No RISC, No Fun!

Peter Gülzow, DB2OS
Translated by Don Moe, KE6MN/DJ0HC

Eine deutsche Version ist ebenfalls verfügbar, klicken sie hier 

"No RISC, no Fun" is the new motto for Chuck Green (N0ADI), Peter Gülzow (DB2OS), Lyle Johnson (WA7GXD), Karl Meinzer (DJ4ZC) and James Miller (G3RUH), who in the shortest time possible have developed a new high-power onboard computer for future satellite applications. If all goes as planned, this new computer will be subjected to its baptism of fire aboard the AMSAT-P3D satellite and demonstrate its capabilities at the same time.

It had already become clear during the development of the P3D satellite that the old onboard computer (Integrated Housekeeping Unit, abbreviated IHU) would be scarcely up to the challenge of future projects. However, no reliable alternative was available and the IHU had been successfully flown on OSCAR-10 and OSCAR-13 in the 1980's. Since it is the most important single element of the P3D satellite, all risk should be avoided. A permanent failure of the onboard computer would result in the complete loss of the entire mission. Therefore the old reliable technology was used.

The old IHU was based upon the COSMAC 1802 technology from the 1970's. The 8-bit CPU is still being manufactured and marketed by Harris. However, we use a radiation hardened version from Sandia Labs that was left over from former projects. The CPU is operated at a clock frequency of 1.6 MHz and thereby executes approximately 100,000 instructions per second. By today's standards that is very little, but in outer space everything happens so very slowly that the software spends most of its time in waiting loops.

For P3D several improvements were envisioned for the IHU since more telemetry and control channels were required, among other things. The basic design of the IHU has not changed significantly compared to OSCAR-10 or OSCAR-13.

For quite some time we have had alternative processors under review. There is a whole series of very interesting processors available on the market for so- called embedded applications, even the Pentium CPU. Many of these CPUs were quickly rejected after brief consideration since they often exhibited excessive power dissipation or their designs were not suitable for use in outer space. Only those CPUs could be considered that were built using fully static CMOS technology and exhibited extremely low power consumption and dissipation. Modern Pentium processors requiring over 5A of power can be named as a bad example since every user doubtless is aware of the annoying fan noise from his computer.

In addition to its radiation hardness, the power dissipation is the decisive aspect in choosing a processor. In a vacuum only approximately 15-20 mW/cm² can be radiated from an ideal black surface. Additional means of cooling are often mechanically complicated. On the other hand, raising the temperature by 10° Celsius reduces the lifetime by one half.

Even more important than the CPU are however the development tools. Since a new IHU would again operate with IPS as its operating system, the plentifully available assemblers and C compilers are not of much value. For the old IHU, the Atari 800 running IPS was used for both software development and the control software at the command stations. For some time now, comparable programs for a PC running Windows-95 have been available, although no IPS tools for developing and testing new IPS programs are available. James Miller, G3RUH, has been using the Acorn RiscPC for a long time, which is widely available in England, and has ported the IPS system in addition to all command software.

The Atari ST with the 68000 CPU briefly represented a further alternative for software development. However, this project was terminated after Atari abandoned the computer business. The Acorn RiscPC is however a readily available system that is capable of compiling the complete IPS code for the IHU. Such a cross development tool is very important since a whole series of software changes and enhancements will be required for the IHU to support such tasks as 3-axis stabilization. Furthermore, we do not wish to forego the comfortable environment of present-day computers. A cassette recorder was the only storage device available for the Atari 800.

The Acorn RiscPC is based upon the SA-110 StrongARM from Digital Semiconductor (DEC) and is operated at a clock frequency of 233 MHz. Still the RISC processor remains fairly cool since it dissipates less than 300 mW. For the end of 1997 Digital has announced an enhancement to the SA-110, the StrongARM SA-1100 with clock frequencies of 133 MHz and 200 MHz. The SA-1100 is a 32-bit RISC processor with an instruction cache of 16 KB and a data cache of 8 KB. It differs from the SA-110 used in the Acorn RiscPC mainly due to the integration onto the chip of a MMU and a whole set of functions for peripheral devices. Among other things, this includes four different serial multi-function interfaces, parallel interfaces, DMA, PCMCIA and LCD drivers, real-time clock, timer and clock generation, as well as functions for power management. The target market of this CPU is portable computers, such as PDA's, notebooks, organizers, etc.

