satellite was designed to make high spatial and temporal resolution observations of particles and
electromagnetic fields in the auroral zone in the altitude range from 350 km to 4175 km to study
small-scale plasma interactions in detail. The dc and ac electric fields are measured with
three-axis dipole antennae with 56 m, 8 m, and 5 m baselines. A three-axis fluxgate magnetomete
measures the vector dc magnetic field and a three-axis search coil measures the vector ac magnetic
field.  The fields processing system receives signals from the electric field probes and the  field
sensors that cover a frequency band from dc to ~4 MHz. There are several types of processed data.
Survey data has continuous coverage over the auroral zone (and full orbit coverage for fluxgate
magnetometer only). Burst data covers a few minutes of the most intense region of the auroral zone
at the highest time resolution. A subset of the burst data, high speed burst memory data, is the highest
resolution waveform data at 2x106 samples per second. Electric field and magnetic field data are
primarily waveforms and power spectral density as a function of frequency and time. There are also
several types of focused data processing, including cross spectral analysis, fine-frequency plasma
wave tracking, high frequency polarity measurement, and wave-particle correlations.

I. Introduction

The primary scientific goal of the FAST mission is to study auroral plasma processes at high time
resolution.  Specifically, the mission was designed to investigate auroral electron acceleration, ion
heating, and wave-particle interactions [Temerin et al., 1990]. These phenomena involve a rich
variety of plasma processes which include electrostatic shocks [Mozer et al., 1977], double layers
[Temerin et al., 1982], field-aligned electrons [McFadden et al., 1986], edge precipitation,
Langmuir and whistler wave emissions [Gurnett et al, 1969], auroral kilometric radiation (AKR)
[Gurnett, 1974], ion conics [Klumpar, 1986], ion beams, and the formation of the auroral density
cavity. The FAST mission was designed to have one to three orders of magnitude higher resolution
than previous auroral missions which have identified many of the auroral processes but were
unable to spatially or temporally resolve them. Secondary scientific goals include exploratory
research of high time resolution phenomena, correlative studies with ground-based radar and
optical observations, conjunctive studies with Polar and other NASA spacecraft, and studies of
low-altitude equatorial plasmas.  The FAST satellite was launched into an ~83° inclination
orbit with a 350 km perigee and 4175 km apogee in August,1996. The satellite is oriented in a
"cartwheel" attitude which has the spin axis nearly (negative) normal to the orbital plane. It is spin
stabilized with a spin period of 5 s. The satellite crosses the auroral zones (which form ovals at
~65°-70° magnetic latitude North and South) four times an orbit. The orbit was designed to have a

Northern apogee during January and February of 1997 for coordinated ground-based and optical
observations. In situ observations of auroral processes at high resolution require high data rates from
all instruments which would exceed telemetry capability. Because the auroral zones are only a
fraction of the FAST orbit, it is possible to acquire data at several times the maximum telemetry rate
if the data are stored on-board. Central to the FAST satellite is a large (~128 MByte) solid state
memory common to all instruments. The FAST data system is described elsewhere [Harvey et al.,
this issue] so we provide only a brief overview. The on-board memory accepts two primary types of
science data from all of the instruments: survey data which has continuous coverage (~40 minutes)
over the entire auroral zone and burst data which has only a few minutes of coverage at the highest
resolution. The position of the auroral zone, although typically at ~65°-70° magnetic latitude, varies
enough that on-board triggers are required for the burst data. Each instrument produces composite
signals called "trigger data" that can be used by the central processor to trigger the burst data
acquisition. Survey data has three sub-types: "full orbit" data which is typically for the fluxgate (dc)
magnetometer only, "slow survey" which has ~0.06 s resolution waveforms in the sub-auroral

and polar cap regions, and "fast survey" which typically has ~0.5 ms resolution. The basic idea is
as follows. The FAST satellite always acquires magnetic field data. As it approaches auroral
latitudes, the science instruments are put in "slow survey" by a time-tagged command. Once the
auroral electron precipitation is detected, the science instruments are configured into "fast survey".
If signals of scientific interest are seen by the burst triggers, burst data is collected for between ~5 s
to several minutes. Collection typically includes 25% of data prior to the trigger. There may be
several burst collections depending on the available memory. Burst electric and magnetic field
waveforms have ~30 ms resolution and the High-Speed Burst Memory waveforms (HSBM) have
0.5 ms resolution. The survey and burst data can be stored on board until telemetered to ground.

The FAST electric field and magnetic field instrument, herein called the "fields instrument",
includes the electric field and magnetic field sensors, the electric and magnetic field signal
processing, and a wave-particle correlator. The instrument was designed to take advantage of the
FAST satellite data system by providing full orbit coverage for the fluxgate magnetometer, a
variable rate survey data, and high-resolution burst data. This article will cover the electric field
sensors, and the electric and magnetic field signal processing system, and the wave-particle
correlator on the FAST satellite. The magnetic field sensors [Elphic et al., this issue] and the
booms will be describe in separate articles.

(Figure 1b) positions. There are ten spherical sensors for measuring the electric field. Eight sensors
are on four wire booms in the spin plane and two are on rigid booms that are deployed along the spin
axis. The vector dc and ac electric field signals are derived from the Voltage difference between
pairs of spherical probes which form dipoles of 5 m or 56 m in the spin plane [Mozer, 1973]. The spin
plane wire booms have two sensors each to make multi-point measurements which can be used to
determine the wave vector of coherent emissions with cross spectral analysis. The length of the spin
axis dipole is 7.7 m. Six of the ten sensors can operate as Langmuir probes, measuring the electron
current to a fixed potential probe in the plasma to determine the electron density. The dc magnetic
field (to ~100 Hz) is measured by a three-axis fluxgate magnetometer and the ac magnetic fields
(~10 Hz to 4 kHz on two axes, ~10 Hz to 500 kHz on one axis) are measured by three search coils.
The three-axis fluxgate magnetometer is on a 2 m boom opposite an identical boom which carries
the three-axis search coil. The satellite is oriented so that the magnetic field lies within 6° of the
spin plane in the auroral regions.

