I've been designing semiconductors for 20 years and I thought I would learn something but this article said nothing other than he wished there were more radiation hardened devices out there.
You can learn more from wikipedia than this article.
Here’s something. Rad-hardened parts are mostly BS. An MSP430 takes 20krads without shielding. Process shrink seems to increase hardness, not decrease. I used flash FPGAs with parity circuits because flash was supposed to be better for SEU. This is all theory; it’s really hard to test SEUs. Store firmware as a low-rate RS and do circuit parity checks to trigger firmware refresh. There is probably a smarter way to do circuit ECC without the EC as a weakness, but I don’t know. That was years ago, and we were just space cowboys giving the middle finger to the rad-hard parts business.
Not hugely informative, but nice. The title over-promises pretty wildly.
That's practically an understatement. It just barely touches on the challenges of extreme temperature range and radiation hardening. The difficulty of getting your design to pass worst case analysis can be rather dispiriting when you have to drastically derate it to account for -55 to 125 degrees C temp range. Further insult comes from derating to account for total dose radiation effects (depending on orbit), although the article seems to only mention the need to account for single event upset events. Handling SEU's is actually a pretty interesting design challenge. Total dose, on the other hand, is just a bummer.
Where are you getting that temp range? The last payload electronics I designed (an RF PA) only had to handle -20 to 70 C, and the actual temp swing is less. Of course this is inside a bus.
Military electronics are just as bad, as they have to operate from the Antarctic to the desert, and be thrown from the back of a helicopter onto concrete.
Tangential, but designing devices to handle shock is surprisingly hard as well. I was once tasked with building 900MHz GPS transponders that would allow us to see the shape of a tether being pulled behind an airplane. The device was about the size and shape of a soda can. Occasionally they would get dropped, impacting at terminal velocity.
One early prototype had a li-ion pouch cell (yes, I'd do things differently now). After an impact, I pulled it apart and found that the battery, via plastic deformation, had perfectly fitted itself to the contour of the inside of the enclosure. IIRC it was still working when I pulled it apart.
There were some non-NASA boxes that had long periods powered off on the shaded side of the bus and had to operate at powerup at initially very low temperature. And then operate on the sun side for extended periods at an elevated temperature. Depending on the configuration of the bus there were sometimes challenging heat transfer issues. The plate temp wasn't 125 C (I don't remember the exact number) but there was assumed a significant temperature rise from the plate to the electronics. And then additional margin added on to that in case something wasn't quite up to spec in the thermal path. So we had to perform worst case analysis assuming junction temperatures of 125 C. It was pretty awful.
Such fragile things it's amazing how little it takes to destroy a device. And how much energy can be put through the same fragile device.
And how much energy is in one Coulomb. The example I like is two points each with one Coulomb repel with a force of one millions tons.
A recent artie I read spoke about the reaction wheels of old spacecraft failing. Solar flares caused arcing in the metal ball bearings the would pit the bearing race. So if a steel ball bearing can't take it imagine fragile electronic devices.
> And how much energy is in one Coulomb. The example I like is two points each with one Coulomb repel with a force of one millions tons.
To have any meaning, you also need to consider the distance between the points. Any two charged points can repel each other with a force of one million tons given a particular distance.
> And how much energy is in one Coulomb. The example I like is two points each with one Coulomb repel with a force of one millions tons.
Coulombs and therefore their related units such as Farads are famous in Physics for being far too large, so this is not particularly surprising (assuming a sane distance). The capacitance of the largest capacitor bank in the world is about 0.2 Farads [1].
Keep in mind that capacitance and capacity are not the same. Case in point, the capacitor mentioned can hold a peak voltage of 24kV, which is allows for a million times more energy to be stored than the same capacitance at 24V peak. Capacitors with far lower peak voltages can have capacitance exceeding several Farads.
Yeah, I was taking some liberties with the definition of largest.
They're still a ridiculously big unit though. Generally the biggest man-made structures on Earth are measured in mega*. Single digits are usually used for things that are roughly apple-sized.
But in electricity/electronics, we deal with the entire range of metric prefixes… terabytes, nanofarads, gigaohms, microseconds.
I'm curious as to what you mean by "biggest man-made structures" though. For me that brings to mind tall buildings, long walls and big dams. I suppose that the Great Wall is 20 "megametres" long, but normally long dimensions will be in kilometres, heights are in 100s of metres and dams are in billions of cubic metres. Care to share some examples?
> But in electricity/electronics, we deal with the entire range of metric prefixes… terabytes, nanofarads, gigaohms, microseconds
But you don't. You'll almost never use whole Farads, for example. It's even on the Wikipedia page: related units, nF, uF.
> Care to share some examples?
