My thoughts on the implications of Dragon’s successful flight, over at PJMedia.
30 thoughts on “America Has Returned To Space”
The somewhat depressing thing is, the same unimaginative little people who predicted and bemoaned “the gap” if SLS wasn’t fully or better funded, will be the same rodents who bash on and nip at SpaceX’s heals for any accidents that occur in the future. Memories of the Shuttle disasters won’t prevent them from saying that “This accident happened because some capitalist put profits ahead of safety!”.
They could use Dragon much like it is and it would be about as safe as Shuttle (if not safer). STS didn’t have a launch abort system. Yet they are working on a launch abort system and have been testing the rocket engines for it. A system like that should definitively be safer than Shuttle.
It’s already safer for at least two reasons. It’s much less complicated meaning less to go wrong. It’s shape is self stabilizing with a much better heat shield protecting a much smaller critical area.
Also, twice it’s demonstrated the ability to splash down accurately on it’s target.
And here I thought it was safer because the orbital return vehicle was placed above debris sources from the rocket itself.
That too.
My comment there:
According to information on the Vandenberg website, the minimum inclination you can launch into from there is 51 degrees. This is just slightly less than the 51.6 degree minimum inclination the Russians have been launching into from Tyuratam (Balkinour) for decades. The latitude at that site is about 46 degrees but range constraints (not wanting to drop rocket stages on China – some people have no sense of humor) limits them to 51.6 degrees, regardless of whether they’re going to the ISS or sending a satellite out to GEO. That’s why the ISS is in a 51.6 degree inclination orbit (as were the Salyuts and Mir stations).
The Russians pay a huge performance penalty for launching satellites to GEO from Tyuratam. The biggest part of the penalty is the amount of energy it takes to lower the inclination to near zero degrees. Years ago, I read in Aviation Week that their Proton booster could put 20 metric tons into LEO. It could send 2 metric tons to GEO (the upper stage took up the other 18 tons) or 5 MT to Mars.
If SpaceX tried to launch to GEO from Vandenberg, it would face similar performance penalties. Good thing the Falcon Heavy is so massively powerful. Still, it seems unlikely they’d do that because its just so inefficient. More likely, they’ll modify their launch pad at the Cape (which used to launch Titan IVs) or build a new launch site such a the proposed one near Brownsville, Texas.
The first Falcon Heavy launch pad is being built at Vandenberg. There’s a lot of speculation as to what the first payload will be. Often, first launches carry an inert test payload so you don’t risk loosing something expensive. Personally, I’d love for them to send a Dragon capsule on a looping flight around the moon. This would require some modifications to the guidance system (which depends heavily on GPS signals that wouldn’t be usable) and boosting the TT&C subsystem to handle long distance commanding and telemetry. Not only would this flight be exciting, it would give SpaceX a chance to test their heat shield on a return from deep space. It was designed for this but it still needs testing.
I don’t think such updates would be very difficult, especially since they have so much internal volume to mount communications equipment.
Does the Dragon already have a star tracker? (I think I have a much better algorithm to implement one of those, BTW)
Perhaps the biggest modification to the TT&C system would be to mount some deployable, steerable high-gain antennas on the service trunk similar to those on the Apollo CM. They might need to boost the transmitter power but not too much if the antennas are good enough. Dragons are far more automated than the Apollo capsules but I suspect their telemetry data rates are considerably higher, too.
They launched the Clementine lunar satellite in 1994 from Vandenberg and it worked well. I think a Falcon Heavy would have plenty of power to send a Dragon on a loop around the moon from Vandenberg. If they add some propellant tanks inside the Dragon, they might even be able to do an unmanned reprise of the Apollo 8 mission profile and actually enter lunar orbit.
Or they could send it to EML-2 and let it beam back picks of the moon and the earth form deep space. They could also use it as a functioning Dragonlab mission.
NASA considered easily adding automatic uploads to the Apollo GNC but decided to let the astronauts input the information as a redundancy check and to keep them in the loop.
A steerable high-gain antenna could take years to develop, unless of course SpaceX prototypes it by slapping two stepper motors or servos on a stick and mounting a Dish from Radio Shack, in which case it will take them a couple weeks to have a final system up and running. ^_^
Yeah, if SpaceX tried to buy a steerable, deployable high-gain antenna from existing contractors, it’d probably go the same as what happened when a cryogenic valve manufacturer was asked for a quote. They quoted an absurdly high price (tens of thousands of dollars, if not more). SpaceX did the valve themselves for almost nothing.
