Details on ISRO built Multi-head Star Sensor (under Laboratory for Electro Optics Systems)
Star Sensor, also known as star-tracker, is a high-accuracy 3-axis attitude sensor used onboard spacecrafts.
Basically, a star tracker is a electronic camera connected to a microcomputer. The camera part popularly called sensor head consists of camera control electronics and camera head electronics with baffle.
Its accuracy about the boresight is poorer than about the cross-axes. This is improved by using two sensor heads with staggered Fields-Of-View (FOVs) and three to avoid break during occultation of any head. All the sensor heads have identical processing operations. So, the Processing Unit (PU) is made common. This minimizes the system electronics, power consumption and also thermal dissipation on each Camera Heads (CH), allowing more efficient cooling of CCD and improving sensor performance.
The resulting multiple Camera Heads (CHs) are operated remotely by the common PU. Thus, a programmable Video Processor (VP) is designed for the CH as an efficient data acquisition co-processor to the PU. The VP works in parallel freeing PU for attitude computation from the data acquired from multiple CHs. VP acquires CCD images and pre-processes them to reduce data size, speeding up PU processing.
History & Geography of ISRO star trackers :
Laboratory for Electro Optics Systems(LEOS) of ISRO indigenously developed different types of star trackers.
The first generation star trackers of LEOS are based on 16-bit processor like 8086 processor operates in only few traditional classic modes like acquisition and track and process only few stars with limited update rate due to many constraints.
The second generation Mark-II star trackers are characterized by low weight, low volume and low power consumption with ERC-32 processor. To meet low weight and optimized optical performance, a seven element optics weighing about 350g indigenously developed in LEOS is used. A radiation hardened area array CCD of size 1024X1024 is used as detector. The processing electronics of sensor consists of ERC32 SPARC processor working at 12 MHz speed, in addition to main processor a custom made Video Processor (VIP) is used to perform the CCD related operations, this Video Processor acts as a co-processor for the main microprocessor.
Electronics consists of 3 different types of memories - PROM for Boot program, EEPROM as secondary storage and RAM as main memory. In addition to these memories, VIP has its own storage to deposit acquired and processed digitized data of star image. Once the data is deposited in the shared memory of VIP, the main processor fetches these data and performs the specified operations.
Video Processor (VP) Operations
The VP is programmed by the PU for each frame of image data acquisition and then initiated at a synchronized time. The VP then sequentially executes the instructions, controls and sequences all the associated peripherals to acquire star image data from the CCD, pre-processes it and stores the data in a suitable format to be transferred back to the PU. In this way, the VP allows the PU to select suitable heads and schedule their operations as required without actually involving in the image readout sequence. The VP is designed to ensure execution of a single action at any given time since the CCD does not support parallel operations.
The functions of the VP are implemented in two sections. The first section, the Video Acquisition section, consists of the Fetch Unit, Execute Unit and Transmit Buffer Write Unit. The second section, the communication section, consists of the SpaceWire Protocol, Transmit and Receive Buffers and the Bridge.
When sensor is activated initially, it has no information about the satellite’s orientation, this is known as the Lost-in-Space (LIS) or Initial Acquisition, the Acquisition mode can be separated into three main parts.
Once the initial attitude is estimated, this is used to estimate the orientation of the subsequent images, this is known as Tracking. It is based on predicting the current orientation and its rate of change accurately from previously obtained information. In track mode unlike in LIS instead of processing complete array of detector only selected area of detector is processed by defining a window called track window, this is necessitated to improve the throughput of star tracker. Mark-II track mode operation starts with predicting probable star positions in the FOV based on the previously computed attitude, rate and acceleration. These positions are used to define active windows in the FOV.
NOTE :
The star tracker with this software is flown in Indian SARAL mission and post performance of tracker is excellent. The tracker provides attitude with required accuracy with specified update rate. All the software logics built in the system are functioning normally in all space conditions and different satellite operations and make the sensor work-horse for future ISRO programmes ranging from remote sensing applications to inter planetary missions, which includes missions like Navigation satellites, Mangalyaan and Chandrayaan-2.
High speed autonomous embedded software for high accuracy star sensors - IEEE Conference Publication
These sensors are really important. They are what would be considered dual-use technology. They have a very important military application.
Star Sensor, also known as star-tracker, is a high-accuracy 3-axis attitude sensor used onboard spacecrafts.
Basically, a star tracker is a electronic camera connected to a microcomputer. The camera part popularly called sensor head consists of camera control electronics and camera head electronics with baffle.
Its accuracy about the boresight is poorer than about the cross-axes. This is improved by using two sensor heads with staggered Fields-Of-View (FOVs) and three to avoid break during occultation of any head. All the sensor heads have identical processing operations. So, the Processing Unit (PU) is made common. This minimizes the system electronics, power consumption and also thermal dissipation on each Camera Heads (CH), allowing more efficient cooling of CCD and improving sensor performance.
The resulting multiple Camera Heads (CHs) are operated remotely by the common PU. Thus, a programmable Video Processor (VP) is designed for the CH as an efficient data acquisition co-processor to the PU. The VP works in parallel freeing PU for attitude computation from the data acquired from multiple CHs. VP acquires CCD images and pre-processes them to reduce data size, speeding up PU processing.
History & Geography of ISRO star trackers :
Laboratory for Electro Optics Systems(LEOS) of ISRO indigenously developed different types of star trackers.
The first generation star trackers of LEOS are based on 16-bit processor like 8086 processor operates in only few traditional classic modes like acquisition and track and process only few stars with limited update rate due to many constraints.
