the full, printer-ready, 37-page Postscript copy. You can retrieve aurora.ps.gz, a gzip-compressed version of it, after browsing this copy.)

The Aurora Alarm:
An Automated Detection System for the Northern Lights

by
Mark A. Haun

Prepared for
School of Engineering
Walla Walla College
College Place, Washington

Submitted in Partial Fulfillment of
the Requirements in
ENGR 496, 497, 498 Engineering Seminar
6 June 1996

Abstract

To increase one's chances of seeing the northern lights, the uncertainties of existing aurora prediction methods can be bypassed by sensing the light from the aurora directly. The Aurora Alarm is a device which accomplishes this objective through the use of a narrowband optical filter, a photomultiplier tube detector, and a flexible microprocessor system for high voltage control and alarm determination. It features a wide dynamic range, flexible alarm levels, serial interfacing to a PC or simplified direct hookup to an alarm, remote programmability, single-supply operation, and a built-in heater for operation in cold weather. A prototype incorporating a compact printed circuit board design and core software routines shows good performance with numerous possibilities for future applications.

Acknowledgments

The successful completion of the Aurora Alarm reflects the contributions of many people. I would like to thank my project advisor, Professor Carlton Cross, for his guidance in every phase of this project. During the mechanical design and construction of the prototype, the practical suggestions and assistance provided by Professors Ralph Stirling, Don Riley, and Ryan Mowat, Jim Forsyth at Technical Support Services, and fellow student Andrew Depaula saved me countless hours. I also wish to thank Physics Professors Claude Barnett and Gordon Johnson for their willingness to answer my optics questions and provide lab space, and Robert Batten for his Mentor Graphics tutoring.

A special thanks goes to Ray Muller at Hamamatsu Corp. for the generous donation of the photomultiplier tubes and high voltage power supplies used in this project. I am also indebted to Bill Grenier at the Andover Corp. and Calvin Grandy at Omega Optical Inc. for their donation of the optical filters. Without their support, this project would not have been possible.

Mark Haun
College Place, WA
4 June 96

1   Introduction

The beauty and power of the aurora capture the interest of many people, particularly those living at northerly latitudes where the phenomenon is known as the northern lights. Where auroral displays are less frequent, as they are in most areas of the US, it becomes very difficult to observe an aurora without either a large investment of time or some means of prediction (see Figure 1). At the latitude of Walla Walla College, auroral displays are not as rare as most people think, but they do occur while most people are asleep, indoors, or under light-polluted city skies. It is likely that there would be considerable popular interest in observing the aurora if people knew when to look.

Current prediction techniques rely on the link between solar activity and magnetic disturbances on earth, which are closely tied to the occurrence of the aurora. For example, a large solar flare is often followed within two or three days by enhanced auroral displays. Fluctuations in the strength and direction of the earth's magnetic field (so-called "magnetic storms") are also good predictors of auroral activity. These and other solar and geophysical data are widely available over the Internet and shortwave radio; Today's Space Weather, a service of the Space Environment Center, is a good example. While predictions based on such indicators increase one's chances of observing an aurora, they still suffer a high failure rate. Furthermore, it becomes very time-consuming to follow this information on a daily basis, and if a dark-sky site is not located nearby, considerable time may be spent chasing down false alarms.

By detecting the light of the aurora directly, it is possible to bypass the shortcomings and uncertainties of other prediction methods. A single reference in the amateur literature1 reported successful results with this method and included an all-analog design which is now somewhat outdated. Microprocessor control of an optical aurora detector seemed to offer many advantages. Such a device could be programmed intelligently to alert the user to auroral activity worth watching given the current sky conditions, moonlight interference, and other light pollution. A computer interface would make data recording easy and allow flexible user alerting through email or automatic telephone call-up. With these possibilities in mind, and because of the author's own desire to improve his chances of seeing an aurora, the Aurora Alarm project was born.

Figure 1: How to See an Aurora
Possible solutions Associated problems
Move north Inconvenient
Look outside every night Sleep loss, light pollution
Check forecasts of solar
and geomagnetic activity 		Time consuming, unreliable
Direct optical detection Technical challenge

2   Design Objectives

The primary objective of the Aurora Alarm project was to design an instrument capable of detecting the light from the aurora and alerting the user to any activity strong enough to be worth going outside to see. Several design goals were laid out from the beginning:

3   Hardware Design

Figure 2: Simplified Block Diagram

Figure 2: Simplified Block Diagram

 Figure 2 is a block diagram of the Aurora Alarm (a complete schematic can be found in Appendix A). The major components form a digital control system in which a microprocessor monitors the detector signal and takes control action to regulate it by changing the detector's sensitivity. The design can be divided into three parts: the electro-optical system, the microprocessor system, and the temperature controller.