Figure 1. (a) A three-dimensional view of the electric and magnetic field sensors on the
FASTsatellite as designed. The electric field instrument has eight spherical sensors that
are on four spin plane wire booms (two each) and two that are on rigid axial booms. All of
the spherical sensors can operate in "Voltage Mode" (marked with "V") in which they
measure the local plasma potentialwith respect to the payload. The electric field signals are
measured by pairs of sensors which form 56 m, 7.7 m, or 5 m dipoles. Six of the ten sensors,
marked with "I" can operated in current mode where the electron current is measured
for deriving plasma density. The fluxgate and search coil magnetometers are on ~2 m
booms. The search coil assembly is rotated 21° out of the spin plane.(b) The deployed state
as of September, 1996. The wire boom carrying sensors 3 and 4 did not fully deploy.
 

Figure 2 shows a functional overview of the fields instrument. The signals from the spherical
sensors are passed through a Boom Electronics Board (BEB) located at the boom deployment
unit. The BEB drives bias currents and voltages of the electric field sensors (set by processor
command), controls the boom motors during deployment, and provides housekeeping signals to the
instrument processor. The fluxgate magnetometer sensor is driven from a card located inside the
main electronics box.

Figure 2. A block diagram of the FAST satellite Electric and Magnetic field instrument.
 

The sixteen electric and magnetic field sensors share a common signal processing system that is
located in the central instrument electronics box. The signals are fed to two analog conditioning
circuits. The low-frequency analog conditioning covers the frequency band of dc to ~16 kHz and
the high-frequency conditioning typically covers the frequency range from ~3 kHz to 2 MHz
(4 MHz maximum). The analog conditioning circuits have differential amplifiers to form the electric
field signals from pairs of sensors and filters to isolate configured pass bands. Analog switches and
analog multiplexors allow the instrument to operate in a variety of configurations. Low-frequency
analog signals are digitized with nine, 16-bit analog to digital (A/D) converters. Survey data

(continuous coverage over the auroral zone) are eighteen electric and magnetic field waveforms
from one A/D converter. The survey waveforms have sample rates of 2,048 samples/s (ten signals
at ~1 kHz Nyquist) or 512 samples/s (eight signals at ~250 Hz Nyquist). Six of the remaining eight
A/D converters sample six signals at 32,768 samples/s (~16 kHz Nyquist). The remaining two
A/D converters operate in one of two different modes. They can sample one signal each at 32,768
samples/s or four signals each at 8,196 samples/s (~4 kHz Nyquist). The signals are labeled by the
band width: 16 kHz, 4 kHz, 1 kHz, or 250 Hz. The 4 kHz and 16 kHz signals are burst wave forms,
typically recorded for several minutes during the most intense part of an auroral pass.

Telemetry limitations are such that 4 kHz and 16 kHz signals cannot have continuous coverage.
Continuous coverage of the power spectral density of the 16 kHz signals is computed on-board
by a Digital Signal Processor (DSP) which averages several Fast Fourier Transforms (FFT).
The 1024 point FFT covers the frequency range from dc to ~16 kHz with 32 Hz bandwidth and
~100 dB dynamic range. The DSP can also perform a cross-spectral analysis to determine the
phase difference and coherency of pairs of signals. The spectra produced by the DSP are treated as
survey data, giving continuous coverage in the auroral zone. High-frequency signals are processed
three ways (Figure 2). A Swept Frequency Analyzer (SFA) produces power-frequency-time
spectra typically from ~10 kHz to 2 MHz (the sweep range is adjustable) with ~80 dB dynamic
range and 15 kHz band width. The typical gain setting in the auroral zone has an electric field
dynamic range of 10^(-15) to 10^(-7) (V/m)^2/Hz and a magnetic field dynamic range of 10^(-12)
to 10^(-4) nT^2/Hz (at 100 kHz). The time resolution is typically 62.5 ms (31.25 ms is the fastest
resolution). The SFA spectra are logarithmically amplified (in analog) then A/D converted at 8 bits.
The 62.5 ms spectra are treated as burst data. Averaged spectra (125 ms to 4 s resolution) form
survey data. The SFA unit also contains a Plasma Wave Tracker (PWT) which gives fine frequency
(~50 Hz) resolution and time resolution (31.25 ms) coverage over a narrow frequency range (16 kHz)
that lies between 0 and 2 MHz.
The High-Speed Burst Memory (HSBM) digitizes four high-frequency signals (~3 kHz - 1 MHz) at
2x10^6 samples/s with 10-bit resolution. The data are stored in 2.5 Mbyte buffers that cover
~0.25 s periods.  Dataintervals are selected by dedicated triggers which monitor wave power in both
the high-frequency (200 kHz - 2 MHz) and low-frequency (~1 kHz - 16 kHz) bands. The HSBM
has very limited time coverage (<1% duty cycle) due to the high conversion rates.
The Broad-Band Filters (BBF) rectify four electric and/or magnetic field signals
to determine the amplitude envelope of 200 kHz - 2 MHz wave emissions versus time. They have
a dynamic range of ~60 dB. For the same four signals, the number of zero crossings are counted,
which represents the wave frequency of narrow band emissions. The phase shift between each pair
of the four selected high-frequency signals (six phase difference signals) are also measured on
board. These phase differences can be used to determine the high-frequency wave polarization.
The high-frequency wave amplitudes, zero crossing rates, and polarizations are telemetered as

survey data. The Wave-Particle Correlator (WPC) uses two electric and/or magnetic field signals
and twelve of the stepped electron electrostatic analyzer anode signals [Carlson et al., this issue] to
measure oscillations in electron fluxes in one of two selected frequency ranges. The low frequency
range is from ~500 Hz to ~16 kHz and the high frequency range is from ~200 kHz to ~2 MHz.
The on-board correlation function is computed in digital circuitry.