THURSDAY, MAY 31, 2012
1030 GMT (6:30 a.m. EDT)
SpaceX will soon close Dragon’s navigation bay door to prepare for the ship’s re-entry. The compartment holds the spacecraft’s thermal and laser rendezvous sensors, star tracker system, and the grapple fixture used by the space station’s robotic arm.
All I can find on the tracker is a tweet from Elon Musk.
Navigation bay pointing to deep space and star map being generated by star tracker one. Yes!
I wonder if they’re using an off-the-shelf solution like the ones from Space Micro or another vendor, or something developed in-house?
I think every tracker out there is based on picking geometric shapes out of a star field (Example) but I came up with a fast and easy way to turn a star field into a rotation-insensitive number that uniquely identifies each star (center on a star and in effect spin the image to turn the neighboring stars into a bullseye barcode), with only a few duplicate hits in a catalog of several hundred thousand stars, and all without using floating-point. If the tracker could center a star in its FOV, you could even do it with a really odd 40 or 50 pixel CCD (as concentric rings of constant area that output a zero or one) that would directly read as a database key identifying the star. For variable stars or stars on the edges of the rings, you input the alternate possibilities into the database so you get a hit either way. If you don’t get a hit you just pick a different star in the same field of view and try again.
It seemed to work well when I was playing around with it a year or so ago (with no actual hardware, just star catalog data and simulations). I was experimenting with 3/4 to 2 degree fields of view and 32-bit to 48-bit keys, for use as a small finderscope that would output high-accuracy RA and Dec as soon as you pointed it at a star field, which would be a very handy accessory for Dobson alti-azimuth telescopes. The same idea should work better in space (vastly more reliable images and no skyglow).
How do you deal with the non-stars? That is, things with vaguely the same apparent magnitude (Venus, Mars, ISS, whatever).
The non-stars are pretty easy, as are supernovae and other oddities. The concentric rings have a maximum real and pixel diameter (set when you chose the hardware). Let’s call that the barcode diameter, for lack of a better term.
You make the barcode diameter a little less than a third of the field of view, so that at least nine complete, non-overlapping barcodes could fit onto the image plane of the sensor, with a little slop left over because a good star won’t necessarily be in the center of these circles. Since any non-star or new star can only screw up the area within a barcode-diameter of itself, there would be many other areas on the sensor where you could still find a valid star in the catalog, unaffected by the anomaly.
With a 1 megapixel sensor this would mean the barcode diameter is a little less than 300 pixels, and you only need 40 or so rings for the barcode, so it should work out okay. More megapixels is obviously going to be better in this regard. With some though, you could probably have a bigger barcode and still ensure valid data.
For satellite use, it would be desirable to have mutliple tracker windows pointing in different directions, so that even in low orbit some sensors are always pointing into deep space, and also so that the system can ignore a camera that won’t produce a valid position without further processing to handle planets and moons (as when Jupiter and it’s moons are interfering with an entire camera’s view).
And of course you have to use mulitiple encodings for some stars because their barcode would include variable stars at the theshhold of making a ring a one or a zero, and to allow for a little bit of slop on the limiting magnitude of the image, which would probably vary a bit with camera temperature and such.
When I was playing around with it initially, I played with the number of concentric rings (number of bits in the number), barcode diameter (in degrees), and limiting visual magnitude (number of stars included in the database) so that each bit had about an equal probability of being a one or a zero. In denser regions along the galactic plane you of course get numbers with more ones than zeroes, and in other regions ones get pretty sparse, but it was easy to find good settings to cover just about any type of system you wanted.
Oh, I forgot another handy trick.
Suppose you’re encoding a millon star catalog into 40-bit index keys (I’d actually use a hash table), which is roughly encoding 1e6 stars in a number field that’s 1e12. For each star, flip each of the 40 bits for an alternate code to handles a bit that flipped because a star went nova, an asteroid or satellite was in the field, or a star mysteriously went out. That’s still only 40e6 entries in a number field of 1e12, so the odds of a duplicate entry are still remote, and it makes the system robust enough to handle all sorts of anomalies in the image.
For a quick stab at setting the parameters, pick how many bits you want, allow that half the bits should on average be a one (so 20 stars for the average 40-bit field), and that determines the real diameter of the barcode given a limiting visual magnitude (the number of stars in the database), or vice versa (the required limiting magnitude to give a good bit spread for the barcode diameter).