Since this project was completed to fulfill the requirements of the Electrical Engineering concentration at WWC, this report will focus primarily on the electrical design of the system, summarizing other facets of the design where necessary. In particular, the details of the mechanical construction, which took a large share of the project time, will not be discussed in depth.

The optical system

3.1   The Electro-Optical System

The photon flux from an aurora is very low, but the energy is strongly concentrated in only a few spectral lines. Based on published measurements of auroral brightness,3 a photomultiplier tube (commonly abbreviated PMT) was selected as the light detector (see Appendix B for the actual sensitivity calculations). This device operates on the principle of the photoelectric effect, in which electrons are generated by photons striking a photocathode. In the PMT, the photocathode is followed by a series of electrodes, called dynodes, which are biased successively at higher and higher positive potentials by about 100 V per stage. Electrons striking a dynode can generate multiple secondary electrons for the next stage, and the PMT uses this electron multiplication, or avalanche, process to achieve gains approaching 107 internal to the tube. The actual electron gain can be varied from this maximum all the way down to unity by changing the high voltage supplied to the tube. The gain versus supply voltage curve is roughly logarithmic.4

A narrowband optical interference filter with a half-power bandwidth of 10 nm is used to increase the signal-to-noise ratio available to the PMT. It is centered on the strong auroral oxygen emission line at 557.7 nm, and greatly attenuates starlight, moonlight, and local light pollution while passing the greenish-yellow color which dominates the human visual response to most auroras. The filter actually has a nominal center wavelength of 560 nm, but the passband can be shifted slightly to the lower wavelength by tilting the filter, because it operates on the basis of interference in thin films. The proper tilt is determined empirically with a spectrograph calibrated against sources having emission lines of known wavelengths. White light can then be shone through the filter and the tilt adjusted to center the filtered spectrum around 557.7 nm. The zero-tilt transmittance curve of the filter used is included as Appendix C.

To sense the overall sky brightness, a small photodiode is used as a secondary detector. It is placed in front of the filter and so is exposed directly to the ambient light entering the optical system at all wavelengths. Its operation is not critical to the detection of auroras, but it allows for intelligent alarm level compensation based on the sky conditions, amount of moonlight, etc.

The complete optical assembly consists of the optical filter mounted in a bracket with an adjustable tilt, a converging lens of similar diameter as the filter, the photomultiplier tube, and the photodiode. The filter/lens assembly can be adjusted up or down on threaded bolts and is positioned so that the PMT photocathode is located at the focus of the lens. A short section of pipe with a clear plastic cover cemented to one end is used to cover and protect the optical elements; the base of the pipe is fixed over a hole drilled in a diecast aluminum box which holds the PMT, high voltage supply, and circuit board.

High voltage to the PMT is supplied from a DC-to-DC converter manufactured by Hamamatsu Corp. which is built into a standard PMT base/socket and includes the necessary voltage divider for supplying each PMT dynode voltage. A low voltage input of 2-8 V is converted by this supply into -300 V to -1100 V for the PMT. The microprocessor system can control this voltage in software as will be discussed later.

Figure 3: Analog Signal Processing Circuitry

Figure 3: Analog Signal Processing Circuitry

Figure 3 shows the analog processing circuitry for the PMT and photodiode signals. The PMT signal is first amplified by a transresistance op-amp stage with a gain of 106 and a time constant of 1 s. It is followed by a standard noninverting stage with a variable gain between 1 and 100. The photodiode amplifiers are the same except for a higher gain in the transresistance stage 107. The large time constants are used to reduce noise since the signals of interest are only changing slowly.

Both signals are digitized to a full scale range from 0-4.095 V. For the PMT, this places the anode current for full scale output between 40 nA and 4 µA. This is entirely appropriate as the absolute maximum rating for anode current on the PMT is 0.1 mA and an average current of less than 1 µA is recommended. The photodiode current for full scale output ranges from 4 to 400 nA, well above the specified dark current.

In order to achieve single supply operation, CMOS op-amps were chosen with rail to rail output swing and input common-mode tolerance to both rails. When powered from regulated +5 V and ground, they have the additional advantage of interfacing nicely with the unipolar 0-4.095 V inputs of the analog-to-digital converter.