Figure 3. The spectral coverage of FAST satellite Electric and Magnetic signal processing.
 

Burst data (Figure 3) supplement the survey data with 16 kHz Nyquist frequency waveforms, high
time resolution SFA spectra (62.5 ms), PWT (Plasma Wave Tracker), and WPC (Wave-Particle
Correlator) data. In addition, the HSBM (High-Speed Burst Memory) records four channels of
waveforms, typically three E and one B, with a frequency range from ~1 kHz to 1 MHz. The
HSBM has a reduced coverage due to the very high data rate.
 

There are ten spherical sensors for measuring the electric field, two on each of four radial wire
booms and one on each of two axial booms, as shown in Figure 1. The 8 cm diameter spherical
probes contain electronics that operate in one of two selectable modes. All of the sensors can
operate in "Voltage mode" in which they measure the potential of the nearby plasma with respect to
the spacecraft. In Voltage mode the probes are biased with a fixed current. Six of the sensors
(Figure 1) can operate in "Current Mode" as Langmuir probes which measure plasma current. In
current mode, the probe is biased at a fixed potential. The radial booms are 2.5 mm diameter wires
which support the sensors and carry power and signals between the sensors and the spacecraft. The
centrifugal force is supported by a kevlar braid that surrounds two coaxial cables and eight insulated
wires. The kevlar braid is covered with aluminized kapton and a silver coated copper wire braid
which is exposed to the plasma. The exposed conductor is segmented into several sections that have
controlled potentials. Figure 4 is a detailed diagram of the wire boom sensors.

Figure 4, Top: The physical layout of the radial boom spheres, stubs, and guards. There are
two spheres on each wire separated by 5 m. The 8 cm diameter spheres house a
preamplifier. The outer sphere operates in Voltage mode only. The inner sphere (nearer the
spacecraft) operates in Voltage or Current mode. Adjacent to each sphere are 2.4 m stub
sections which are biased at a selectable Voltage with respect to the sphere potential.
Three 10 cm long guard sections are also shown. Bottom Left: The axial boom physical
layout. There is one sphere on each 2 m rigid boom. Each sphere has a 20 cm stub on the
inside and a 10 cm stub on the outside. The axial spheres can operate in Voltage or Current
mode. Bottom Right: The electrical/mechanical sphere assembly. Inside of each sphere are
one or two circuit boards for Voltage mode and/or Current mode operation. Surrounding the
circuitry and the wires is a shield which penetrates ~0.2 cm beyond the sphere. The shield
is driven at the sphere potential over the full bandwidth, thus reducing any unwanted
capacitance between the sphere surface and the electronics to less that 1 pF. The shield
also reduces cross talk between spheres to a negligible level.
 

The electric field is derived from the difference in potential between two probes in Voltage mode.
The probe potential is determined by a balance of electron current, ion current, photo-electron
emission, secondary electron emission, and a bias current. The probe surfaces are coated with
carbon that produces known photo-emission characteristics. Bias currents are adjusted from
-100nA to 100 nA in steps of ~0.8 nA to minimize errors. Bias tables for low and high plasma
densities,eclipse and sun, and spacecraft configuration are stored on-board and are
automatically adjusted as the spacecraft passes through the terminator.

The "stubs" are 2.4 meter sections of wire exposed to the plasma immediately beside the sphere.
A 2.4 meter stub section is added to the outside of the outer sphere for symmetry. The outer
conductive surface of the stubs are driven at a fixed potential with respect to the sphere that is
adjustable from -2.5 V to 2.5 V. By holding the stub potentials fixed with respect to the nearby sphere,
the photo emission current between the sphere and the nearby wires can be controlled so that photo
emission modulations on the wire minimally effect the potential of the sphere. The voltage control of
the stubs is resistive to maintain electrical stability.

Three "guards" are 10 cm sections immediately beside the stubs. The potentials of the guards are
adjustable from -10 V to 0 V with respect to the outer most probe potential. The guards are typically
biased at -5 V to restrict photo-emission current between the spacecraft and the spherical probes
and, between the two probes. Optimized stub and guard biases are also stored in the on-board bias
tables.

The spin axis booms are rigid, 2 m stacers that have a single sphere with short (20 cm and 10 cm)
stubs and no guards. The preamplifier design of the axial spheres is identical to the inner radial
spheres. The axial probes can operate in Voltage or current mode.

Plasma resistance to the spherical probes is expected to vary from ~10^6 omega to >10^9 omega
and the capacitive coupling to the plasma is typically ~5 pF. Preamplifiers located inside the
spherical probes are designed to have very low stray capacitance (<1.0 pF) and very high input
resistance (>10^11omega) so that electric fields from DC to ~2 MHz can be measured. An
aluminum shield (Figure 4) electrically shields the probe surface from the circuitry inside the probe
and the wire which runs through the probe. The same cover also provides radiation shielding.
Radial and axial preamplifier response is diagramed in Figure 5a and the signal processing coverage
of the electric field is diagramed in Figure 5b. Signals less than ~300 Hz are typically resistively
coupled to the plasma while those greater than ~300 Hz are typically capacitively coupled to the
plasma. The drop in preamplifier gain to 0.8 at frequencies greater than 300 Hz is due to the cross
over from resistive to capacitive coupling.