I think it’s even simpler than that George. Put a magnitude threshold on your side. Suppose your threshold were so high that only two stars in a field met it. Focus on either one and you’ve got a dupe record. Now lower the threshold to include a third star. Both of the original stars are now a unique entry in your table. You don’t need that many stars for a precise fix. Your concept is brilliant and much easier to use than even the pulsar GPS linked above.
or a star mysteriously went out
Like an occultation. Could it be used to improve efficiency in neo detection? You’ve got to figure computing power comes into play there somewhere. An improved approach seems possible to me.
Possibly. I played around with the angular or image based approaches for a while, but they can be computationally expensive (the angular methods require lots of floating point and trigonometric functions, and generally even use floating point data base searches).
The image itself can be treated as a huge binary number, but it is plagued by the image rotation problem. By centering on a star and smearing the stars into rings you reduce the problem to a one dimensional brightness curve, much like a spectrogram, and my first thought was to use that like an analog signal and use the same software techniques you’d use in spectroscopy. But the stars are narrow points and by ignoring their brightness the image could directly be reduced to a single, repeatable, unique number.
It might be faster than running a straight XOR or subtractive image comparison because those are telling you exactly where (in x y) the difference is, and you don’t need to know that until you know if there’s any meaningful difference at all.
What I would try for your application is to search the center area for a fairly bright star (x0, y0) and then scan through the image (for all x, y), computing the square of the distance from x0, y0. If dist2 is less than some cut point, add the image(x, y) to histogram(dist2). (I avoided the square root because it slows things down without improving anything.)
Then instead of a 2-d image comparison problem, you’ve got a 1-d curve comparison, and if you ignore brightness like I do for my star tracker algorithm, you’ve really just got a binary number to look up, along with all the possible flipped bits. If the non-flipped is in your database there’s nothing new, and if a bit-flipped version is in the database then the image might hold a new object.
Among other ways this should be faster is that I’d only have to find the centroid of the central star. The rest of the operation is just dumping CCD counts into bins, then summing the bins into 40 groups, then running 40 compares to generate the index number for the star lookup.
One of the other things I like about the approach is that it turns star field identification into more of a digital signals problem, with a new object acting like a dropped bit in a communications stream. Any EE has lots of ready math to characterize that problem, as opposed to coding a way around a quadrilateral with an incorrect vertex, even though that’s how a human would ponder on the problem.
Perhaps we’re so used to looking at star patterns and staring at constellations that we approached the star tracker problem that way, instead of something crazy like trying to turn the image into an audio signal where each star says its own name in its native tongue. ^_^
Kudos on this method.
Analyzing spectroscopy can -still- be a massive headache… But you really don’t care. Apply a threshold to wipe the miniscule motes and you’ve got your stbar code.
Three of these CCDs mounted orthogonally should basically be the entire physical sensor for comprehensively determining X, Y, and Z without slewing anything. And the extra CCDs are going to end up lighter than the servos anyway.
Excellent.
A great method George. Very robust because you pick a few stars to get your fix. If some don’t have hits in your DB, so what? If you have dupes, the other star fixes will rule them out. Very cool.
Technically, Dragon isn’t the only vehicle capable of bringing back cargo, since Soyuz has very limited return capability. I gather it’s about 100 kg total, it can’t be more than 28″ in diameter, and stowage volume is going to be very limited unless they return with one less human on board. Dragon is vastly more capable.
I’ve tried to post the following comment twice at PJM and it hasn’t shown up yet. Two subsequent comments showed up right away. I don’t know what the problem is. Does the word “fascism” trigger the spam filter?
Terms like “public-private partnership” and “pay to play” are pretty much euphemisms for fascism. Fascism is a form of socialism where individuals nominally own businesses, but their activities are heavily regulated and controlled by government. It has become the dominant political/economic system worldwide.
In such a situation, the government picks winners and losers. Businessmen who make political donations are essentially paying protection money. “Go harass my competitors, but leave me alone.”
Often the calculus is to donate not to who you actually support ideologically, but to who you expect to win the election. There could be negative consequences if you guess wrong. There is also a lot of hedging of bets with donations to both candidates and parties.
If everybody is playing that game, it’s almost suicidal not to join in.
I’ve had three more subsequent comments show up, and this one still hasn’t. Interesting.
Market Socialism != Fascism. In the current Market Socialist system the main difference is how the State applies the taxes it collects from the market (sales taxes, income taxes, etc).
