BLOG
Computers for Interstellar Missions – Part 1
by Dimos Homatas
In this article, we aim to open up discussions on hardware electronics and computers for the Icarus and interstellar space missions in general. Introduction In today’s society, everybody is witness to the enormous advances in computer technology from ENIAC’s day until today. Home computers routinely implement features which a few decades ago were found only in mainframe systems. Virtualization, multi-core processing and even the existence of personal computers are but a few examples. To briefly quantify these advances, let consider IBM’s classic mainframe computer S/360 which was a great commercial success, since its introduction in 1964. After 10 year of continuous technological evolution, its benchmarks peaked at approximately 1.7 MIPS (Million Instructions Per Second). Compare this to a modern commercial home CPU, such as Intel’s I7 processor, capable of 147,000 MIPS and we see five orders of magnitude improvement! Note we are comparing the performance of a mainframe with a commercial CPU. The main drive behind this evolution is the integrated circuit. The continuing miniaturization of ICs led to the coining of various terms that quantify transistor density, such as Small Scale Integration, Medium SI, Large SI, Very Large SI and finally Ultra Large SI. An empirical observation, which turned into a trend and was finally branded as a Law, is Moore’s Law [1]: “The number of transistors that can be placed inexpensively on an integrated circuit has doubled approximately every two years”. Now, let’s make clear that the term “Law” is indeed a law in a sense: it has become an industry standard, dictating the targets for microprocessor development. To that end, Moore’s Law serves as the most credible tool that will help us predict the advancement of complex electronics. In the original Project Daedalus papers [2], factors such as processor speed, memory and storage are taken into account using the standards of that time. A cursory comparison reveals the limited computing resources they had available. In turn, these led to significant concerns relating to issues such as (a) problems arising from the time needed to compile a program, (b) the need for working memory and (c) the cost of software development to mention but a few. Do those problems still apply today? Are they relevant? Well, the answer seems to be both yes and no. For example, the same code that used to put a computer in an endless loop 50 years ago can still be used today to produce the same situation. The only difference is the processing speed. An endless recursion will produce a stack overflow error all the same. The difference is that in modern computing environments the error will surface much faster. On the other hand, higher processing speed, larger and faster working memory and huge amounts of storage provided the flexibility to write progressively complex software using various programming techniques like object oriented programming and adding more and more levels of abstraction. It seems that with modern computers, many problems that were present in Daedalus (but still faced with ingenuity) are of no relevance today. Still, the task of creating a mostly autonomous system that is faced with decades of space travel in the hostile environment of space, reaching a target star, gathering and transmitting back information, is not solved by raw processing speed and huge storage. Icarus is not a simple communication satellite. The complexity and the duration of its mission raises the bar for the computing needs. In other words, we need to pack in a so far unprecedented amount of computing resources. Hardware considerations – The Radiation Hazard Of all the hazards a computer may face in space, the biggest threat is radiation. Radiation which comes in various flavors as well (take your pick): protons, electrons, heavy ions and cosmic rays, encountered in a variety of different space environments, from Low Earth Orbit to Interstellar Space. There are two main categories of effects on electronics (a) Total Ionizing Radiations and (b) Single Event Effects [3]. As an example, let’s take the Single Event Effect (SEE). A common error is the bit flip (the change of state of a single bit from 0 to 1 or vice versa). A SEE can be caused by a single charged particle passing through a circuit. Now, a bit flip in allocated memory space can be catastrophic for the execution of a program, or could significantly distort data. There are several techniques of handling this issue, such as cyclic redundancy checks, cross referencing memory and any number of “safe” digital encoding/decoding methods which monitor data integrity. All of which have limitations, which must be scrutinized by the Icarus Computer and Data Management design team. One solution for this and other kinds of problems is the technique of Radiation Hardening (RH). RH is simply this: use circuits with a scale of integration closer to the low end rather on the high end, thus reducing the possibility of the particle flipping a bit. So, the general principle so far is that the more low-tech we go, the