Figure 5. (a) The sensor preamplifier response as a function of frequency under 100 MW,
5 pF source impedance. The drop of gain from 1.0 to 0.8 at ~300 Hz represents the cross
over from resistive to capacitive coupling. The response on radial booms has an effective
a two-pole roll off at ~500 kHz due to losses driving the 28 m cable. The axial boom (2m
cable) preamplifier response has twice the bandwidth. (b) The theoretical noise level of the
FAST 56 m electric field antenna system and the dynamic range of survey waveform, DSP,
SFA, and HSBM signal processing systems.
 

In current mode, sensors 6, 7, 9, and 10 have a dynamic range from ~0.5 nA to 2x10^4 nA
representing a density range from ~0.2 cm^(-3) to ~10^4 cm^(-3) with typical auroral electron
temperatures (~1 eV). Spheres 2 and 3 measure from ~10 nA to 5x105 nA for low-altitude
coverage. The bias Voltages on sensors 2, 3, 6, and 7 can be set between 0 V and 20 V with respect
to one of two base potentials, the nearby Voltage mode sensor or the potential derived from 0.8 the
nearby Voltage sensor and 0.2 the spacecraft potential. Initial testing on orbit has shown the first base
potential can cause undesirable spacecraft charging so the latter base potential is always used.
Sensors 9 and 10 are always referenced from the payload and can be biased from -5 V to 45 V.
 

A Boom Electronics Board (BEB) is located in each of the four wire boom deployment units and one
is located on the radiation shield of the main spacecraft to control the axial booms. The five boom
electronics boards perform the following functions: (1) Receive commands from the IDPU
(Instrument Digital Processing Unit) including bias levels for the sensors, stubs, and guards, (2)
provide analog bias Voltages and currents to the sensors, (3) provide power to the sensor
preamplifiers, (4) distribute the analog signals from the sensors to the fields signal processing system,
(5) turn on and off wire boom motors from IDPU command, and (6) return housekeeping signals of
bias levels, temperatures, and the state of deployment of the wire booms. A block diagram of the
radial BEB is in Figure 6.

Commands from the IDPU are sent via a serial digital interface to a programmable gate array
(ACTEL) which contains all of the digital logic on the BEB. Each BEB has a unique address so they
can be commanded individually. Digital commands include the bias, guard, and stub levels which are
fed to an 8-bit digital to analog converter (DAC) which are referenced to the dc level (low pass at ~1
kHz) of the sensor (the guard is referenced to the outer sensor). The command interface also
receives commands to set the operating mode of the inner sphere preamplifier (voltage mode or
current mode), turn on and off the deployment motors, and to select the housekeeping signals.

 Figure 6: A block diagram of the radial boom electronics board.
 

The floating power converter provides power for the BEB and the sensor preamplifers. The sensor
power supply has a reference to the sensor potential with a reference range from -52 V to 52 V.
The dc to ~300 Hz dynamic range of the sensors is from ~-45 V to 45 V which can measure ±1.6 V/m
on the 56 m dipoles and +11 V/m on the 8 m axial dipole. The 5 m dipoles saturate at the same electric
field amplitudes as the 56 m dipoles, except for short wavelength emissions. The sensors have a
operating range of ~-8 V to 8 V for signals >~300 Hz. The 56 m and 5 m dipoles can measure up to
300 mV/m wave electric fields and the 7.7 m axial dipole can measure up to 2 V/m wave electric fields.

The FAST instruments were defined under the concept of science "modes", whereby the
instruments can be configured to emphasize specific scientific investigations. Science modes also
allow for optimizing investigations at various local times and altitudes as well as performing follow
up investigations of any phenomena discovered during the mission. For example, a science mode
that emphasizes auroral kilometric radiation studies would emphasize the high-frequency signal
processing. Another use of the modes is to control the overall data rate. Southern hemisphere
auroral zone crossings typically do not have ground contact (the data must be stored on board)
whereas northern crossings typically do. Science modes with lower data rates are used in the
southern auroral zone.

The fields instrument was designed with several layers of flexibility to accommodate a variety of
science modes. The first layer of flexibility is in the sensor configuration. Six of the ten electric field
sensors can be operated in one of two configurations. The ten sphere signals and six magnetic field
signals (three dc signals from the fluxgate and three ac signals from the search coil) are the primary
inputs to the fields signal processing unit. The sphere signals are passed to a series of differential
amplifiers that form long baseline (56 m) and short baseline (7.7 m or 5m) electric field signals that
are conditioned to dc coupled low-gain (16 kHz bandwidth) and ac coupled high-gain (typically 3
kHz to 2 MHz bandwidth) signals. A series of multiplexors allows selection of raw sensor signals,
differential (electric field) signals, search coil magnetometer signals, and the fluxgate magnetometer
signals to several signal processing systems. Finally, each of the signal processing systems has a
variable data rate. The configuration of the sensors, the signal selection, and the individual data
rates of each signal processing system are the three main elements of a science mode.

V. Signal Processing and Data Products

The signal processing descriptions are organized by the data products they produce following the
overview diagram in Figure 2: (1) survey waveforms, (2) burst waveforms, (3) survey VLF
spectra, (4) survey HF spectra, (5) burst HF spectra, (6) HF waveforms, and (7) survey HF power
and frequency. The wave-particle correlator is described in a separate section. Several signal
processing systems may share common resources. For example, the low-frequency analog
circuitry conditions signals for both the survey waveform and burst waveform data. The digital
signal processor uses burst waveform A/D converters to produce the survey spectra.

survey signals are summarized in Table 1. The signals that are most often selected in the science
modes are high-lighted.

Five of the survey waveforms are filtered to a DC to 250 Hz band. The five signals are three sensor
Voltages (to monitor spacecraft potential) and two signals representing the VLF (~1 kHz - 16 kHz)
wave power. The VLF signals are rectified wave signals that have been logarithmically amplified.