The second generation Mark-II star trackers are characterized by low weight, low volume and low power consumption with ERC-32 processor. To meet low weight and optimized optical performance, a seven element optics weighing about 350g indigenously developed in LEOS is used. A radiation hardened area array CCD of size 1024X1024 is used as detector. The processing electronics of sensor consists of ERC32 SPARC processor working at 12 MHz speed, in addition to main processor a custom made Video Processor (VIP) is used to perform the CCD related operations, this Video Processor acts as a co-processor for the main microprocessor.
Electronics consists of 3 different types of memories - PROM for Boot program, EEPROM as secondary storage and RAM as main memory. In addition to these memories, VIP has its own storage to deposit acquired and processed digitized data of star image. Once the data is deposited in the shared memory of VIP, the main processor fetches these data and performs the specified operations.
Video Processor (VP) Operations
The VP is programmed by the PU for each frame of image data acquisition and then initiated at a synchronized time. The VP then sequentially executes the instructions, controls and sequences all the associated peripherals to acquire star image data from the CCD, pre-processes it and stores the data in a suitable format to be transferred back to the PU. In this way, the VP allows the PU to select suitable heads and schedule their operations as required without actually involving in the image readout sequence. The VP is designed to ensure execution of a single action at any given time since the CCD does not support parallel operations.
The functions of the VP are implemented in two sections. The first section, the Video Acquisition section, consists of the Fetch Unit, Execute Unit and Transmit Buffer Write Unit. The second section, the communication section, consists of the SpaceWire Protocol, Transmit and Receive Buffers and the Bridge.
- Fetch Unit (FU):It fetches instructions sequentially from internal RAM, decodes them and writes data to the relevant registers of Execute Unit.
- Execute Unit (EU): It acquires image data by driving CCD vertical and horizontal readout clocks in a phased and sequential manner.
- Transmit Buffer Write Unit (TxBufWU): It receives data from the EU and writes it into the transmit buffer in a SpaceWire appropriate format.
- SpaceWire codec: SpaceWire is a bidirectional full-duplex serial communication protocol. It is a low-power high-speed protocol operable at 2Mb/s to 400 Mb/s.
- Receive and Transmit buffers: They are dual-buffered FIFOs implemented in on-chip RAM to store the PU instructions and CCD star image data respectively.
- Bridge: It forms the link between the SpaceWire core and video acquisition section. It provides necessary interface inputs to the SpaceWire codec for configuring the communication between CH and PU.
When sensor is activated initially, it has no information about the satellite’s orientation, this is known as the Lost-in-Space (LIS) or Initial Acquisition, the Acquisition mode can be separated into three main parts.
- Star centre estimation.
Detection of star is a highly challenging task, especially when sensor has a high noise level relative to the signal level makes detection difficult, because the illuminated pixels do not 'stand out'. To estimate star centre in LIS mode, a method called binning is used, in binned mode instead of processing pixels by pixels a group of adjacent pixels defined by bin factor is combined as a single pixel. The processing of the binned pixel outputs is carried out in two stages, first stage is a detection stage, where the binned pixels which have been illuminated by a star called “litpixels” are identified using defined threshold with Sobel operator. Secondly, using a cluster of contiguous pixels those have been marked as "litpixels", the exact position of the source of the illumination is estimated. This is done by assuming that contiguously illuminated pixels have been illuminated by a single star whose image on the pixel array is circular. The estimated star centroid is converted to direction cosines co-ordinates called measured co-ordinates. - Star identification.
The process of star identification is to associate body-frame measured star image directions with the catalog reference inertial directions. LEOS first generation star trackers used in many remote sensing satellites uses Pyramid star identification algorithm presented by Mortari, here after we refer this as Algorithm-A. The Algorithm-A has many disadvantages in real time space environment, mainly the worst case run time of Algorithm-A is high. A state-of-art star identification algorithm here after this algorithm is referred as Algorithm-B is designed and developed for Mark-II star tracker. The success rate of Algorithm-B, is close to 100%, means provides identification solution at all times, which is major requirement of inter planetary and scientific missions. The successful star identification provides reference coordinates for the selected measured star vectors. - Attitude estimation.
Attitude estimation requires 2 set of vectors namely, measured vector and reference vector. As already explained measured vectors are out come centroid estimation process and reference vectors are out come of identification process. In this star tracker attitude is estimated in form of quaternions, two algorithms namely QUEST and Second Optimal Estimator of Quaternion ESOQ2 are studied and finally ESOQ2 is implemented.
Once the initial attitude is estimated, this is used to estimate the orientation of the subsequent images, this is known as Tracking. It is based on predicting the current orientation and its rate of change accurately from previously obtained information. In track mode unlike in LIS instead of processing complete array of detector only selected area of detector is processed by defining a window called track window, this is necessitated to improve the throughput of star tracker. Mark-II track mode operation starts with predicting probable star positions in the FOV based on the previously computed attitude, rate and acceleration. These positions are used to define active windows in the FOV.
NOTE :
The star tracker with this software is flown in Indian SARAL mission and post performance of tracker is excellent. The tracker provides attitude with required accuracy with specified update rate. All the software logics built in the system are functioning normally in all space conditions and different satellite operations and make the sensor work-horse for future ISRO programmes ranging from remote sensing applications to inter planetary missions, which includes missions like Navigation satellites, Mangalyaan and Chandrayaan-2.
High speed autonomous embedded software for high accuracy star sensors - IEEE Conference Publication
These sensors are really important. They are what would be considered dual-use technology. They have a very important military application.
