Slashdot Mirror
Looking At the Hardware and Software of NASA's New Horizons (imgtec.com)
alexvoica writes: Last week we learnt that Pluto has blue skies and ice water thanks to a series of high-resolution images provided by the New Horizons probe. But how is the probe taking these photographs and sending them back to NASA? What hardware and software systems are inside and who built them? Luckily, the New Horizons engineering team kindly answered these questions (and many others) in a detailed interview.
Here are some fun facts from my discussion with the engineers. The chipset: It might sound strange to some but NASA used to be a chip maker. Before using standard MIPS or Intel CPUs for probes like New Horizons, NASA had to design custom-built processors since the commercial solutions available at the time were not designed to handle the intense workloads of space travel. Inside New Horizons we find a radiation-hardened, MIPS-based Mongoose-V processor worth $40,000 apiece and built using a grant from the Goddard Space Flight Center. The camera: New Horizons has a multispectral 1 megapixel camera; sending a single 1200 x 900 image back to earth takes approximately 3-4 hours. The comms: Forget 4G LTE, New Horizons uses the very best! The probe relies on NASA's Deep Space Network (DSN) to make its long-distance calls. DSN is the largest and most sensitive scientific telecom system in the world and was also used to guide the astronauts aboard the Apollo 13 mission back to earth. Tom Hanks and Kevin Bacon remain forever grateful. The memory: New Horizons includes 16GB of flash memory which provides plenty of storage space for photos and other scientific data. The operating system: New Horizons runs on Nucleus, a popular operating system designed by Mentor Graphics. Coincidentally, Nucleus is also at the heart of the ARTIK 1 platform for IoT launched by Samsung only a few months ago.
Here are some fun facts from my discussion with the engineers. The chipset: It might sound strange to some but NASA used to be a chip maker. Before using standard MIPS or Intel CPUs for probes like New Horizons, NASA had to design custom-built processors since the commercial solutions available at the time were not designed to handle the intense workloads of space travel. Inside New Horizons we find a radiation-hardened, MIPS-based Mongoose-V processor worth $40,000 apiece and built using a grant from the Goddard Space Flight Center. The camera: New Horizons has a multispectral 1 megapixel camera; sending a single 1200 x 900 image back to earth takes approximately 3-4 hours. The comms: Forget 4G LTE, New Horizons uses the very best! The probe relies on NASA's Deep Space Network (DSN) to make its long-distance calls. DSN is the largest and most sensitive scientific telecom system in the world and was also used to guide the astronauts aboard the Apollo 13 mission back to earth. Tom Hanks and Kevin Bacon remain forever grateful. The memory: New Horizons includes 16GB of flash memory which provides plenty of storage space for photos and other scientific data. The operating system: New Horizons runs on Nucleus, a popular operating system designed by Mentor Graphics. Coincidentally, Nucleus is also at the heart of the ARTIK 1 platform for IoT launched by Samsung only a few months ago.
In the dawn of Silicon Valley the fabs counted on NASA and Military orders. For quite a long time they could count on 70+% of the production going towards NASA and military contracts. Almost no one else could afford the products at the time. Eventually Intel broke that mold by making a huge bet that they could slash the product costs and a wave of volume would follow to make the price point profitable. It was a huge risk.
A decade ago I spent about two years on an embedded system running Nucleus, spent several months fixing bugs in the threading primitives, including the core spin-lock mutex that worked about 99.999% of the time under low-load conditions, but whose failure rate rose rapidly with load to about 2%. So much fun. Parts of that codebase looked like they were written by very low skill programmers.
Why not just use an off-the-shelf processor and put the PCBs in shielded, shock mounted enclosures?
Several reasons. 1) Weight matters. Adding shielding and enclosures adds weight and thus cost. Sometimes the cost of the component is dwarfed by the cost to launch said component. 2) Off the shelf hardware doesn't always work for applications like these for a variety of reasons. Sometimes it's fine but not always. 3) A lot of this stuff was developed a LONG time ago and has to work with a lot of legacy systems. Off the shelf solutions don't always work for some of the problems they face. 4) Once they have a proven design, it is a non-trivial task to get a new piece of hardware qualified.
Radiation shielding in space is harder than you might think. You can't just add a thin lead sheet or other dense material around critical circuits and be done. When photons get above a certain energy level, they pretty much blast through anything dense. What you really need for shielding for those types of high energy photons is a _really_ thick layer of low density shielding, say several miles of a gas under about 1 atmosphere of pressure. But this is simply not practical on today's spacecraft. There are other approaches to shielding such as layering high and low density materials, but in a spacecraft shielding is always limited by volume and mass constraints.
The other option is to deal with radiation by building chips with redundancy. The idea is that if one part of the circuit gets temporarily zapped, two other parts are still functional and the majority is probably the correct answer. You also build the electronics so that if everything goes south they can reboot and recover.
NASA knows what they are doing here!
It is a myth. But the myth is grounded on real differences in approaches between the Soviets and the US at the time. Two examples:
I had a friend who went as part of a US military team as a translator during one of those mutual nuclear disarmament treaties in the 80s. The Soviets had sent their own team to the US. Each team was allowed to inspect any area large enough to conceal a "treaty-limited item", which was carefully laid out in the treaty. So the US team was sent over with laser measuring devices to figure out what they could inspect and what they couldn't. The Soviets thought the devices were really clever. What their teams had been sent over with was... a stick. If the stick fit, they could inspect it.
Another example: in the US, you know how if you get pulled over (at least at the time), the police officer would take your license back to the police car, key the number into a computer or radio it to someone who would key it in, this would look up the number in a central database for existing violations (letting them know if they needed to for example take the license away), then it'd register the new violation in the database, the officer would then get all the info, print up or write up the ticket using that, and hand the license back, right? The Soviet system was a bit... simpler. The officer would take your license and punch a hole in it. If you had too many holes, they'd keep it. ;)
The War of 1812... the good 'ol days when the federal government actually tried to save New Orleans.