All of the survey waveform signals share a single 16-bit A/D converter (Crystal Semiconductor
CS5016) that constantly samples at 32768 samples/s. The survey waveform output rate is
configurable by factors of two from 16 samples/s to 2048 samples/s for the 1 kHz waveforms, and
from 4 samples/s to 512 samples/s for 250 Hz waveforms. The signals are under-sampled except
at the maximum rate.

The three-axis fluxgate magnetometer signals are converted at rates from 16 samples/s to 2048
samples/s before being input into a recursive digital filter that is in a programmable gate array. The
recursive filter acts as a one-pole low-pass filter at , which is ~21 Hz at
maximum sample rate. The output rate of the fluxgate magnetometer signals are at 1/4 the sample
rate varying from 4 samples/s at the minimum rate to 512 samples/s at the maximum rate. Fluxgate
magnetometer data are plotted in panel (c) of Figure 15.

In typical operation, the fluxgate magnetometer signals and the 250 Hz signals are output at 8
samples/s for full orbit and slow survey coverage. The 1 kHz waveforms are output at 32
samples/s in slow survey. In fast survey, the data rates are at maximum if the spacecraft is in
contact with a ground station, typical of Northern auroral passes. Otherwise, the fast survey sample
rates are typically 1/4 or 1/16 of the maximum rate.

Burst Waveforms

Burst waveforms are at significantly higher sample rates of 32,768 samples/s (16 kHz Nyquist) or
8,196 samples/s (4 kHz Nyquist). Up to 14 signals out of 40 signals (Table 2) can be selected for
A/D conversion. Sixteen are electric field signals from long baseline and short baseline dipoles that
are dc or high-gain, ac-coupled signals. Three search coil magnetometer signals, ten spheres
sensor signals (which can include Langmuir probe currents), six Langmuir probe signals, four BBF
signals, and the analog signal from the Plasma Wave Tracker are also fed into a bank of analog
switches which are configured by the science mode.

 
There are eight 16-bit (Crystal Semiconductor CS5016) A/D converters dedicated to burst

Table 2 summarizes the available signals and their properties. The entries ending with HG are ac
coupled, high-gain signals with a frequency band from ~3kHz to 16 kHz. The high-pass filter at 3
kHz has one pole. Otherwise, the signals are DC coupled. The sensitivity in Table 2 is the one-bit
level, while the range is limited either by the maximum range of the sensor or the A/D converter.
A/D converters 7 and 8 can be multiplexed to receive four 4 kHz bandwidth signals. If the signal is
available at 4 kHz, an "m" is added (e.g. 8m). Otherwise the signal is available at 16 kHz. The
highlighted signals are most often configured. VQUAD is the measurement of the quadrupole
signal (V1+V4-V5-V8). PWT is the plasma wave tracker output described later. The BBF
channels are the 200 kHz - 2 MHz wave amplitude, also described later.

The burst digital data ar continuously available to the Instrument Data Processing Unit and to the

Digital Signal Processor at an overall data rate of 4.194 Mbit/s which far exceeds the telemetry
capability so the data are recorded for several minute intervals. The periods of data capture are
selected by the instrument data processor. The selection criteria are from trigger signals supplied
by the fields instrument and the particles instrument.

The primary function of the Digital Signal Processor (DSP) is to provide continuous coverage of
the spectral power density of the electric and magnetic field in the frequency range from DC to 16
kHz. It also has three other functions: provide spectral power density of the high-frequency (~1
kHz to 1 MHz) electric and magnetic field, perform cross spectral analysis of the low-frequency
and high-frequency electric field pairs, and calculate the auto-correlation function of four fixed
energy channels from the Electron Spectrograph.

The low-frequency spectrum analysis averages 2^n (n is configurable from 0 to 7), 1024-point Fast
Fourier Transforms (FFT) of the digital waveforms from the burst A/D converters. The resulting
spectra have 32 Hz resolution in frequency and from 32 ms to 4 s time resolution. The selection,
sensitivity, and range of the burst waveforms are described in Table 2.
 A block diagram of the DSP is in Figure 7. The eight burst waveforms are fed into a dedicated

Direct Memory Access (DMA) designed into an programmable gate array (ACTEL). A second
custom DMA was designed to access the High Speed Burst Memory waveforms (1 MHz
waveforms at 2 Msamples/s) and four channels of the electron electrostatic analyzer. All of the
data passes from the DMAs into the RAM through single high-speed serial interface in the DSP. In
all, there are sixteen inputs which are individually selectable. Typical selection are the first six
low-frequency waveforms.

A third programmable gate array acts as a controller. Its primary functions are (1) to direct start up,
reprogramming, and resets of the DSP processor, (2) to receive commands from the IDPU
processor, (3) to provide timing and control of the DMA PGAs, (4) to detect single event latch up,
single event upset, or malfunction of the DSP and restart, and (5) to protect the code area in RAM.

The DSP processor is a 32-bit floating point ATT-DSP32C with an input clock at 32 MHz (below
the 50 MHz maximum). Radiation testing was performed at the University of California at
Berkeley to determine that the total dose tolerance exceeds 200 kRad. Four radiation hardened
static CMOS RAMs were used to make a 32-bit by 32 K memory. Bi-polar ROMs are used to hold
the code.

Due to the high power consumption (2.5 W) and slow access speed (100 ns) of the ROMs, the
start-up sequence copies the ROM code into the RAM code area then turns off the ROMS.

Figure 7: A block diagram of the DSP.                        Figure 8: The dynamic range coverage
The DSP receives data from the burst                        versus frequency for DSP spectra. DSP
A/D converters (low-frequency), the                        spectra have continuous coverage in the
HSBM (high-frequency), and the                              auroral zone.  The DSP averages FFTs of
electron electrostatic analyzers. The                         electric and magnetic waveforms.  The
primary function of the DSP is to provide                   time resolution of the 1024 point FFTs
continuous 16 Hz - 16 kHz wave power                   varies from 31 ms to 4s.
spectra in the auroral zone.