In Fascism everyone (including the corporations) is subordinate to the State in a highly hierarchical fashion: Strikes are outlawed. Production itself is probably dictated by the state directly (i.e. instead of not giving tax breaks or subsidies if you over produce or under produce a good they will either put you out of business or incarcerate you) as well to control economic flows. In Market Socialism it is not.
You would be hard pressed to find a country today which doesn’t subsidize a given industry. It often used to be food or energy supplies. Then there was transportation and telecoms. If the state grants Bell a monopoly on telephone systems isn’t that picking winners and losers in an anti-market fashion?
The subsidies exist for a reason. If there are major supply disruptions they may cause economic problems which are difficult to recover from. We can see this today with the so called commodities crisis where goods like oil, gas, coal, metals, grains got much more expensive. I still remember in the 90s hearing about mines being closed on a weekly basis because they weren’t profitable while today it is the other way around. Speculators were buying mines around the world for peanuts and closing them. Meanwhile the country has a shortage of strategic materials. One of the reasons Hitler lost WWII was because of strategic material shortages on commodities like oil, tungsten, which the Allies had in plentiful quantities. The DoD used to subsidize such production by preferring to buy from US companies. As such criteria are relaxed any nation eventually has to face the consequences.
“The subsidies exist for a reason.”
Mostly bad reasons. They usually end up as picking favored industries and businesses at the expense of the economy as a whole.
“If there are major supply disruptions …”
I’m seeing major supply disruptions in oil recently exasperated by interference from the government in domestic production. The recent massive disruption in real estate and banking was the result in government policies subsidizing home ownership.
“The DoD used to subsidize such production by preferring to buy from US companies.”
Domestic production is less likely to be interrupted when war breaks out, and less likely to have subtle sabotage from foreign parties. Sufficient cause for such a policy.
If you don’t control your property, do you really own it? In either market socialism or fascism do you control your property, or does the state?
The recent massive disruption in real estate and banking was the result in government policies subsidizing home ownership.
Agreed. There were many tax breaks intended to increase home ownership that eventually led to the housing bubble. Besides that there were historically low interest rates and it was a lot easier to get credit. Too easy.
Oil prices are high for several reasons. One is the dearth of refining capacity. The refining infrastructure in the US, for example, is old and no one wanted to invest on it. Another is lack of crude oil. This means it is necessary to use lower grade hydrocarbons which are much more expensive to process. The US did not invest on local production because it was cheaper to buy elsewhere. The thing is we got to the point where global crude oil extraction just isn’t enough to supply the demand. Thankfully Canada did invest on tar sands processing. Otherwise we would probably have a major world war by now rather than localized conflicts (which by coincidence seem to be mostly in oil rich nations).
When I look at the current stock market I cannot help but shudder at the thought that a company like Facebook gets tens of billions of dollars for selling a sliver of the company while people who actually work in real products like A123 are worth squat in comparison. The stock market is even more bonkers than I ever presumed was possible.
Before WWII US companies worked on polymers and synthetic rubber to reduce reliance on expensive strategic material imports. I do not see it happening to the same degree today. Some things were done in the recent past like replacing copper with fiber optics in telecoms. Or the increasing use of carbon composites in aerospace, Or the shift to digital photography. That reduced the amount of strategic materials necessary to make many products. However much remains to be done and I’m not seeing it happen at any decent depth or breadth.
Regarding Solyndra (which seems to be a name thrown a lot lately) it failed for all sorts of reasons. CIGS materials got more expensive while the Chinese manufacture silicon solar cells for a pittance. There is a sort of Moore’s law for silicon PV costs and the Chinese entry into the market caused the whole predictions to shift. We are getting prices for silicon PV (W/$) that were originally estimated to only happen in the 2020s. No one expected them to cut costs to such a degree. Such things can happen with any investment. IMO giving hundreds of millions to Solyndra vs tens of billions to Facebook makes the US government seem like a prodigious spendthrift investor in comparison to the people who should know better…
When I look at the current stock market I cannot help but shudder at the thought that a company like Facebook gets tens of billions of dollars for selling a sliver of the company while people who actually work in real products like A123 are worth squat in comparison. The stock market is even more bonkers than I ever presumed was possible.
I don’t see this issue myself. Facebook fell after it came out at a particularly unrealistic price. It also had a profit of a billion dollars last year. While it’s current P/E ratio still looks unrealistic, it is at least near the ballpark for a company earning a billion dollars in a year and expected to earn more that than in the future. Keep in mind that a large part of the reason Facebook stock can stay so high is because very little of it is on the market.