3.2   The Microprocessor System

Figure 4: Digital and Interface Circuitry

Figure 4: Digital and Interface Circuitry

 A DS5000T microcontroller from Dallas Semiconductor forms the heart of the Aurora Alarm's microprocessor control system (see Figure 4). This module is 8051 compatible, which enabled much of the code development to be done using the 8051 development tools available at WWC. The DS5000T offers several features that were helpful in meeting the design goals. All of the program and data memory in the DS5000T is battery backed-up RAM, and a boot loader is included which allows the part to be programmed in-circuit over the serial interface. Also included are additional power management functions, a watchdog timer, and a real time clock.5

Since all of the necessary memory is included inside the DS5000T, no external bus is needed, thus freeing all of the port pins for other functions. To allow for future design changes, however, only ports one and three are used in this design. The two transistors in Figure 4 are used to manipulate the RST and /PSEN lines in a special way. When the /PROG signal at the interface connector is grounded, both transistors turn on, pulling RST high and /PSEN low. This puts the DS5000T into its serial boot loader mode for program updates.

The Aurora Alarm's serial interface is a simple three wire (RXD, TXD, and ground) type. The 8051-style serial port on the DS5000T is used, along with a MAX231 level converter to handle the voltage translation between RS232 and TTL levels. The microprocessor is clocked at 7.3728 MHz which is a "magic number" for baud rate generation and allows the use of most standard serial port speeds. The spare pins on port three are available for general purpose bit I/O; four of these are routed to the interface connector and defined as auxiliary inputs and outputs for applications which do not use the serial port. Series resistors are provided in these lines for current limiting, and schottky diodes prevent the CMOS inputs from falling much below ground.

Figure 5: A/D, D/A, and High Voltage Control Circuitry

Figure 5: A/D, D/A, and High Voltage Control Circuitry

Port one on the DS5000T is used for communication with the 12-bit A/D and D/A converters, which connect the processor to the PMT and photodiode signals and provide for control of the high voltage supplied to the PMT (see Figure 5). Since both chips use serial communication interfaces, it was possible to use just seven port pins on the microprocessor to control both functions.

The MAX186 12-bit A/D has eight single-ended input channels of which only three are currently used. Channel 0 is assigned to the PMT and channel 1 is assigned to the photodiode. Channel 2 is connected to the thermistor voltage divider in the temperature control circuit; this is an addition to the original circuit and is not shown on the schematics included here. The chip is configured for unipolar analog inputs with a range of 0-4.095 V -- a good match for the CMOS op-amp outputs. The on-chip voltage reference is used to supply a 4.095 V reference to both the A/D and D/A.

The DAC8043 is a simple 12-bit D/A with a current output. It is used to drive the high voltage control circuit which must supply a controlled voltage between 0 and 8 V to the high voltage supply (DC-to-DC converter). The use of single-supply op-amps in this situation poses a problem since it is desired to use the inverting input as a current summing point. All voltages in the circuit must be positive, so a current sink at the summing point is needed, yet only current sources are available since the summing point is at ground. To get around this problem, an ICL7660 "positive-to-negative" voltage converter is used to provide a voltage below ground which can sink current through a resistor connected to the op-amp inverting input. In the final configuration, a D/A value of all zeros requires 8 V at the high voltage supply input. A D/A value of all ones supplies enough current into the summing point to force the high voltage supply input below 2 V, effectively turning off the high voltage to the PMT.

Great care must be taken to insure that the PMT anode current does not exceed the absolute maximum rating of 0.1 mA, as might happen if it is exposed to bright light with the high voltage supply turned on. A certain amount of protection can be obtained in software, since the microprocessor can take action to reduce the high voltage if the PMT output becomes too high. A brute-force method is required in the case of sudden light changes, however, and this is provided by the 400 ohm current limiting resistor in series with the 2N2222 pass transistor. It limits the current into the high voltage supply at about 20 mA, which in turn limits the PMT supply current to safe levels.

3.3   The Temperature Controller

Figure 6: Temperature Control Circuitry

Figure 6: Temperature Control Circuitry

 Figure 6 is a schematic of the Aurora Alarm's temperature control system. If the unit is operated in conditions which put components at risk of temperatures much below freezing, the temperature adjustment can be set to supply enough extra heat in the enclosure to maintain it at the desired temperature.