The code can be augmented or patched by command from the IDPU processor. The code area
is hardware protected after the start-up sequence is completed and before the science mode
configuration is sent to the DSP.

The controller can sense a latch up or upset in two ways: (1) if the current to the DSP exceeds a
preset value or (2) if the DSP fails to send the controller a coded command within a specified
period. If an upset occurs, the controller can "warm start" the DSP processor without having to
turn on the ROM or receive commands from the IDPU. No latch up or upset events have occurred
in the first eleven months of operation during which there have been several solar events.

The dynamic range of the DSP VLF spectra are displayed in Figure 8. The data are logarithmically
compressed to 8 bits. The electric field spectra range from 2x10^(-13) (V/m)^2/Hz to 2x10^(-3)
(V/m)^2/Hz on the 56 m high gain (V5-V8HG) signal. The 5m dc coupled signals have a dynamic
range starting at 10^(-10) (V/m)^2/Hz. The range of the signal processing is optimized for auroral
processes which are well above the sensor noise levels. The magnetic spectra of the 21" search coil
range from ~1x10^(-10) (nT)^2/Hz to ~5x10^(-3) (nT)^2/Hz at 1kHz. The search coil dynamic
range is limited by the sensor noise at frequencies greater than ~500 Hz. Auroral electric and
magnetic spectra from the DSP are displayed in panel (f) of Figure 15.

The primary functions of the swept frequency analyzer (SFA) are (1) to provide continuous
(survey) coverage of the high-frequency (~10 kHz - 2 MHz) electric and magnetic fields, (2)
provide high time resolution (62.5 ms), high-frequency spectra of the electric and magnetic fields,
and (3) provide fine frequency resolution observations of narrow-band emissions such as auroral
kilometric radiation. The third function is designated as the Plasma Wave Tracker, which is
described in the next section.

A block diagram of one of the SFA channels is given in Figure 9. Table 3 describes selection and
the sensitivities of the signals from the high-frequency analog (Figure 2). The SFA has four
channels. In three of the channels, the high-frequency signal is passed into a 3-pole low-pass filter
at 2 MHz. The fourth channel, which can be used by the plasma wave tracker, has a 4 MHz filter.
The filtered signal is mixed with a sweeping reference, passed through a crystal filter with 15 kHz
band width, then mixed again to 50 kHz intermediate frequency. The resulting signal is rectified
and logarithmically amplified before digital (8-bit) conversion by the high-frequency A/D
converter (see Figure 2). The 256-point HF spectra have ~8 kHz steps (over-sampling the 15 kHz
band width) covering from 0 to 2 MHz. Since the sweeping reference is produce by a digital
frequency synthesizer, the sweep can be configured to have maximum range of 500 Hz, 1 MHz, 2
MHz, or 4 MHz. The sweep rate can be configured to 31.25 ms or ~62.5 ms (typical operation).

The survey data, called "SfaAve", averages 2^n sweeps, where n is set from 0 to 7. The sweep rate
changes from "slow survey" to "fast survey", typically from n=6 to n=3, thus increasing the data
rate and time resolution of the sweeps (typically 4 s in slow survey and 0.5s in fast survey). Burst
data transmits every sweep, typically 16 per second. The spectral coverage of the SFA is displayed
in Figure 10 SFA data are displayed in panel (e) of Figure 15.

The primary function of the plasma wave tracker is to provide fine frequency resolution spectra of
narrow-band emissions. This is accomplished using the fourth channel of the SFA. The signal from
the digital frequency synthesizer is set at a fixed frequency (fo) that is (1) fixed by configuration,
(2) dynamically set at the electron cyclotron frequency by the IDPU processor, or (3) dynamically
set by the number of zero crossings in the wave form. The fixed frequency is the center of the 15
kHz high-resolution band.

After passing through the 10.7 MHz crystal filter (Figure 9), the signal passes to the second mixer
which is set at 10.692 MHz when configured for the plasma wave tracker. The resulting signal
(~0.5 kHz to 15.5 kHz) represents the frequency band fo-7.5 kHz to fo +7.5 kHz. The PWT signal
then is digitized as a burst waveform (see Table 2, entry 20) and/or processed by the DSP. The
frequency resolution of the PWT signal is limited by the jitter in the digital frequency synthesizer
to ±70 Hz. An example of PWT data is given in panel (d) of Figure 15.

The high-speed burst memory (HSBM) produces 0.5 micros resolution digital waveforms of four
signals, typically three electric field and one magnetic field, with ~0.1% coverage in the auroral
zone (typically ~2 seconds out of 30 minutes). Alternatively, the HSBM can supply the DSP
(described above) with digital HF waveforms to be converted into HF spectra by averaging FFTs.
The HSBM is located in the fields instrument signal processing and is not part of the main
instrument burst memory. A block diagram is in Figure 11. Spectral coverage and dynamic range is
displayed in Table 3.

The HSBM has three main sub-systems. The analog section has selectable three-pole low-pass
filters of 125 kHz, 500 kHz, and 1 MHz corresponding to three selectable A/D speeds of 250
ksample/s, 1 Msample/s, or 2 Msample/s. Four 10-bit MP7694 A/D converters are augmented
with a high-speed sample and hold. The four A/D converters are on a 40-bit bus that is
continuously written to a 10 Mbyte RAM. The trigger system uses the rectified and logarithmically
amplified low-frequency wave power (see survey waveforms), the high-frequency wave power
(see later), and the plasma density (see survey waveforms) as inputs. The trigger A/D converter
alternately samples two of the selected inputs labeled "trigger A" and "trigger B", typically a
low-frequency and a high-frequency power level.