A123 has the problem that it’s going to get dropkicked by Chinese industry. That depresses its stock price.
IMO giving hundreds of millions to Solyndra vs tens of billions to Facebook makes the US government seem like a prodigious spendthrift investor in comparison to the people who should know better…
No it doesn’t. Facebook is a far more serious company than the remarkably poorly run Solyndra. And I think you’d find that the people who invested in Facebook, especially after it came on the market were a combination of suckers, banks running the IPO (who were obligated to buy stock to keep the price up), or smart players operating with Other Peoples’ Money and a conflict of interest. That’s not your typical “know better” crowd.
The refineries we have are old because of NIMBY. There is almost as much Luddite opposition to new refineries as there is to nuclear power. The refineries that we have are in great shape and continue to be modernized, de-bottlenecked and upgraded on a constant basis, simply because it’s all we have to work with.
The somewhat depressing thing is, the same unimaginative little people who predicted and bemoaned “the gap” if SLS wasn’t fully or better funded, will be the same rodents who bash on and nip at SpaceX’s heals for any accidents that occur in the future. Memories of the Shuttle disasters won’t prevent them from saying that “This accident happened because some capitalist put profits ahead of safety!”.
They could use Dragon much like it is and it would be about as safe as Shuttle (if not safer). STS didn’t have a launch abort system. Yet they are working on a launch abort system and have been testing the rocket engines for it. A system like that should definitively be safer than Shuttle.
It’s already safer for at least two reasons. It’s much less complicated meaning less to go wrong. It’s shape is self stabilizing with a much better heat shield protecting a much smaller critical area.
Also, twice it’s demonstrated the ability to splash down accurately on it’s target.
And here I thought it was safer because the orbital return vehicle was placed above debris sources from the rocket itself.
That too.
My comment there:
According to information on the Vandenberg website, the minimum inclination you can launch into from there is 51 degrees. This is just slightly less than the 51.6 degree minimum inclination the Russians have been launching into from Tyuratam (Balkinour) for decades. The latitude at that site is about 46 degrees but range constraints (not wanting to drop rocket stages on China – some people have no sense of humor) limits them to 51.6 degrees, regardless of whether they’re going to the ISS or sending a satellite out to GEO. That’s why the ISS is in a 51.6 degree inclination orbit (as were the Salyuts and Mir stations).
The Russians pay a huge performance penalty for launching satellites to GEO from Tyuratam. The biggest part of the penalty is the amount of energy it takes to lower the inclination to near zero degrees. Years ago, I read in Aviation Week that their Proton booster could put 20 metric tons into LEO. It could send 2 metric tons to GEO (the upper stage took up the other 18 tons) or 5 MT to Mars.
If SpaceX tried to launch to GEO from Vandenberg, it would face similar performance penalties. Good thing the Falcon Heavy is so massively powerful. Still, it seems unlikely they’d do that because its just so inefficient. More likely, they’ll modify their launch pad at the Cape (which used to launch Titan IVs) or build a new launch site such a the proposed one near Brownsville, Texas.
The first Falcon Heavy launch pad is being built at Vandenberg. There’s a lot of speculation as to what the first payload will be. Often, first launches carry an inert test payload so you don’t risk loosing something expensive. Personally, I’d love for them to send a Dragon capsule on a looping flight around the moon. This would require some modifications to the guidance system (which depends heavily on GPS signals that wouldn’t be usable) and boosting the TT&C subsystem to handle long distance commanding and telemetry. Not only would this flight be exciting, it would give SpaceX a chance to test their heat shield on a return from deep space. It was designed for this but it still needs testing.
I don’t think such updates would be very difficult, especially since they have so much internal volume to mount communications equipment.
Does the Dragon already have a star tracker? (I think I have a much better algorithm to implement one of those, BTW)
Perhaps the biggest modification to the TT&C system would be to mount some deployable, steerable high-gain antennas on the service trunk similar to those on the Apollo CM. They might need to boost the transmitter power but not too much if the antennas are good enough. Dragons are far more automated than the Apollo capsules but I suspect their telemetry data rates are considerably higher, too.