Blockschaltbild des StrongARM SA-1100 von DIGITAL

We quickly recognized that the SA-1100 would be an ideal processor for a new IHU. Naturally it is fully static and exhibits an exceptionally low power dissipation of 200 mW at 133 MHz and 250 mW at 200 MHz. Its processing speed is rated at 230 MIPS (Dhrystone 2.1) at 200 MHz. It is therefore the most powerful processor on the planet as far as the MIPS per Watt rating is concerned!

On May 17, 1998 the entire operation of StrongARM, PCI bridges and network components, along with the chip fabrication facility in Hudson, Massachusetts, was transferred from Digital Semiconductor to Intel. Intel will continue development and manufacture of the SA-1100 and subsequent products. In the future Digital will only pursue further development of the DEC Alpha microprocessor and its systems. Hence P3D will also fall under the category "Intel Inside". Anyone interested in learning more about the StrongARM can find considerable information at Intel's website on the Internet under the URL: http://developer.intel.com/design/strong/

Unfortunately no data or reports are available regarding radiation hardness. Dynamic memory and processors customarily employ CMOS differential amplifiers at switching levels in the range of 100 mV. These components commonly fail at relatively low radiation levels of 1 kRad and become unusable. Fully static CMOS components have however a significantly higher "natural" radiation hardness that lies at least one, often even 2 or 3 orders of magnitude higher. Specific components can certainly be irradiated, but only their successful deployment in outer space will truly clarify the matter.

What would be more appropriate then than to fly the StrongARM SA-1100 on AMSAT P3-D?

Fascinated by the possibilities of a new IHU based on the StrongARM SA-1100, a working group was quickly formed. Our P3-D project manager, Karl Meinzer, DJ4ZC, immediately offered a chance to fly with P3-D, provided that the project is completed on time for integration and there are no problems with its mass. A free location was quickly found within the satellite for the module. It goes without saying that the previous IHU onboard computer will not be replaced. Additionally AMSAT-DL decided to support the project financially since it will prove to be important for future satellite projects.

In December AMSAT-DL held a 3-day design meeting in Marburg so that the new onboard computer could be realized as soon as possible. Prior to the meeting important technical details were discussed and determined, including the mechanical dimensions, weight and power consumption. A method had to be devised to integrate the new module into P3-D with the least possible effort and the fewest alterations to the cable harness. Since the originally planned 10m broadcast transmitter could not be completed on time, the new onboard computer has occupied its place. The required changes to the cable harness were therefore minimal.

An internal name for the project was also quickly found while enjoying a round of beers: YAHU, the acronym for "Yet Another Housekeeping Unit".

A list of system goals was started and all participants in the working group contributed many new ideas for the development of YAHU, also known as IHU-2. Lyle Johnson had already created a substantial portion of the circuit design and FPGA logic. The team discussed each subsystem individually. Lyle Johnson was also nominated by the group to be project leader for the new IHU-2.

Although YAHU is significantly more complex than the old IHU, the effort and space for individual components could be drastically reduced through the use of so-called FPGA's. The CMOS logic arrays employ the co-called "anti-fuse" technology and are thus particularly well suited for use in outer space. A radiation hardness level of at least 50K to 100 KRad (Si) can be expected. In all, three FPGA components will be used to hold the entire boot logic, command decoder, memory logic, EDAC, etc.

