Microsoft word - 1$nt01!.doc
Application and experience of CAN as a low cost OBDH bus system
MAPLD 2004, Washington D.C. USA, 8th – 10th September, 2004
Surrey Satel ite Technology Ltd, University of Surrey, Guildford, GU2 7XH, UK.
This paper gives an overview of Surrey Satel ite Technology Ltd. (SSTL) use of CAN bus
on its recent missions. It gives a description of the SSTL CAN topology and goes into
depth concerning software protocols. The paper also covers chipsets (including COTS)
used in order to implement CAN in space.
The Control er Area Network (CAN) is a bus system used for the communication of
discrete packets of information across a distributed electronic system. It is ideal y suited for
CAN was adopted by the automotive industry in the late 80’s to handle the increasing
connectivity requirements of in-car electronic systems. CAN was original y developed by
Bosch and is now formalised in ISO-11898 Parts 1 and 2.
CAN features include prioritization of messages, configuration flexibility, multicast
reception with time synchronization, system wide data consistency, multi-master and
automatic retransmission of corrupted messages.
SSTL CAN Implementation
SSTL has been using CAN as an on-board telemetry/telecommand bus since 1995.
Initial y CAN was used with a centralised TTC system on FASAT-Alpha/Bravo. SSTL then
migrated to a distributed CAN solution, as commonly used in the automotive industry. Al
TTC functions have been exclusively implemented using the distributed CAN bus on 6
LEO missions (UoSAT-12, SNAP-1, AlSAT-1, UKDMC, NigeriaSAT-1 & BilSAT-1). The
CAN topology featured in figure 1 has been tried and tested:
Nodes within the spacecraft connect to both primary and redundant buses via a latching
relay. On power up, the relay of a node switches to communicate on the primary bus. If a
node does not receive a CAN message for 5 minutes, it assumes bus failure and switches
to the redundant bus. Modules connected to the bus fal into 2 categories:
• A data processing module such as an On Board Computer (OBC) has 2
connections to the bus. One is through an 8-bit CAN microcontrol er which provides
telemetry when the main processor is off. The other is via a CAN peripheral
• Less intel igent modules are connected to the bus via the 8-bit microcontrol er which
performs al the control/telemetry gathering needed.
Figure 1 – CAN Bus Architecture
This architecture ensures the survival of the mission upon a single hardware failure
anywhere in the system. Failures of the transceivers and other CAN interface hardware
are accounted for by incorporating complete hardware redundancy.
CAN SU Protocol
SSTL has developed its own relatively simple higher layer protocol on top of CAN. CAN for
spacecraft usage (CAN-SU) forces peer to peer addressing and is optimised for telemetry,
telecommand and buffer transfer. The development of CAN-SU was driven by SSTL
practices and had the fol owing requirements:
• Repeat build of spacecraft sub-systems for different missions
At the time of CAN-SU’s creation, 29 bit CAN identifiers were being introduced and
therefore it is limited to an 11 bit identifier. This has proven to be a limiting factor to the
The CAN-SU protocol splits a CAN frame up in the fol owing way (with the exception of the
buffer transfer) as shown in Figure 2 below.
Figure 2 – CAN-SU protocol in a CAN frame
The 11 bit CAN identifier field is split into 2 parts. The first part is the ID of a task running
on a processor (eg AOCS task), the other being a sequence number that is used in the
buffer transfers. Since the CAN-SU identifier field is only 8 bits long, the maximum number
of addressable processes in the system is 256. Not wanting to limit the amount of
telemetry channels on the spacecraft to 256 the decision was made to use the CAN data
field to hold an address field. This specifies a telemetry/telecommand channel giving a
total of 256*210 possible telemetry channels across the spacecraft and 210 channels per
task. Since the addressing of telemetry and telecommands across the CAN uses not only
the CAN ID field but also part of the CAN Data field, the use of remote frames2 is not
possible. CAN-SU therefore only makes use of CAN Data frames2.
The protocol forces peer to peer addressing by featuring the “From ID” in the CAN data
field, this al ows the receiving node to know where to acknowledge the request.