The op-amp is operated here as a comparator with positive feedback to provide some hysteresis. The thermistor in the voltage divider connected to the noninverting input has a large negative temperature coefficient. As the temperature falls, the thermistor resistance will eventually increase to the point where the noninverting input to the comparator has a higher voltage than the inverting input supplied by the reference divider. When this happens, the comparator will switch states, driving the transistor into saturation and dissipating about 1.3 W of heat in the heating resistors until the temperature again increases to the set point.

Circuit board and PMT in enclosure

4   Prototype Implementation and Performance

4.1   Printed Circuit Board Design

An important part of this project was the layout of a printed circuit board, as it gave me the chance to use the Mentor Graphics software available at WWC. This software is very complex, and learning to use it took a substantial portion of the total time spent on the project. A computer-aided layout has definite advantages, however.

The first step was to enter the schematic into the Design Architect module. When this was done and the completed diagram had been checked for consistency, the Boardstation module and other helper applications were used to define the physical outline and pin definitions of each component, the sizes of the pads and traces to be used, and general board parameters like the overall size and the number of layers. Components were then placed on the board and traces routed. Since the schematic was linked to the routing process at all times, the software was able to prevent mistakes such as connecting the wrong pads or forgetting to connect others. The end result was a PCB layout guaranteed to be error free (assuming a correct schematic). The ability to rotate and move components at will prior to routing resulted in a compact, high density board. The final printed circuit board for the Aurora Alarm is a two layer board with plated-through holes and measures about three by four inches. The A/D and D/A chips define a clear boundary between analog and digital areas of the board. The analog and digital regions have separate ground planes which are connected together at a single ground point. Computer-aided PCB layout proved effective: after stuffing the board with components and powering it up, it worked the first time!

4.2   Software Design

A complete listing of the DS5000T assembly code used for testing the Aurora Alarm is given in Appendix D. It includes all of the core routines for such functions as initializing the hardware, reading an A/D value, writing a D/A value, getting the current time from the real time clock, performing 16-bit arithmetic operations, converting binary numbers to ASCII decimal representation, and sending data over the serial port.

The current version of the code implements a simple digital controller which attempts to keep the PMT output at a fixed level, currently set to one volt input at the A/D. Proportional control action is taken at each sample to vary the PMT high voltage supply as required to keep the PMT output constant. With this type of control algorithm, the high voltage level itself becomes the output of the system, providing a convenient logarithmic brightness scale due to the logarithmic gain vs. supply voltage dependence of the PMT.

At each sample, the software generates a one line status message with the current values for the high voltage control, PMT output, photodiode output, and temperature. All of these are given in human-readable decimal form and can be easily imported into graphing packages for later data analysis. The time required to send one line of status information determines the time between samples; at 1200 bps a sample rate of about 1 Hz is obtained.

Since the software has been used mainly for lab testing so far, no alarm algorithms have yet been programmed. These can be implemented in a host PC with the current DS5000T code or added to the DS5000T code itself. The exact choice and algorithm used will depend on the requirements of a specific installation. In most cases, a simple adjustable threshold would work, but it might be desirable to vary the alarm threshold automatically with data from the photodiode about current sky conditions.

4.3   Performance

The true test of the Aurora Alarm will be the detection of an actual aurora. Since this project was carried out during the bottom of the eleven year solar activity cycle and auroras are not common at WWC to begin with, lab tests and theoretical calculations had to suffice. It is anticipated that at least one real aurora will be detected during the first year of operation, which was about to begin as this report was being written.

The factors determining the sensitivity of the PMT to auroral light are straightforward (see Appendix B). The photon flux at the aperture, angular field of view, transmittance of the filter, quantum sensitivity of the PMT, and current amplification of the PMT are all reasonably well-defined quantities and are sufficient to establish the ability to detect auroras. Lab tests in a darkroom confirmed the unit's ability to sense very low light levels near the limits of human visual perception. An "artificial aurora" composed of several green LEDs with peak emission around 560 nm caused a measurable change in the high voltage level from many feet away even in the presence of normal indoor room light.

The performance of the control loop is less predictable. In theory, the system should have dominant first-order behavior due to the long time constant in the signal processing circuits. In practice, however, the system can go unstable if the proportional gain is set too high, with the high voltage alternately saturating at the upper and lower limits. Since speed is not at all critical in this application, a detailed analysis of the system was not performed; the proportional gain was simply lowered far enough to make the system stable.