The digital logic is contained in three programmable gate arrays (ACTEL) and works as follows.
The RAM is divided into four buffers, one of which continuously accepts data (input buffer), two
are holding the buffers with the highest trigger A and trigger B levels (holding buffers), and the
forth buffer outputs data to the selected port (output buffer). If the trigger level exceeds the level
in one of the holding buffers, the instrument (1) waits for 3/4 of the input buffer to be written
keeping 1/4 of the buffer for data prior to the trigger event and (2) swaps the input buffer and a
holding buffer thus overwriting the holding buffer. This process continues until the output buffer
has been completely read. At that point, one of the holding buffers, alternating between the two,
becomes the output buffer. The new hold buffer is assigned a zero trigger level.

The logical process above has several additional features. One can assign a minimum trigger level
to the HSBM so that a trigger event cannot occur unless the minimum requirement is met. There
are two settings of trigger position with in the buffer at either 1/4 or 1/2 of the buffer for data
prior to the trigger. The triggers can be set "retriggerable" or "absolute". The former allows for a

Figure 11. A block diagram of the                          Figure12. A block diagram of the broad
high-speed burst memory. Four                             band filter, phase, and zero crossing processing.
high-frequency channels are low-pass              Four high-frequency channels are high-pass
filtered then digitized. The data are                        filtered then rectified to measure the high-
selected by high- or low-frequency                     frequency wave amplitude at high time resolution.
wave event or plasma density cavity                    The high-frequency signals are digitized by a
by a dedicated trigger system and are                  comparator.
stored in a 10 Mbyte RAM. The RAM
buffers can be output as waveform data
(Port 1) or transferred to the DSP which
produces a average of FFTs.
trigger event to restart if the trigger level increases later in the event so that the peak is always at
the trigger position. The latter, does not allow for a restart. The maximum buffer size is 2.5
Mbytes, allowing for 538,068 10-bit samples in four channels. The buffer sizes can be 1/2n of the
maximum with n between 0 and 7. Finally, on can go into "time-based" triggering whereby the
triggers events occur by time rather than by plasma wave amplitude. The last feature is typically
used when the DSP generating spectra.

The main function of the Broad-Band Filter (BBF) signal processing is to provide high-time
resolution power, frequency, and phase information for high-frequency signals (Figure 12). Four
ac coupled, high-frequency signals are processed. The selection, sensitivities and range are
described in Table 3. The selected signals are high-pass filtered at 200 kHz with a three-pole
passive LCR filter. Since the sensor preamplifier response rolls off at ~500 kHz, the BBF
emphasizes auroral kilometric radiation which falls in the 200 kHz to 500 kHz band. The four
signals are rectified and pseudo-logarithmically amplified and then passed to the high-frequency
A/D converter (Figure 2) at typically 1 ms resolution. The BBF has a 60 dB dynamic range.

The selected analog signals are also fed to a comparator producing a digital representation of the
zero crossing. The digital signals are input to counters which yield the frequency of a dominating,
narrow band signal. Six pairs of the digital signals are processed by a phase discriminator which
measures the relative phase of the signals to determine wave polarization. The zero-crossing
counters and the phase shifts have 4 ms time resolution.

The main scientific objective of the wave-particle correlator instrument is to identify the energy
and pitch angle of particles that are interacting with waves. Measurement of the amplitude of
particle oscillations and their phase relation with the wave allows for the evolution of the wave
and distribution function to be studied in detail. Wave-particle correlator instruments essentially
provide a direct observation of wave-particle interactions.

The FAST wave-particle correlator detects oscillations in particle flux by integrating the product
of the wave electric field and particle flux [Ergun et al., 1991a; Lin et al., 1995]. This technique
can be used to determine both the amplitude and phase (with respect to the wave phase) of
oscillatory currents. "Resistive" currents are in phase with the electric field and determine energy
flow between the wave and particles [Ergun et al., 1991b], whereas "reactive" currents are in phase
with wave potential and can indicate nonlinear kinetic processes such as particle trapping
[Muschietti et al., 1994].

The goal of the direct wave-particle correlator is to measure the oscillatory perturbation in the
electron distribution function (f1) that is at the ambient wave frequency. If the electric field wave
form is known, the oscillatory perturbation in the particle distribution can be derived from the
correlation function: 

where tint is the integration period. The numerator of Equation (1) contains two terms. Term B
gives us the resistive perturbation whereas term A should have no contribution since E oscillates
and fo is constant. However, since |f1|<<<fo, term A can be a source of error (called the dc bias
error) if  .   The wave-particle correlator design minimized the dc bias error.

One channel of the wave-particle correlator is diagramed in Figure 13. The electric field signal

Figure 13. A simplified block diagram of  a              Figure 14. A block diagram of the FAST
one channel wave-particle correlator.                     satellite wave-particle correlator.  Each
The electric field signal processing is                       electron channel produces a total count
diagrammed on top. Comparators produce              and two phased counts. The Duty cycle
two digital signals that represent polarities            monitor measures the dc bias error
of the wave signal that has been split into
two phases that differ by 90o. The wave
polarity signals are correlated with the
particle pulses (bottom) with digital logic.
The f0Count (or f1Count) and is incremented
only if the polarity of the f0 (or f1) wave
signal is positive. The correlation
measurements are (2f0Count - Total Count)
and (2f1Count - Total Count).

from the antenna is filtered to the desired frequency band. The instrument has selectable pass
bands of 200 kHz to 2 MHz for Langmuir wave and auroral kilometric radiation studies, and
500 Hz to 20 kHz for lower frequency wave studies. The filtered electric field signal is fed into
an analog phase splitter which consists of two all-pass filters that have phase responses which
differ by 90° over a broad frequency range. (An all-pass filter is a unity gain circuit that has a
non-zero phase response.) The phase response of the FAST satellite high-frequency phase splitters
are diagrammed in Figure 14. The high-frequency phase splitter has a phase difference of
90°±10° for frequencies between 100 kHz and 2 MHz. The usable frequency range (phase
difference of 90°±45°) is from 40 kHz to 5 MHz. The low-frequency phase splitters (not shown)
have a usable frequency range from 300 Hz to 40 kHz. The all pass filters have near unity (±10%)
gain over the usable frequency range.