They launched the Clementine lunar satellite in 1994 from Vandenberg and it worked well. I think a Falcon Heavy would have plenty of power to send a Dragon on a loop around the moon from Vandenberg. If they add some propellant tanks inside the Dragon, they might even be able to do an unmanned reprise of the Apollo 8 mission profile and actually enter lunar orbit.
Or they could send it to EML-2 and let it beam back picks of the moon and the earth form deep space. They could also use it as a functioning Dragonlab mission.
NASA considered easily adding automatic uploads to the Apollo GNC but decided to let the astronauts input the information as a redundancy check and to keep them in the loop.
A steerable high-gain antenna could take years to develop, unless of course SpaceX prototypes it by slapping two stepper motors or servos on a stick and mounting a Dish from Radio Shack, in which case it will take them a couple weeks to have a final system up and running. ^_^
Yeah, if SpaceX tried to buy a steerable, deployable high-gain antenna from existing contractors, it’d probably go the same as what happened when a cryogenic valve manufacturer was asked for a quote. They quoted an absurdly high price (tens of thousands of dollars, if not more). SpaceX did the valve themselves for almost nothing.
Yes it does.
THURSDAY, MAY 31, 2012
1030 GMT (6:30 a.m. EDT)
SpaceX will soon close Dragon’s navigation bay door to prepare for the ship’s re-entry. The compartment holds the spacecraft’s thermal and laser rendezvous sensors, star tracker system, and the grapple fixture used by the space station’s robotic arm.
All I can find on the tracker is a tweet from Elon Musk.
I wonder if they’re using an off-the-shelf solution like the ones from Space Micro or another vendor, or something developed in-house?
I think every tracker out there is based on picking geometric shapes out of a star field (Example) but I came up with a fast and easy way to turn a star field into a rotation-insensitive number that uniquely identifies each star (center on a star and in effect spin the image to turn the neighboring stars into a bullseye barcode), with only a few duplicate hits in a catalog of several hundred thousand stars, and all without using floating-point. If the tracker could center a star in its FOV, you could even do it with a really odd 40 or 50 pixel CCD (as concentric rings of constant area that output a zero or one) that would directly read as a database key identifying the star. For variable stars or stars on the edges of the rings, you input the alternate possibilities into the database so you get a hit either way. If you don’t get a hit you just pick a different star in the same field of view and try again.
It seemed to work well when I was playing around with it a year or so ago (with no actual hardware, just star catalog data and simulations). I was experimenting with 3/4 to 2 degree fields of view and 32-bit to 48-bit keys, for use as a small finderscope that would output high-accuracy RA and Dec as soon as you pointed it at a star field, which would be a very handy accessory for Dobson alti-azimuth telescopes. The same idea should work better in space (vastly more reliable images and no skyglow).
Fascinating. Sounds like you could build one of those for a song. I kind of like the idea of a gallactic GPS system using pulsars.
How do you deal with the non-stars? That is, things with vaguely the same apparent magnitude (Venus, Mars, ISS, whatever).
The non-stars are pretty easy, as are supernovae and other oddities. The concentric rings have a maximum real and pixel diameter (set when you chose the hardware). Let’s call that the barcode diameter, for lack of a better term.
You make the barcode diameter a little less than a third of the field of view, so that at least nine complete, non-overlapping barcodes could fit onto the image plane of the sensor, with a little slop left over because a good star won’t necessarily be in the center of these circles. Since any non-star or new star can only screw up the area within a barcode-diameter of itself, there would be many other areas on the sensor where you could still find a valid star in the catalog, unaffected by the anomaly.
With a 1 megapixel sensor this would mean the barcode diameter is a little less than 300 pixels, and you only need 40 or so rings for the barcode, so it should work out okay. More megapixels is obviously going to be better in this regard. With some though, you could probably have a bigger barcode and still ensure valid data.
For satellite use, it would be desirable to have mutliple tracker windows pointing in different directions, so that even in low orbit some sensors are always pointing into deep space, and also so that the system can ignore a camera that won’t produce a valid position without further processing to handle planets and moons (as when Jupiter and it’s moons are interfering with an entire camera’s view).
And of course you have to use mulitiple encodings for some stars because their barcode would include variable stars at the theshhold of making a ring a one or a zero, and to allow for a little bit of slop on the limiting magnitude of the image, which would probably vary a bit with camera temperature and such.
When I was playing around with it initially, I played with the number of concentric rings (number of bits in the number), barcode diameter (in degrees), and limiting visual magnitude (number of stars included in the database) so that each bit had about an equal probability of being a one or a zero. In denser regions along the galactic plane you of course get numbers with more ones than zeroes, and in other regions ones get pretty sparse, but it was easy to find good settings to cover just about any type of system you wanted.
