The latest news from the Terra Bella team
Building the World's Most Capable Microsatellites, Part 2
Wednesday, March 19, 2014
we explored why the time is ripe for small imaging satellites. While building SkySat-1 and the data infrastructure that surrounds it, we have developed a set of design principles that are steeped in the fast-paced, creative engineering culture of Silicon Valley but also draw extensively from 60 years of US space experience.
Here in Part II we expound upon these principles, explaining how they enable us to build phenomenally capable products within tight time and resource constraints. We start with one that sounds obvious…
Do the right thing
Finding time to do the right thing is often difficult in a cost- and time-constrained environment. But the cost of the time needed to think deeply about a problem early on or solve an issue when it first arises rather than sweeping it under the rug will pay off many times down the road.
Striking the balance between time to do the right thing and the need to move faster, faster, faster is difficult but can be make-or-break for a small organization. We challenge our team every day to do the right thing.
That dimension matters
There is a famous saying (often attributed to Einstein):
Make things as simple as possible but not simpler.
Starting with the goal of simplicity has huge implications as a design matures. It makes debugging and testing easier. It keeps cost down and it allows for easier evolution of the design.
Simplicity doesn’t just happen - it requires deep thought, discipline and many iterations. “Can we make it simpler?” is a question you will hear in almost every design review at Skybox.
Modularity is an aspect of design that takes discipline and a long view, but enables fast progress in the face of uncertainty. It is absolutely necessary in a startup environment because it allows the team to react quickly to changing business requirements and postpone decisions.
A truly modular system (software or hardware) allows for change in a component or subsystem without large, system-wide impact. This flexibility makes it possible to postpone decisions until better information is available. Turns out that in a startup better information flows in every day.
Let's swap this one out
Hiring hands-on engineers that build on a modular architecture makes it cheap to try things, fail and try again. The goal of testing is not necessarily to succeed - it is to uncover points of failure early and fix them. That is how you find weaknesses in a design and create true robustness.
Our hardware engineers are itching to get out from behind a computer monitor and our software engineers build tests before they implement features. One of our core beliefs is that there’s no such thing as too much testing if you can afford it.
Where does this screw go? A little to the right…
In fact, we find excuses to integrate and test things early in a design cycle. Even if contrived, the time spent putting things together early is paid back many times over as the design matures.
Much of engineering is about striking balances. Recognizing that is an important first step. But we also need guidelines to help us in finding those balances. Here’s a snapshot of some of the balances we weigh in engineering everyday at Skybox.
George E. P. Box wrote that “… all models are wrong, but some are useful.”
Analysis is critical to design - without it quantifying the merit of design choices is practically impossible. However it is all-to-often that design teams spend inordinate time and resources on complex models that nature shows to be inaccurate anyway
Your timestep is too large
We believe one of the true arts of engineering is a right-sized model - one that balances sufficient fidelity to predict useful things while being simple enough that it is both manageable and understandable. Essentially Occam's Razor applied to analysis. Interestingly, there are actually hard statistical arguments that smaller models are better.
The best models guide the designer’s intuition and help them make decisions.
You can’t do everything yourself… Do everything yourself!
This is one of the hardest balances for us and one we wrestle with every day. We care more than anyone else about what we’re building and we have the best team in the world. So why shouldn’t we make it all?
But with a small team building a large and complex system, it is simply unrealistic to do everything in-house.
Can you hear me now?
We try to pick the important stuff - the stuff that will really move the needle - and make sure to do those things ourselves. When given a choice of two things we could make and only the resources to do one, we always pick the one that is the biggest lever. And we’ll never choose to build a piece that is a commodity - we just don’t have the time.
A corollary to this is to have good in-house infrastructure. Being able to build things ourselves enables quick hardware iteration cycles and the flexibility to pull something in-house if we need to. To that end, we’ve built lab and shop space here to make everything from quick-turn prototypes to parts that will fly in space.
Ford or Ferrari?
Engineers love performance - squeezing as much capability out of a system as is possible. And while optimization is important for building small, tightly integrated systems with tight resource constraints, it can also lead to a fragile design.
Effective design optimization balances performance in an ideal case with resiliency in the face of off-nominal conditions.
Redundancy is NOT the root of all evil
Redundancy is a powerful tool in the fight to build a robust system. It is also seemingly in direct conflict with the simplicity maxim. Thus redundancy has become a bad word in many scaled-down aerospace systems.
We believe redundancy is a good thing, it just has to be applied intelligently. A modular architecture enables targeted redundancy without permutational complexity and we make redundancy a goal
not a requirement
The delicate balance of simplicity and targeted redundancy can lead to a robust, capable and elegant system.
Leverage other industries
In engineering if there’s an easy way out of a challenge that doesn’t compromise other principles it’s foolish not to take it.
Technical problems in one industry are often similar to technical problems in other industries. And while there are factors and constraints in what we’re doing that are absolutely unique, there are also many areas where great technologies exist in other industries that are applicable to our needs and can save a lot of time. Here are a few places where we “cheat” by pulling from other industries.
We have heavily leveraged technology from the consumer electronics, industrial automation, and automotive industries.
The consumer electronics industry has invested $100 billion’s over the past 50 years in packing miraculous technology in tiny footprints. Your iPhone is the direct beneficiary of this. Here are five technologies that are as enabling to a satellite imaging platform as they are to an iPhone:
Super-computer level processing performance in tiny space / power footprint
Sensitive, fast and high resolution CMOS image sensors
Compact, high efficiency solid state radio components
High energy-density lithium-ion batteries
Dense, fast and robust solid-state FLASH memory
Unfortunately, the technology used in space electronics lags its commercial counterpart by 2-3 generations due to risk aversion and the cost of the traditional space-hardening processes.
Moreover, massive investment in process control has made today’s COTS parts incredibly reliable - especially those targeted at automotive or industrial applications.
One of the most critical enablers for our satellites are proprietary design, test and screening processes that enable us to fly cutting-edge COTS parts in the space environment. Commercial parts let us prototype quickly and cheaply, utilize technology years ahead of typical space missions and still have high confidence in the reliability of the system.
In Part III we will show how we applied these principles to the design of SkySat-1, elaborate on the way we balanced capability vs. size, and why we think the SkySat platform is right in the sweet-spot.
Posted by Jonny Dyer, Chief Engineer
No comments :
Post a Comment
Subscribe by Email
Official Google Blog
Lat Long Blog
Official Android Blog
Geo Developers Blog
Android Developers Blog