The “C” field specifies the CAN-SU message type (eg. telemetry request, telemetry
response etc). It is excluded in some packets of a buffer transfer in order to maximise data
throughput. The packet type is then determined by the sequence number within the CAN
ID field. The protocol features several message types some of which are detailed in the
2.1.1 Telecommand Exchange
Figure 3 – Example of CAN-SU Telecommand Exchange
The telecommand exchange (Figure 3) is a simple request/acknowledge transfer using the
CAN Data frame2 message format. Since the address of the telecommand is comprised of
both the ID field and the address in the data field and that a processor task typical y uses 1
ID, overruns of the receiving hardware buffer can occur, therefore an acknowledgement
message is necessary. The data field of the telecommand request comprises of an
address and the data or argument related to that telecommand address. The telecommand
acknowledgement includes a response field featuring any necessary data response.
2.1.2 Telemetry Exchange
The exchange of telemetry is il ustrated as shown in the fol owing diagram:
Figure 4 – Example of CAN-SU Telemetry Exchange
Telemetry is retrieved using a simple request/response method (Figure 4). Once again
using CAN Data frames2, the request packet features a 10 bit address and the response
2.1.3 Buffer transfer
CAN has a smal data payload per packet (8 bytes max) which minimises bus latency.
Therefore, it is not wel suited to large data exchanges between nodes. Therefore, CAN-
SU incorporates a buffer transfer scheme that speeds up large data transfers. This is
Figure 5 – Example of CAN-SU Buffer Transfer
The transfer is initiated by either a put or a get request to/from the data source. This is
acknowledged by the corresponding sink/source. The get request features a 24 bit buffer
identifier and a 24 bit length. Next in the transfer is the burst request. This features the
window size (number of packets in a burst). This packet also features the 1 bit “Br” burst
number used to keep sync of the bursts (in the “X” byte). The burst features a start burst
and up to 6 burst packets. A burst packet features 7 bytes of data and the from identifier.
After as many bursts are transferred as are necessary for the transfer, the session is
terminated by a Done and Done OK handshake.
2.2.1 COTS CAN Solution
SSTL specialise in providing low cost, predominantly LEO missions to a variety of civil and
military markets. It has been the ongoing philosophy to fly exclusively commercial off the
shelf technology (COTS) while mitigating the risk of radiation effects by module level
redundancy and passive fail-safe system design. This is possible as the LEO orbits used
range between 600Km and 1000Km where the radiation effects are relatively benign.
The total ionising radiation dose is in the region of 1Krad per year (at the component level
with 5mm of aluminium shielding), SEUs occur at a rate of approximately 1 SEU per Mbyte
of SRAM per day (from empirical observations) and observed SEL events are very rare,
with only 3 or 4 suspected cases in SSTL’s history.
It has been found that almost al CMOS COTS electronics wil survive 10Krads with no
significant degradation in performance. SELs are mitigated by over current trips on the
power system and cold redundant modules. And SEUs are mitigated by EDAC where
possible. Further SEU protection provided by the design of the satel ite system to be safe
should the on-board computer crash. For example, the thermal control is passive and wil
work with the satel ite in any attitude.
The use of COTS components al ows SSTL to rapidly utilise new technology as it enters to
commercial market. SSTL first flew CAN bus technology in 1996 and now uses CAN as
the telemetry and tele-command bus for al missions. See table 1 for components flown:
Philips CAN Transceiver: Physical CAN driver, current production
Phil ips PCA82C250: Physical CAN driver, going obsolete
Philips P87C592: CAN microcontrol er, obsolete
Philips CAN 8-bit peripheral, current production
Philips PCA82C200: CAN 8-bit peripheral, obsolete
Infineon: 8-bit CAN microcontrol er (A/D, PWM etc. 8051), current
Table 1 – COTS CAN components used on recent missions
Total numbers of a COTS component flown can be substantial. For example, the 8-bit
microcontrol er is used on al 30 to 40 modules in a spacecraft, so over the five year
lifetime of a typical SSTL satel ite, several hundred orbit-years of use is col ected. This
rapid accumulation of failure-free orbit years suggests that COTS is a viable approach to
2.2.2 Radiation Tolerant CAN Solution (RadCAN)
Recently SSTL has started moving out of the benign environment LEO missions. As part
of the British National Space Centre funding, a smal GEO platform program is being
developed under the project GEMINI. In order to meet the tougher radiation environment
and higher reliability (as GEMINI has deployable panels and therefore no passive fail-safe
attitude) A COTS approach is no longer applicable and a more traditional space industry
As SSTL has a lot of experience with the CAN bus, it was decide to develop a Radiation
tolerant CAN node – RadCAN. This was to al ow quicker migration of current designs
between the various environments as required.