Werner Haas DJ5KQ, Peter Gülzow DB2OS und Chuck 
Green N0ADI mit der YAHU
Werner Haas, Peter Gülzow and Chuck Green with the YAHU Prototype
(Photo: W.Gladisch, AMSAT-DL)

While participating in the P3-D systems integration in Orlando in March 1998, Chuck Green (shown in the picture above) presented the first YAHU prototype which he had just populated with components and soldered. The YAHU consists to a large extent of SMD parts. Subsequently the prototype was sent by courier to Tucson, Arizona, where Lyle Johnson was eager to begin testing it. Chuck Green is responsible for the entire circuit board layout and manufacture. He had previously demonstrated his superb craftsmanship on the IHU, RUDAK and other circuit boards.

As a comparison of their relative speeds, the 32-bit IPS version running on the SA-110 clocked at 200 MHz will be nearly 5,000 faster than on the old Cosmac 1802 IHU. Thanks to its vertical structure and compact code, IPS-32 can run almost completely within the two cache memories. However, the data cache in the newer SA-1100 was unfortunately reduced from 16 KB to 8 KB. The speed improvement is nevertheless enormous.

So that the new YAHU does not become an end in itself and merely kill time in waiting loops on P3D, the functionality has been extended well beyond that of a normal onboard computer. Thus the YAHU will have a modem interface at the 10.7 MHz IF level with A/D and D/A converters (I/Q). The job of the DSP will naturally be performed by the StrongARM. This will allow, in addition to higher data rates, other modulation and encoding techniques to be evaluated for suitability, such as would be needed for a deep space mission to Mars.

But not only that, YAHU also will have its own eye. In the last 12 months there have been interesting new developments in the area of camera and CCD chips. A novel new technology, the so-called CMOS APS sensors can be very simply integrated into digital circuitry. In contrast to the previous CCD sensors, these picture sensors can be read out like an EPROM. After applying an X and a Y address, the 8-bit value of the picture information is available for the selected pixel. The complicated timing and digitizing of picture data become totally unnecessary. In addition it exhibits superb picture characteristics, such as a very large logarithmic dynamic range of nearly 120 dB (6 light decades) in comparison to a normal CCD sensor with only 60 to 70 dB. The well- known "blooming" effects of overloaded CCD sensors also disappear. The camera proposed by Peter Gülzow would have a resolution of 512 x 512 pixels with an 8-bit resolution for the brightness information (black/white). The hardware effort is minimal, as is also the interface to the YAHU. The radiation hardness of CMOS APS sensors is exceptionally good at nearly 1 Mrad. Customary CCD sensors have a comparatively low radiation hardness of approximately 10 Krad.

CMOS APS Kamera im Größenvergleich
CMOS APS camera in a size comparison
Bild aufgenommen mit dem YACE Prototypen
Extreme light differences, as recorded with the YACE prototype
(Photo: Peter Gülzow)

Originally it was planned to mount two of these cameras on the satellite. One camera on the upper side and the other on the lower side. The idea was to have YACE ("Yet Another Camera Experiment") film and thus document the separation of P3D after launch. Comparably spectacular pictures have already been provided by TEAMSAT, launched with Ariane-502, which, as it turns out, employed the same camera technology. Due to space limitations we have decided however to mount only a single camera on the upper side. Following separation it will still provide pictures of Earth and also be used by YAHU as a navigation instrument in order to determine the orientation towards the Earth precisely and to correct it with the momentum wheels. A subsequent job could be as star sensor for determining flight orientation, a task that the StrongARM could easily handle. The pictures from the YACE camera will initially be stored in the 8 MB memory of the YAHU. Without compression no more than 32 pictures could be so stored. Appropriate JPEG compression would allow many more pictures, even a film sequence to be stored. The YACE camera is not intended to nor can it compete with the SCOPE experiment.