I don't really mind the difference between giga/mega/kilo. I was really just talking about a ballpark where we don't want the biggest thing ever to be unity in our everyday unit.
You do raise a valid point with height. This is because the gravitational field in some sense makes length directional: 3km up is very different to 3km along. Clearly, we need some vector based measures so we could scale them sensibly, g-hat and x-hat : ).
Can you think of many man made structures who have just one of some extensive property in an everyday unit?
I read that article too and am skeptical. There is a less-than-perfect correlation between space weather an wheel friction increases, but these are not exposed bearings on the outside of spacecraft. These are deep inside, behind layers of metal parts. I don't see the mechanism for creating the imbalance of charges necessary to arc inside the bearings. If it is happening, I would expect this to be far more common in terrestrial bearings. Cars build up static charges. We aren't seeing their bearings degrading so suddenly when after they drive through lighting storms or other static charge imbalances.
You're looking at it the wrong way. Electronics work on earth by definition, as that's where we want them to work. You can be surprised that electronics can be made so fragile and still work on earth, but being surprised they work is begging the question.
> To mask the effects of upsets in the FPGA configuration memory, temporal or structural redundancy can be applied to the system. Temporal redundancy involves the
replication of a computation or logic function in time to mitigate failures that occur during one of the redundant computations. Structural redundancy involves the replication of selected circuit structures to remove single-point failures. Failures in the circuit can be masked by performing the logic or computing function in more than one circuit location.
> The most common form of structural redundancy is to apply triple-modular redundancy (TMR) [24]. As shown in Fig. 6, TMR involves the triplication of all circuit resources
and the addition of majority voters at the appropriate circuit outputs.
Reminds me of this Stack Overflow question "Compiling an application for use in highly radioactive environments" where the top answer is from someone who worked on mini satellites: https://stackoverflow.com/q/36827659
To some extend, space electronics have a lot in common with stuff used to manage industrial plants.
For examples in Chinese megascale steel mills, while SCADA's are used for monitoring and higher level control, operational aspects are handled by handcrafted PLCs made of modern analogues of 7400 series, 8t sram, redundant sensors, quoruming and BCH coding, mram, mems oscilators, and all solid DC-DC conversion, and all of that is provided with manual overrides
Most important feature of all of this is that it provides hard, near physical, impediment to wrong inputs from higher level control systems (SCADAs.)
> We’re enabling people to reach their full potential by bring them Internet access and that’s really inspiring for me. On the other hand, they also get access to Twitter –- you win some, you lose some.
Haha, brilliant.
What I always wonder though is how many of those people can actually afford a devise to make use of satellite internet. I guess it's still important to at least enable them to get it.
I always had the impression 99% of all space-electronic problems come from the fact that you have to lift stuff from earth.
You get plenty of sunlight to power your stuff and since gravity isn't as bad as on earth you can build stuff big enough to be shielded from almost everything.
You could install hundreds of tons of radiation shielding if you had easy lifting tech, but I still would point at the radiation as the root problem, not the difficulty of getting things into space.
I read so many stories about companies with new ideas for satellite internet. But i am visiting my parents next weekend. Their house is off grid in rural BC. They only have one internet option: a dish pointed to a geostationary sat which I will no doubt have to repoint on saturday. They pay insane rates for practically no bandwidth. There is only the one company: explorenet. When exactly are any of these innovations going to hit the canadian market?
Without asking too much detail, approximately where is there house? Is there any chance there is a regional WISP they could get service from?
In general, consumer grade VSAT service via geostationary satellite should be a last resort, if anything else is available. The economics of launching 5000 kilogram satellites into geostationary orbits mean that transponder kHz in Ku/Ka-band spot beams need to be significantly oversubscribed.
The actual cost of satellite capacity, translating dedicated (1:1 ratio) Mbps into transponder capacity, plus the cost of running earth stations on the ISP uplink side, can range from $1800 to $5500 per Mbps per month.
In order to make any money at all off a large number of $100 to $150/month consumer grade VSAT services it needs to be radically oversubscribed.
Xplornet has a particularly bad reputation as an ISP in general, which doesn't help.
Things like SpaceX's starlink or other upcoming LEO/MEO services like Oneweb are promising. But not available yet.
Point to multipoint wireless last mile via 2.4/3.5/3.65/5.2/5.8 GHz bands can be much more effective. It's even possible for WISPs to offer 75 Mbps x 25 Mbps packages based off the latest Ubiquiti 802.11ac gen2 platforms, Cambium PMP450 or Mimosa A5 AP radio platforms.
Disclaimer: I work in both satellite Internet and point-to-point/point-to-multipoint microwave and millimeter wave.
Egmont BC. Only 80km (50 miles) from Vancouver. Not exactly remote by Canada or even British Columbia standards. It is also only 79km from Whistler, but look at what stands between Egmont and whistler to get an idea BC terrain.
