The full energy per charge range (~15eV/e-32 keV/e) is spanned in 28 approximately logarithmically spaced discrete steps in the ROM table. Below ~2 keV/e the ratio of successive energy steps is approximately 1.33 and the energy passbands are contiguous or slightly overlapping due to the preacceleration of the ions. Above 2 keV/e the ratio of successive energy steps is approximately 1.24 and the gaps between adjacent energy passbands progressively increase due to the decreasing effect of the limited preacceleration voltage. At 32 keV/e, where the preacceleration is reduced nearly to zero, the passband approaches the internal analyzer resolution of approximately 8%. In the standard science mode the full energy range is covered in two spacecraft spins with even steps covered on one spin and the interleaving odd steps covered on the next spin (see Figure 13). Since the intrinsic angular resolution of the instrument is approximately 10 degrees and the spacing between energy sweeps is 22.5 degrees of spacecraft rotation, only about half of the angle space near the spin plane is covered by each detector half during a given spin. The TIMAS view cone angle of ~146 degrees, which is an odd half multiple of 22.5 degrees, was selected so that near the spin plane the energy sweeps of one side of the view cone fall midway between the energy sweeps of the other side of the view cone approximately one half spin later (see Figure 13a). This interleaving of look directions breaks down as one moves to detector sectors further from the spin plane; however, gaps left between successive measurements also decrease since the instantaneous 11 degrees field of view covers an increasing span of spin angle.
Systematic gaps in angular coverage are also avoided when only one MCP half is being operated. This is accomplished by off-setting the even energy sweeps relative to the odd energy sweeps by 11.25 degrees (i.e. one half a sweep period, see Figure 13b). Thus a single detector sector covers a checker board pattern in energy-angle space over a two spin period. The conjugate detector sector from the other MCP half fills in the missing spaces.
In the other basic science mode, the low energy sweep mode, only the 14 lowest energy steps ( 2.0 keV/e) are covered and these are repeated every spin (i.e. odd and even steps combined). However the offset between even and odd spins is retained to achieve the angular interleaving discussed above. For the background measurement that is performed once each energy sweep period, the energy passbands of EA1 and EA2 are offset relative to one another so that no ions are transmitted by the optical system. Thus any events detected by the MCP are due to noise or penetrating radiation. These background rates can be subtracted during detailed data processing on the ground.
The RAM control tables mentioned above can be used to correct for errors in the ROMs, unanticipated changes in analyzer constants, shifts in power supply gains, or simply to achieve an optimal measurement sequence for a special science objective.
Data collected during one spin are processed for telemetry during the succeeding two spin periods. There are two spin buffers (see Figure 7), one into which new data are being accumulated and the other from which data are being processed. Data are stored in the spin buffer according to M/Q (any 4 M/Q values selected by ground command), E/Q (15 values, background is treated as a 15th energy step), detector sector (28) and spin sector (16). This corresponds to approximately 27k elements per buffer, each of which is 16-bits deep. Each event which is processed by the detector ADCs and converted to a corrected 6-bit radial position plus detector sector identification via the 32k ROM passes through the hardware front-end event processing system. Events are suppressed, however, during the power supply settling time for each step (~9 ms) and during the 16th interval used for high voltage backsweep. The front-end event processing system includes the mass LUT which identifies events corresponding to any one of the four selected M/Q values by a combination of the 6-bit radial code and the 5-bit energy-step code. That is, the range of radial positions spanning the particular M/Q peak at each E/Q is encoded with the appropriate 2-bit mass identification. Events not meeting the criteria for selection are not stored into the spin buffer. For the background energy step the full radial range is divided into four fixed intervals, each interval being encoded as one of the four M/Q values for cataloging purposes. The M/Q LUT is in RAM and can be altered by command to correspond to different M/Q values, either using tables of precoded information in ROM or directly via an uplink.
In parallel with the data accumulation in the spin buffer, detailed mass spectra at the full 6-bit radial resolution are accumulated. Mass spectra are sorted according to E/Q and solid angle of incidence. The E/Q selection is made via any one of 32 ROM tables stored in the energy-angle LUT. Up to 8 E/Q bands are selected via each table. Note that an E/Q band can consist of either a single E/Q step or a range of adjacent E/Q steps, thus not all E/Q steps will necessarily correspond to a selected band. The solid angle sorting is selectable through any one of four ROM LUTs. The baseline solid angle division selects 6 nearly equal solid angles spanning the 4 pi sphere, one each parallel and antiparallel to the spacecraft spin axis and four equally spaced around the spin plane. There are two mass spectra buffers, but in contrast to the spin buffers discussed previously they are not alternated each spin. In fact, the minimum accumulation time into a given buffer is two spins and the maximum is 32 spins. The mass spectra buffers are also 16 bits deep but are protected against overflow that might result from the greater integration time.
The total number of "Fast" (~140 ns deadtime per event) counts and the total number of events processed by the event selection electronics are recorded in the Fast Event (singles) Counter (FEC) and Processed Event Counter (PEC) respectively (See Figure 7) for each of the 240 active accumulation periods during each spin. These count rates are used later to renormalize the spin buffer data before further processing.
The Event Mode is an alternative mode of science data acquisition in which individual 20-bit (6 radial, 5 detector sector, 5 energy step, and 4 spin sector) events are selected for telemetry. Every nth event is selected such that all regions of phase space are approximately equally sampled and such that the total number of events selected is consistent with available telemetry. In ground data processing these sample events can be renormalized to the Fast event count rates in each accumulation interval. In the event mode the on board data processing is minimal.