safer we are. This principle seems to be in conflict with the Icarus design philosophy, which calls for advanced solutions resulting from projections of the current state-of-the-art. We really do want all the high-tech goodies we can fit in. For example we could easily design computers that use cheap, commercially available Error Correcting Code (ECC) Memory. The situation is different regarding processors, but there are solutions there as well. For instance, three identical systems running the same programs concurrently and voting for the accepted result. Or even better a distributed design with no single point of failure [4]. By the way, both have their uses – smaller voting systems could be placed at robotic equipment, while the main computing facility would work using a distributed model. We can count several other examples of possible failures due to different kinds of radiation and the effect on electronics, such as channel degradation and transistor over-doping, to name a few. So how might the Icarus protect itself during its almost century-long journey? Some early thoughts point towards the Icarus employing some type of magnetic shielding [5]. Much like the Earth’s or the Sun’s magnetosphere, such a system would be able to deflect dangerous cosmic rays from inflicting damage to the electronics of Icarus. Would this be enough? Well, perhaps it is sufficient for the duration of the travel. But what happens when Icarus reaches the target star and possibly deploys multiple probes, orbiters or even landers? Suddenly the task of radiation shielding becomes more crucial, since several new factors make their appearance. For example the target star’s solar wind. Is it sufficient to deflect cosmic rays? Or an orbiter going through a planet’s Van Allen belt. It seems that radiation shielding in one form or another for the electronics is inevitable. NASA has performed some important research on “Commercial Microelectronics Technologies for Applications in the Satellite Radiation Environment” [6], essentially looking for ways to use commercial off-the-shelf (COTS) technologies for satellite applications. A notable example is the 1997 Mars Pathfinder mission which used an COTS Motorola radio modem for its communications with Earth [7]. The NASA taskforce proposed a methodology for addressing design specifications regarding radiation hazards for hardware:
- Define the hazard
- Evaluate the hazard
- Define requirements
- Evaluate device usage
- Engineer with designers
- Iterate process as necessary
The methodology in itself is pretty solid. In fact, it shares some modern software development methodologies. The main problem with applying this methodology to Icarus comes with the first step: Define the hazard. Can we do this accurately? To what extent can we safely ascertain the characteristics of a radiation environment that is light years away? These are all considerations that are being addressed as part of Project Icarus – an extremely complex and multi-faceted spacecraft systems engineering challenge. Our objective is to rigorously inform the Icarus designers, allowing for the latest available state of the art technology that may be available through reasonable projections. However, having current computing technology on-board Icarus is a goal on its own: so far most spacecrafts use technology much older than any of us have at home. As the Mission’s 2nd Term of Reference [8] states: “The spacecraft must use current or near future technology and be designed to be launched as soon as is credibly determined”. It would be an irony if we had to use older technology for Icarus. After several modifications (some small, some not so small) we should be able to give Icarus as much computing power as needed in order to provide the facilities required by every mission module. The fact that this technology will be obsolete upon arrival, should be considered as irrelevant 
References
- Moore, Gordon E. “Cramming more components onto integrated circuits”, 1965.
- T.J. Grant. “Project Daedalus: The Computers”, 1978.
- Kenneth A. LaBel, Allan H. Johnston, Janet L. Barth, Robert A. Reed, Charles E. Barnes: “Emerging Radiation Hardness Assurance (RHA) issues: A NASA approach for space flight programs”, 1998.
- Andreas C. Tziolas, “Distributive Computing Architectures for Icarus Daedalus-Like Interstellar Missions”, Presentation at BIS Symposium: Daedalus 30 years later, September 2009.
- Adam Crowl, “Shields for Icarus: Part 2 – Navigational Deflectors for Real“, Project Icarus Blog Article, 2010.
- Kenneth A. LaBel, Michele M. Gates, Amy K. Moran, Paul W. Marshall, Janet Barth, E.G. Stassinopoulos, Christina M. Seidleck, Cheryl J. Dale: “Commercial Microelectronics Technologies for Applications in the Satellite Radiation Environment”, 1997.
- Scot Stride, Microrover Radio Modem, NASA-JPL, 2007.
- Project Icarus: Terms Of Reference, 2009.
Pingback: Tweets that mention Computers for Interstelllar Missions – Part 1 — Project Icarus -- Topsy.com
I fail to see how, AFTER arriving at the system, radiation shielding is any more challenging than it is here, where we have lived with it for many decades.