I've looked into every option. The irony is that the nearest cell tower is just over a kilometer away. The problem is the rocky terrain and dense pacific coastal rain forest. Trees heavy with water droplets suck up everything, from sound to radio. It is shocking how a heavy mist dulls everyone's cellphone boosters. Voice calls only last a second or two even on dry days. Everyone uses text messaging.
There have been many multipoint proposals. The problem is the rugged coast. Even a 1000' tower wouldn't have line of sight to every house. It would take all sorts of relays atop individual hills. And those relays need power, which is tricky. Buried lines aren't an option (rock) and towers are expensive (forest).
Yeah, that is going to be a hard location to reach. Took a look at it from Google Earth / satellite view for a few minutes. The best option I can realistically think of is for a group of 7 to 20 people to share the cost for a larger, much more serious geostationary vsat terminal (not some xplornet consumer grade stuff), like a 2.4 meter ku-band dish with 20W BUC and modern iDirect modem, and find a vsat ISP with ku band spot coverage of the area to pay for access.
You'd be looking at like $800 a month for a better chunk of bandwidth. Then divide that by the number of local users in the Egmont town are you can connect through it, building a very small micropop WISP setup. Something like a mimosa a5c on a pole in a central location and c5c CPE radios with 24-30dB gain dishes on the client side. And a small mikrotik router between the mimosa and the vsat modem.
Divided by enough people it could work out to around $80-100 per residence per month. This assumes that somebody with a modicum of networking clue can run the local end for free, a few hours a week for maintenance and monitoring.
Yup. Thanks for looking. Getting everyone on board with a 10 to 20-house collective would be very hard. The terrain is really unforgiving. All the houses are by the water, with steep rocky hills behind them. Any maintenance is a big issue. That "somebody with a modicum of networking clue" doesn't live in Egmont.
Atm my parents are paying 100/month for sat internet, and another 50 for sat TV. It suits their needs today but they know that when the grandkids are a little older bandwidth will be an issue. When I visit I bring them thumbdrives full of all TV shows they cannot get.
The other best possible option would be a single access point, somewhat up on a hill, possibly mounted to a tree with TV white space radio gear, just across the water from Egmont, with sector antennas aimed at the town. Redline and a few others have commercial TVWS band access point radios for the 500 to 800 MHz bands (various models available) which can cut through trees for non line of sight radio pretty effectively. You'd still need to get some kind of semi-decent dedicated broadband connection to the AP site, such as a 20 Mbps x 20 Mbps to Telus in Sechelt or Gibsons.
Looks like you can get 25 Mbps with a 300GB cap for $120. That's not bad IMO. I'm paying $40/month for 25Mbps but I also live in the city so other things are more expensive.
When exactly are any of these innovations going to hit the canadian market?
When are they going to hit any market? There are plenty of companies promising the world, but it is a difficult, expensive, and time consuming problem. I believe better options will come, but I don't see any on the near term horizon.
Great, but while i see lots of talk of this for ships and planes i've seen not a single file or application from spacex to become a residential provider anywhere, much less in canada. So they remain perpetually 10+ years away.
It is much more likely that it will be economical to use in a similar method to o3b, to bring a decent amount of capacity to a single location (example: totally isolated very rural valley somewhere in northern BC), and then distribute bandwidth from there.
The link budget problems for gain and EIRP, and cost of CPE, may make it cost prohibitive to put individual CPEs on peoples' roofs as a competitor to last mile WISPs.
I can see a scenario where a WISP buys a $7000 to $15000 dual-antenna terminal and a monthly recurring service package in the $700 to $1200/mo range, for a decent chunk of semi-dedicated capacity, and then redistributes access from there.
For those who want to understand how this works, o3b at MEO works on the same general concept, but at a more expensive and high bandwidth scale. It's been operational for years now. There's lots of good reference material out there on O3B.
Take the same two-satellite make-before-break handoff system used by o3b and apply it to a much larger number of LEO satellites, and smaller terminals, that's the general idea of how oneweb and starlink are intended to work.
That really is the elephant in the room where SDR is concerned (along with front-end). More realistically you could change protocol/encoding after launch.
Not necessarily. I've never heard of anyone doing it for space (maybe SpaceX?) but you could have an SDR driving a massive MIMO/beamforming array. The SDR can then dynamically "change the antenna" (change the radiation pattern) with a fixed front-end.
One use case I have imagined is being able to have a single radio module that can be used with a variety of front ends, but honestly I am not sure how big a market that is. Certainly I don't see SDR adding much value to high volume applications.
You can learn more from wikipedia than this article.
https://en.wikipedia.org/wiki/Radiation_hardening