The digital output of the comparator (Figure 13) represents the polarity of the wave signal. The
comparator has positive feedback (not shown) to create a small amount of hysteresis which
determines the sensitivity of the instrument. The sensitivity is generally not critical since one
expects significant results only when large-amplitude waves are present. The FAST satellite
instrument is sensitive to waves with amplitudes between ~0.5 mV/m and ~ 1.6 V/m.

It is crucial that the digital wave polarity signals have a 50% duty cycle over an integration period
(typically several ms), that is, the integrated time that the polarity signal is in a high state (positive
polarity) must equal to the integrated time that the polarity signal is in a low state (negative
polarity), otherwise there will be a dc bias error (Equation 1, term A). Dc bias errors can be very
large when wave signals are strongly modulated. In the brief moments that the signal has too small
an amplitude to overcome the hysteresis of the comparator, the comparator rests at positive or
negative phase, creating a dc bias error. The baseline restorer circuit (Figure 13) is designed to
reduce the dc bias error. It causes the comparator to oscillate if no appreciable signal is present.
The FAST satellite wave-particle correlator baseline restorer maintains a duty cycle of
(50.0 ± 0.2)% for high-frequency wave signals over the integration period (~3 ms). Duty cycle
monitors measure the duty cycle which allows for the error to be further reduced in post flight
analysis.

The correlation function (Equation 1) is approximated by integrating the digital wave polarity
signals with particle events. The circuit is outlined in Figure 13 (lower). The wave polarity signals
and particle events are first synchronized to a fast oscillator (fosc > fwave, fparticle). The
synchronization allows the signals to be integrated with a simple gate and allows for the duty
cycles of the digital polarity signals to be precisely measured. The frequency of the oscillator
should be several times the maximum expected wave frequency and at least two times the
maximum particle count rate. Phase jitter (ø) reduces the correlation measurement by ~ cos(ø/2).

The wave-particle correlator determines  ,an approximation of Equation 3, where
F is the particle flux. One way to carry out the integration would be to increment a counter if a
particle event occurs during positive wave polarity, and decrement a counter if a particle event
occurs during negative wave polarity to determine Cpos-Cneg (counts during positive polarity -
counts during negative polarity). The technique we use is to count particle events that occur during
positive polarity (Cpos) and the total number of events (T = Cpos+Cneg), then determine the
correlation from the relation: Cpos-Cneg = 2Cpos - T. This method requires the minimum
amount of digital circuitry.

The particle events from the particle detectors increment three counters (Figure 13) which are
designated the f0 Count (Cø0), the ø1 Count (Cø1), and the Total Count (T). The ø0 (or ø1) counter
is incremented only if the polarity of the ø0 (or ø1) wave signal is positive. The three counters are
read out and zeroed at a selectable time interval called the integration time. The oscillatory
perturbation to the distribution function can be derived, to first order, directly from these values:

The ratio () enters in because the electric field E is represented by sgn(E). The data are
compressed before transmission to the ground. The transmitted quantities are (2Cø0 - T) and
(2Cø1 - T) that are compressed (pseudo-logarithmically) to 8 bit numbers (7 bits plus sign) and T
is compressed (square root) to 8 bits. The wave-particle correlator has the added feature that data
are not transmitted unless the wave amplitude exceeds a minimum threshold. The latter feature can
result in several orders of magnitude of data compression, depending upon wave activity.

Figure 14 is a block diagram of the FAST satellite wave-particle correlator instrument. The electric
field signal from one of two antennae are selected (depending on spin phase) to give the mostly
parallel (to the magnetic field) signal or mostly the perpendicular signal, depending on the
configuration. The instrument can also be configured to operate at high frequency (200 kHz -
2 MHz) or low frequency (500 Hz - 50 kHz). The selected wave polarity signals are correlated
with twelve fixed energy channels of the FAST EESA instrument (Electron Electrostatic Analyzer,
see Carlson et al., this volume). The look directions of the electron channels are configurable. The
channels that are closest to the desired pitch angle are automatically selected (de-spun on board).

The baseline restorer (Figure 13) was designed to reduce the dc bias error in the wave polarity
signal, but a small bias error may still exist. A significant bias error in a low-frequency signal
appears if there is a non-integer number of wave periods in the integration period. There is also a
frequency dependent error from asymmetric response in the digital circuitry (the time it takes to
switch from negative to positive may not be the same as the reverse). Many high speed CMOS
gates and programmable gate arrays have such a property with up to 20 ns differences which can
produces a false correlation of 4% (f1/f0) in a 1 MHz wave. Fortunately, these errors are easily
removed.

Since the digital wave polarity signals are synchronized to an oscillator, the exact duty cycle of the
waves can be measured (Figure 14). The duty cycles (Deltaø0 and Deltaøf1) are measured by
counting the number of (synchronizing oscillator) cycles that the wave polarity signal spends
positive, and subtracting the number of cycles that the wave polarity is negative. This is done by
counting the positive cycles and the total cycles (Figure 14) then subtracting:

Table 4 summarizes the mass and power for the FAST electric and magnetic field instrument. The
radial boom sensors include the deployment mechanism, two sensors, the wire booms, and support
structure for the radiation shield. The axial boom and sensor mass does not include the transmitter
support tubes. The majority of the signal processing and electric field sensors are nominally on
for ~25% of the orbit reflected in the low orbit-averaged power consumption.