Figure 7: High Voltage Control System Performance

Figure 7: High Voltage Control System Performance

Figure 7 is a graph of the PMT signal and the high voltage during a typical test run in the darkroom. At the start of the run, the detector was in total darkness. The PMT signal remained at its lower limit (an offset generated in the amplifier) while the high voltage ramped up and eventually saturated since there was not enough light to yield a PMT output of 1.0 (the set point in arbitrary units depending on the gain setting).

About 47 s into the run, a single green LED on a computer keyboard was uncovered and allowed to reflect off of the ceiling. At this point, the PMT signal jumped above 1.0 and the high voltage decreased to compensate, achieving equilibrium in about 10 s.

About 75 s into the run, the room light was suddenly switched on. This immediately caused the PMT output to saturate and the high voltage to fall rapidly. Equilibrium was established about 25 s later.

The overall performance of the Aurora Alarm has met the original design objectives. The PMT provides good response to an ample range of light levels -- nearly seven orders of magnitude compared with visible auroras which only span about four. The auxiliary photodiode and microcontroller allow for a variety of alarm algorithms. The serial interface works well and allows for remote code updates, but the simple digital I/O can be utilized just as easily for a simpler interface. During cold winters, the heater can keep critical components safe and warm. Finally, everything is powered from a single 12 V supply.

Aurora monitor installation overlooking Walla Walla

5   Future Applications

With creative interfacing and programming, the Aurora Alarm can form the heart of an elaborate aurora warning network. Serial data transmitted from the unit to a host computer can be distributed in a number of ways. A computer with a modem could be used to send alarm notices to a paging service. At WWC, an Internet server is being written which will accept connections at a special TCP port and rebroadcast the aurora data anywhere in the world in real time. Another method being investigated is the use of a voice mail system, which would maintain a database of phone numbers to call with a pre-recorded message when contacted by the aurora host computer.

Intelligent programming can be used to tailor alarm notifications to the wishes of particular users. Two or three alarm thresholds might be defined so that the more casual observers will not be notified of activity which is just barely visible. Users of the system could even submit "on call" schedules if they do not wish to be awakened during certain hours.

The Aurora Alarm is designed to make it easy to see an aurora. If it enables more people to experience the awe and wonder of the northern lights, then this project has been a success.

A   Schematic

This full schematic is only available in encapsulated postscript form. The quality is much better than that of the excerpts in this online version.

B   Sensitivity Calculations

Visible aurora intensities can be classified with a four level "International Brightness Coefficient" (IBC) of I, II, III, or IV according to the strength of the strong green line at 5577 Å. The faintest visible auroras have an IBC of I, corresponding to an approximate 5577 Å photon flux on the detector of I = 4 x 107 photons/s (approx; assuming .05 steradian field of view and detector area of 10 cm2 which are ballpark values for the Aurora Alarm).

At this wavelength,

Therefore, the anode current is approximately

I T N G q = (4 x 107 photons/s)(0.5)(0.03 e-/photon)(107)(1.6 x 10-19 C/e-) = 1 µA

This is well within range of the signal processing circuitry which follows the PMT. It is therefore safe to assume that the signal to noise ratio available to the detector, and not its sensitivity, will be the limiting factor for the detection of auroras. The S/N ratio is more difficult to quantify since it involves so many factors: moonlight, airglow, light pollution from towns and cities, the spectrum of all of these sources, and the bandwidth of the optical filter used.

C   Optical Filter Transmittance Characteristic

(Not yet available in online version.)

D   DS5000T Assembly Code

Two versions are available:

References

[1] Knight, Jesse W.   Monitoring the aurora electronically. Sky and Telescope 63:6, pp. 635-638.

[2] Omholt, Anders.   The Optical Aurora. Berlin: Springer-Verlag, 1971, pp. 5-6.

[3] Chamberlain, J. W.   Physics of the Aurora and Airglow. Academic Press, 1961, pp. 569-571.

[4] Photomultiplier tubes catalog and data book, 1994.   Hamamatsu Corporation, 360 Foothill Road, P.O. Box 6910, Bridgewater, NJ 08807-0910, 1-908-231-0960.

[5] Soft microcontroller data book, 1993.   Dallas Semiconductor, 4401 South Beltwood Parkway, Dallas, TX 75244-3292, 1-214-450-0448.

<-- Back to the aurora monitor page

You are the 34060th person to see this page.

Last updated 15 March 1999 by Mark Haun
Comments? Mail them to markhaun@uiuc.edu