MAQSAT-H und SPELTRA, 19 Sekunden nach der 
Abtrennung von AR502
Separation of MAQSAT-H and SPELTRA, 15 Seconds after launch with ARIANE 502
(recorded with the VTS System of TEAMSAT)

To return now to the YAHU, the command system is compatible with the old IHU and can even use the same command uplink. The downlink and telemetry will likewise be transmitted at 400 Bit/s over the middle beacon (MB). Originally the MB was only intended for the 2m transmitter since the limited bandwidth of the transponder would not support both an engineering beacon (EB) and a general beacon (GB). The MB signal will be generated in the transponder matrix and then fed directly into the IF. Thanks to this fortunate circumstance, we can now use this beacon for YAHU also without having to make any significant changes. Merely one additional wire must be added to the cable harness from the matrix to YAHU. As soon as YAHU is activated upon command via the IHU, it takes control of modulating the middle beacon. Interested parties can thus receive the YAHU telemetry using the same demodulator and software for the 400 Bit/s telemetry from the P3 satellites.

Like its predecessors the YAHU does not contain any ROM and all of its software can thus be loaded or replaced from the ground. The IPS operating system and the operational software will reside in an 128K x 32 x 3 Bit EDAC memory block. The logic for the hardware boot loader is implemented in a FPGA component. Additionally there is also a Flash EPROM of 128 KB which can be written to and erased under software control. The intention is to store a copy of the IPS operating system and the operational software in this flash memory. In order to avoid errors caused by radiation, the contents of the flash memory will be periodically checked and corrected as required. In the case of a computer crash, the command station could restart the computer immediately from the flash memory or reload the entire software from the ground. This technique should prove to be very helpful, particularly under unfavorable link conditions. Additionally 8 MB of unprotected SRAM memory is available for pictures from the YACE camera and other experiments. As already implemented for RUDAK and the various other experiments, YAHU will also have an interface to the CAN bus. This 800 KBit/s local network will allow data exchange between the experiments within the P3D satellite. For example, the pictures from the YACE camera can also be transferred to the generally accessible RUDAK mailbox.

Der erste Telemetrie-Block der neuen IHU-2
The first telemetry test transmission from YAHU
(Photo: Lyle Johnson)

As already mentioned at the beginning, the software for the new YAHU is an important aspect during initial testing of the hardware. James Miller, who as long-term Acorn user is very experienced with the ARM assembler, took over the development of suitable test software for the initial startup of the prototype and is also working on the important software interfaces for the upcoming 32-bit IPS. Step by step the various functions in the YAHU could thus be tested and put into operation. James Miller also ported the standard IPS (16 Bit version) to the ARM architecture and it is already running satisfactorily in the new IHU-2. ARM development tools are available for other environments, notably the IBM-PC and Sun Workstations.  But an IPS environment is presently only available for Acorn Risc Computers. See:  http://www.jrmiller.demon.co.uk/IPS
By early April 1998 it had progressed to the point where YAHU could send its first block of telemetry data. However, Lyle Johnson still had much debugging to perform on the hardware. Some of the modern components many times did not behave as described in the data sheets so that conversations with the manufacturers were often required to resolve the remaining problems. By the end of April 1998 Lyle Johnson and Chuck Green received additional support from James Miller, who made further software tests and changes on site in Tucson. This saved considerable debugging time, but apparently the nights were often very long and tiring, as the following picture shows:

James Miller und Chuck Green
James Miller and Chuck Green during YAHU startup in Tucson, Arizona
(Photo: Lyle Johnson)
Lyle Johnson presenting the YAHU Engineering Unit 
#1 (Photo: Heather Johnson)
Lyle Johnson presenting the YAHU Engineering Unit #1 (Photo: Heather Johnson)

As soon as the YAHU prototype operates flawlessly, the next step is to perform a "fit check" in the satellite and to construct the flight hardware. This also signals a major milestone for Karl Meinzer, since he is so enthusiastic over the development of YAHU and the possibilities of the StrongARM SA-1100 that he has promised to write a very powerful 32-bit IPS especially for this RISC processor. This will enable us in the future to perform more complex mathematical computations in the satellite, as would be needed for autonomic supervision and attitude control using the momentum wheels, the camera as well as the Sun and Earth sensors. In short, IPS-32 will optimally exploit the computational capacity of the StrongARM and catapult future satellite projects of AMSAT into the next century. Until then there is much to do however and we will hopefully report on it soon in further detail.

Übersicht | P3-D Übersicht | Glossar