2.5.2 Level 0 Instrument Source Packet Description

Processing of raw data from Level 0 packets will require extraction of the Instrument Source Packets from the Level 0 Product.

The following sections describe the packet structure and arrangement for five types of ASAR Level0 products, the Image mode, Wide Swath mode, Wave mode, Global Monitoring mode, and Alternating Polarisation mode. A description of decompressing FBAQ-coded data is given last.

Packet Structure

Level 0 Measurement Data Sets consist of Annotation Packet and Instrument Source Packets. The Annotation Packet contains five fields: date code, length of ISP packet, two error counters and a spare word. The Annotations are shown in detail in table 2.14 . The length word in this packet is corrected with any identified transmission errors (as compared to the ISPs) and is more reliable.

The Instrument Source Packets consist of a number of 16-bit words arranged in a Header which includes Identification, Sequence Control and Length. The Packet Data field follows the Header and contains a Data Field (sub-)Header, Source Data Field and optional Packet Error Control.

Bits are displayed left-to-right, numbering from 0, and with least significant byte on the right, as for ordinary numbers. The Packet header is shown in table 2.15 below.

In the Identification word, Version, Type, and DataField are constants, with Application ID represented in 11 bits. These vary by instrument mode as shown in table 2.16 . The Segment counter represents data segment sequence number and is reset to zero for ASAR mode changes. The Length word represents data lengths from 1 to 65536, except the minimum length is actually 1008 due to data transmission requirements.

The Packet Data entries in the table give details of the settings of the radar during operation. The code words for Pulse Repetition Interval, Window Start-time, etc are described in the third column showing the calculation for the code word. To obtain the actual value, manipulate the algebraic expression, so that PRI = codeword/RadarSamplingRate. For example, codeword=10300, RadarSamplingRate=1/19.2MHz, PRI = 10300/19.2E6=536 microseconds.

The compression ratio code is: 00 for 8/4 compression; 01 for 8/4 compression; 10 for 8/3 compression; 11 for 8/2 compression where 8/2 indicates 8 bits compressed to two using FBAQ flexible block-adaptive quantisation.

The Application Identification codes listed in table 2.16 appear in the first word of the Packet Header as the least significant 11 bits. The most significant bits are the pattern 10001. Therefore the first word of the packet for Image Mode will be 8E14.

Image Mode Packets

During Image mode a continuous series of pulses and echoes are acquired. The initial transmissions include a calibration sequence, but this changes to interleaved echo measurement and calibration pulses. Figure 1 shows the arrangement of transmissions and echo receptions.

Table 2.19 shows the Packet Format for averaging mode (4-bit quantisation). This format applies to both Image Mode and Wide Swath Mode. The filler bytes are added to ensure 972 bytes of source packet plus 6byte packet header and 30 byte packet data header are 1008 bytes or longer. This data size is the minimum transmitted in CCSDS format used in the spacecraft downlink.

Table 2.20 shows the format of the Calibration data, which are always quantised to 8 bits. The number of samples in a sampling window is given in the Window Length codeword in table 2.15 .

Table 2.21 shows the format of Noise data. The data are padded to 972 bytes if necessary.

Wide Swath Mode Packets

In Wide Swath Mode the radar is operated in subswaths, Q <= 6, and within a cycle a subswath may be sampled again, up to P samples per cycle, where P<=12. Each packet shown is comprised of data samples described in table 2.19 Source Packet Format Echo Data. Table 2.20 and Table 2.21 for calibration data, periodic calibration data and noise data also apply.

Wave Mode Packets

The Wave Mode packets consist of compressed echo packets (compressed 8 to 2), calibration packets and noise packets, interleaved as shown in table 2.22 , below.

In Wave Mode there are two calibration intervals, a chirp calibration with 2 windows (called simply "Calibration" in the table), and continuous wave (CW) calibration with 97 windows ("Calibration Sequence" in the table). The echo data is 8 windows and noise data is 8 windows.

In the wave mode data, each sampling window contains blocks corresponding to PRIs, and sample codewords with quantity specified in the data header. The 2bit codewords are uncompressed in accordance with FBAQ decoding scheme. This is explained in SECTION TBD. Filler bytes are added when necessary to provide 1008 byte blocks when the packet header (36 bytes) is included.

Calibration packets in Wave mode appear in either 3 windows or 97 windows depending on the calibration pulse (chirp and CW respectively). The data are quantised to 8 bits not compressed.

The Wave Mode noise packet is formatted to sign and magnitude with fixed exponent. The number of samples is defined in the data header Window Length word, divided by the Resampling factor.

Global Monitoring Mode

GM Mode echo packets are as described in Table 2.19 , Averaging Mode.

GM Mode Calibration data are described in Table 2.20 .

GM Mode Noise data are described in Table 2.26 .

Alternating Polarisation Mode

In this mode each sampling window contains either echo data, initial calibration, periodic calibration or noise data. The periodic calibration data is four sampling windows. In table 2.28 following, the format for VV-HH (co-polar) mode is shown. Cross-polar modes VV and HH are similar except VV becomes HV and HH becomes VH.

Packet contents for the AP source packets have been described earlier. Table 2.19 describes the echo data, Table 7 calibration, Table 8 for noise.

Decompression of FBAQ-coded data

FBAQ compresses the raw SAR echo data from 8 bits per sample to 4, 3 or 2 bits per sample, abbreviated 8/4, 8/3, 8/2. This is a lossy compression method, meaning that distortion in the form of "quantization noise" is added. The distortion is proportional to the degree of compression.

Decompression of FBAQ-coded data is accomplished with the use of look-up-tables, abbreviated LUTs. The LUTs map the codewords to normalized floating-point values. There is one LUT for I-channel and one for the Q-channel when corrections for the differences in analog-to-digital (ADC) converters are applied.

The LUTs are determined from the standard deviation and mean of the SAR data. The threshold values are not uniform but are selected to produce results with minimum least-square error, based on statistics of SAR echoes. The reconstructed value is based on the threshold values and code words. The same tables may be used for positive and negative values (the reconstructed values are symmetric about zero).

Table 2.29 , table 2.30 and table 2.31 show the threshold values for 4-bit, 3-bit and 2-bit reconstruction.

The reconstruction thresholds are

eq 2.1

where Dn is given in the tables above, and is the standard deviation and is the mean. For SAR data the mean is zero, and the standard deviation is given in the BlockID.

The decompression algorithm is then

For each range line

For each block in the range line

read the BlockID

Select a LUT corresponding to compression level and BlockID

unpack FBAQ codewords

reconstruct data using codewords and LUT entry

The LUT is an array which is 256 entries across (corresponding to the 8-bit BlockID) and 16-rows (for 4-bit), 8-rows (for 3-bit), and 4-rows (for 2-bit compression). The values of the entries of the six different LUTs (one each for I and Q, three different quantisations) are provided in the Instrument Characterisation File 6.5.2. .