Therefore, the single chip CAN microcontrol er functionality had to be re-created with a
discrete 8051 microcontrol er, memory, EPROM, ADC and FPGA, as shown in figure 5.1
Figure 6 – Discrete elements of the RadCAN solution
Using traditional space approved components a Latch-up immune, highly SEU tolerant,
100Krad solution was produced. However, there is a significant increase in design
complexity and size. The physical impact of migrating from COTS to radiation tolerant is
<0.75W (max), 0.20W (nominal) Estimate 1W
Table 2 – Comparison of a Radiation Tolerant to COTS CAN solution
CAN-SU has been tried and tested over a large number of missions and has been proved
robust. It does not however make the most of CAN. The use of 29 bit identifiers could
increase the amount of data available per telemetry request and also make use of the RTR
bit and other features unavailable through CAN-SU. Future work on CAN protocols at
SSTL is currently looking at the latest generation of 8-bit microcontrol ers to replace the
current infineon device. Initials tests of a Microchip PIC device have looked promising with
the 3.3V core providing a significant power saving. The processor can also be clocked up
to 40MHz providing up to 3.5MIPS if required (compared to the infineons 1MIPS).
As SSTL moves out of purely LEO missions, the need for more robust, radiation tolerant
systems increases. The next generation of RadCAN is being developed as a System on a
chip solution (SoC), integrating VHDL IP cores for a micro-control er, CAN core, memory
and EDAC in a single FPGA as shown in figure 7. The current FPGA baseline is the
mil ion gate radiation tolerant version of the Actel AX.
Figure 7 – IP cores to replace RadCAN
For applications not requiring microcontrol er functionality, RadCAN Lite is being
developed as just the CAN core with a simple state-machine for telemetry and
telecommand services. This design would be targeted at the smal er and cheaper Actel
With 10’s of orbit years of use on SSTL satel ites, COTS CAN has been proven to be a
suitable architecture for the radiation benign LEO missions. The combination of module
redundancy and fault tolerant system design has resulted in no observed failures of the
CAN bus or it’s constituent components.
With the developments outlined above, the CAN bus wil be an attractive alternative to
more traditional space bus architecture for al missions.
1. http://www.caen.it/micro/rd_casta.html: a Rad-Hard 8051 microcontrol er with on chip
2. http://www.can.bosch.com/: bosch’s CAN page
3. http://www.can.bosch.com/docu/can2spec.pdf : the Bosch CAN 2.0 B specification
4. http://www.can-cia.org/: CAN in automation
5. http://www.vector-informatik.com/canlist/: an informative CAN mailing list
An investor needs to be aware that there is a risk associated with making any investment and the return they are rewarded with usually reflects the risk involved. This guide explains some of the main types of risk. 1. Stock specific riskThis is the risk that there might be a reduction in the expected return as a result of some event or circumstance specific to a company. For example, the fortun
Hyperkalemia JOYCE C. HOLLANDER-RODRIGUEZ, M.D., and JAMES F. CALVERT, JR., M.D. Oregon Health & Science University, Portland, Oregon Hyperkalemia is a potentially life-threatening metabolic problem caused by inability of the kidneys to excrete potas- sium, impairment of the mechanisms that move potassium from the circulation into the cells, or a combination of these factors. Acute epi