Other specialized modes include the Engineering Mode in which science data are replaced by engineering data, the MCP Calibration Mode in which the MCP gain replaces the spin sector in the spin buffer, the Mass Spectra Calibration Mode in which the mass spectra are accumulated by individual detector sector rather than the phase space solid angle of incidence, and the memory Dump Mode in which the contents of selected regions of RAM or ROM replace the science data. None of these modes will be discussed in detail here.
We first define a set of Data Products which includes Pitch Angle Distributions (PAD), Medium Resolution Distribution Functions (MRDF), Low Resolution Distribution Functions (LRDF), Mass Spectra (MS), Fast Event Counter (FEC), Processed Event Counter (PEC) and Background Counting Rates (BCR). These Data Products are generated from the data collected in the spin buffers, mass spectra buffers, and counters.
The total instrument operation is phased with an internal spin frame counter from which we define a Super-Spin (SS) period of 32 spins (~ 3.2 min). In general, science mode changes take place on SS boundaries and within a given mode of operation the pattern of measurements and Data Product generation repeats at least at this periodicity. Individual Data Products can range in temporal resolution (integration time) from one spin period to a maximum of one SS period. The various Data Products are briefly described below.
The block diagram in Figure 7 shows the flow of data through the TIMAS system. The data are sequentially processed by two SA-3300 microprocessors, each operating at 4 MHz. The first, referred to as the Data Processor (DP), performs the deadtime corrections, the spin buffer renormalization, the integrations necessary to produce the Data Products, and a 16-bit to 8-bit pre-compression during the spin period immediately following the data acquisition. Each element of the Data Products described above is a 16-bit value (clipped at 2^16 -1 to prevent overflow). Each element is compressed to an 8-bit value using one of a pair of 11-bit lookup tables selected on the basis of whether or not the input value is less than 2^11. The code that has been adopted is approximately logarithmic with a decode uncertainty of 1.8%.
Even after the significant reduction in data represented by the Data Product processing described above, the quantity of non-redundant data could still be approximately 16k words per spin. This is approximately 5 times the TIMAS telemetry allocation. The necessary further reduction is accomplished through prioritization of Data Products for telemetry and through the use of the Rice lossless data compression scheme (Rice, 1979; Rice and Lee, 1983).
Each Data Product is explicitly prioritized for each spin period of a SS period for both processing by the DP and for allocation of available telementry. These priorities are specified in the Processing and Telemetry Priority Table (PTPT) which is programmed into RAM by ground command for each data mode. During each spin period the DP processes Data Products from the previous spin's buffer according to the PTPT priorities and places them into a block of memory that is shared with the Instrument Mode Processor (IMP). It continues processing until either all products are complete or it runs out of processing time. As each Data Product is produced, a flag is set in shared memory to alert the IMP. During the next spin period the IMP uses the same PTPT to select Data Products for Rice compression if specified (each element in the PTPT carries a flag to specify whether or not Rice compression is to be performed) and formats the data into the telemetry buffer. The telemetry buffer is large enough to buffer approximately three spins worth of data. At the beginning of the spin period the IMP determines the buffer space available. It then proceeds to compress and format Data Products per the PTPT until it determines that the next product would overflow the buffer, until it completes all Data Products available in the shared memory, or until it runs out of processing time. Any of these reasons for termination could occur since the compressibility of the data and the time necessary to perform the compression both depend on the characteristics of data. However, the design goal is to be limited by the available telemetry in most cases. With experience we will fine tune the PTPT for various expected plasma conditions to balance the loads on the two processors and to make maximum utilization of the telemetry.
Some of the Data Products are entered into the PTPT as groups, all elements of which have "equal" priority in order to minimize biases in the data selection. There are two PAD groups (PADGP1 and PADGP2) and two MRDF groups (MRDFGP1 and MRDFGP2). For example, if O+, He2+, and H+ PADs are all specified in PADGP1, the first time PADGP1 is reached in the priority table the O+ PAD will be produced. If time permits the He2+ and H+ PADs will also be processed before proceeding to the next lower priority Data Product. If these remaining two PADs are not completed, the next time the PADGP1 is reached the processing will start with the PAD next in order, i.e. He2+. In this way we can achieve more uniform coverage of different species if that is desirable.
The Rice compression scheme relies on patterns in the bit stream and its efficiency is strongly dependent on the characteristics of the data. Simulations of the data expected from TIMAS resulted in compression factors ranging from just greater than unity to as much as a factor of 30 or more, depending on both the absolute count rates and on the energy and angular widths of the distributions. We anticipate a typical compression of ~ 1.7 for most of the interesting data regions to be covered by the POLAR mission.
Each Data Product packet starts with an 8-bit sync-word and a packet header with sufficient information to specify the decompression algorithm and the length of the packet. Given the compression scheme, a single bit error could affect an entire packet; therefore each packet also includes a check-sum for error detection. The length of spin packets can vary significantly from spin to spin. Any "unused" telemetry is zero filled; however, the telemetry buffer is of sufficient depth (approximately 3 spin periods) that this should only occur where the count rates are extremely low.
The Event Mode, Engineering Mode and Dump Mode data are not compressed, thus their telemetry formats are always fixed. These modes are however spin synchronous.