My bet is that ECC memory and fault tolerant software will make it possible to use COTS computer hardware quite easily. Mount most of the electronics behind the fuel, and there will not even be a particularly challenging radiation situation in flight. Once arrived, probes need to be just as robust as the ones we have here, albeit more autonomous.
Unfortunately, once the fuel’s exhausted after the boost phase, the radiation is there to sleet through the processors at will. Project Longshot accelerated all the way to Alpha Centauri, so for most of the time fuel was available as shielding. Unfortunately, such low accelerations increase overall trip time. Clever arrangement of hardware and structure might mitigate some radiation, but it’s most likely that the Icarus team will have to baseline at least some dedicated radiation shielding – otherwise they won’t fit the “near future tech” requirement.
What about superconducting magnets? The strongest we have now produces a magnetic field something like 700,000 times that of the Earth’s.
Thanks for your comments.
Indeed, protection from radiation by positioning the components in various configurations (depended on several factors, spacecraft geomentry included) is a valid method.
An example of this is using shielded boxes to house electronic equipment. These boxes could be placed in a way to form another bigger box, several layers thick perhaps, so it can provide even more shielding to more sensitive equipment (see ref 6).
A consideration that will eventually arise in the design is overdesigning. Where does system security stop and overdesign starts?
This is the part where we need to take into account factors like the environment, shielding/spacecraft mass, etc.
Stumbled on this project recently and gravitated straight to this blog due to my back ground. I find myself curious what you have in mind when you say the project must use advanced solutions from the current state of the art. For example, if you built the ship today and were going to send it out on its way tomorrow would you use the types of chips in the machine under my desk now; something along the lines of a modern i7 or i5 processor core? Or would you settle for a custom build that uses the latest manufacturing techniques but maybe using an older class server design from 5 or 10 years back? The latter would probably be wiser…
I ask the question only to say that latest technology is a bit of a loaded term when talking about the computer industry. The fastest CPUs out there are on the cutting edge of transistor technology in terms of speed and density, which tends to mean they push the envelop on what the current manufacturing is capable of so hard that they burn out quickly.
Cosmic rays and ionizing radiation are not your only concern. Even a perfectly shielded computer is going to generate its own internal heat and wear, and while that isn’t going to cause random bit flips the way a cosmic ray strike will do, it will wear out the substrate and dull the “crispness” (if you’ll allow a metaphor) of the 0′s and 1′s being pushed through the transistors fairly quickly. I think space craft use older computer technologies for their time as a general rule due to the issue of ordinary wear. If you need an integrated circuit to last long, you don’t want anything like the off-the-shelf parts that are at the leading edge of the Moore’s Law curve. You need something far less aggressive that uses stages of the process curve that are a few generations old. Probably unique materials too that are chosen around the need for long life, with a likely sacrifice in raw processing speed and (unfortunately) cost.
That said, the other poster is right, ECC deals with random errors fairly well, but for something like what you have in mind you’ll still need the sort of redundancy you speak of in addition to as much ECC as you can design into your chips. Not just ECC on memory cells, but ECC on internal registers and on as many intermediate buffering points in the data path as you can reasonably afford to design in.
Just a follow up on the previous comment. The same semiconductor process technology that improves compute power also poses novel problems for space travel. At today’s clock frequencies and transistor junction voltages, you will need far more sophisticated fault detection, isolation and recovery mechanisms.
I just stumbled across this project and am struck by both the audacity and enormity of the project! Just on the computer side, not only do you have this redundancy design issue but also longevity, I doubt anyone knows if the COTS processors/memory/etc will work for a hundred years, or even if the 99-year old replacement part waiting to be plugged in will work when its turn comes.
It wasn’t mentioned but I hope it is part of the process to analyze all of the computer issues and failures of recent probes for trends and commonalities that can be fed into the requirements matrix.
Thanks for your comment Mark.
You’re right, the scope of this project is enormous, and there’s clearly a limit to our ability to predict component survivability rates for mission durations that are longer than the length of time that the technologies have existed so far!
Still, we have to start somewhere, and we can make informed estimates of these factors. I anticipate that over the coming decades we – and other designers – will revisit the design to see what we got right and where our estimates and design assumptions went awry. In many ways, that’s what we’re doing to the Daedalus design that was produced back in the 1970s.