Oh, I forgot another handy trick.
Suppose you’re encoding a millon star catalog into 40-bit index keys (I’d actually use a hash table), which is roughly encoding 1e6 stars in a number field that’s 1e12. For each star, flip each of the 40 bits for an alternate code to handles a bit that flipped because a star went nova, an asteroid or satellite was in the field, or a star mysteriously went out. That’s still only 40e6 entries in a number field of 1e12, so the odds of a duplicate entry are still remote, and it makes the system robust enough to handle all sorts of anomalies in the image.
For a quick stab at setting the parameters, pick how many bits you want, allow that half the bits should on average be a one (so 20 stars for the average 40-bit field), and that determines the real diameter of the barcode given a limiting visual magnitude (the number of stars in the database), or vice versa (the required limiting magnitude to give a good bit spread for the barcode diameter).
I think it’s even simpler than that George. Put a magnitude threshold on your side. Suppose your threshold were so high that only two stars in a field met it. Focus on either one and you’ve got a dupe record. Now lower the threshold to include a third star. Both of the original stars are now a unique entry in your table. You don’t need that many stars for a precise fix. Your concept is brilliant and much easier to use than even the pulsar GPS linked above.
or a star mysteriously went out
Like an occultation. Could it be used to improve efficiency in neo detection? You’ve got to figure computing power comes into play there somewhere. An improved approach seems possible to me.
Possibly. I played around with the angular or image based approaches for a while, but they can be computationally expensive (the angular methods require lots of floating point and trigonometric functions, and generally even use floating point data base searches).
The image itself can be treated as a huge binary number, but it is plagued by the image rotation problem. By centering on a star and smearing the stars into rings you reduce the problem to a one dimensional brightness curve, much like a spectrogram, and my first thought was to use that like an analog signal and use the same software techniques you’d use in spectroscopy. But the stars are narrow points and by ignoring their brightness the image could directly be reduced to a single, repeatable, unique number.
It might be faster than running a straight XOR or subtractive image comparison because those are telling you exactly where (in x y) the difference is, and you don’t need to know that until you know if there’s any meaningful difference at all.
What I would try for your application is to search the center area for a fairly bright star (x0, y0) and then scan through the image (for all x, y), computing the square of the distance from x0, y0. If dist2 is less than some cut point, add the image(x, y) to histogram(dist2). (I avoided the square root because it slows things down without improving anything.)
Then instead of a 2-d image comparison problem, you’ve got a 1-d curve comparison, and if you ignore brightness like I do for my star tracker algorithm, you’ve really just got a binary number to look up, along with all the possible flipped bits. If the non-flipped is in your database there’s nothing new, and if a bit-flipped version is in the database then the image might hold a new object.
Among other ways this should be faster is that I’d only have to find the centroid of the central star. The rest of the operation is just dumping CCD counts into bins, then summing the bins into 40 groups, then running 40 compares to generate the index number for the star lookup.
One of the other things I like about the approach is that it turns star field identification into more of a digital signals problem, with a new object acting like a dropped bit in a communications stream. Any EE has lots of ready math to characterize that problem, as opposed to coding a way around a quadrilateral with an incorrect vertex, even though that’s how a human would ponder on the problem.
Perhaps we’re so used to looking at star patterns and staring at constellations that we approached the star tracker problem that way, instead of something crazy like trying to turn the image into an audio signal where each star says its own name in its native tongue. ^_^
Kudos on this method.
Analyzing spectroscopy can -still- be a massive headache… But you really don’t care. Apply a threshold to wipe the miniscule motes and you’ve got your stbar code.
Three of these CCDs mounted orthogonally should basically be the entire physical sensor for comprehensively determining X, Y, and Z without slewing anything. And the extra CCDs are going to end up lighter than the servos anyway.
Excellent.
A great method George. Very robust because you pick a few stars to get your fix. If some don’t have hits in your DB, so what? If you have dupes, the other star fixes will rule them out. Very cool.
Technically, Dragon isn’t the only vehicle capable of bringing back cargo, since Soyuz has very limited return capability. I gather it’s about 100 kg total, it can’t be more than 28″ in diameter, and stowage volume is going to be very limited unless they return with one less human on board. Dragon is vastly more capable.
I’ve tried to post the following comment twice at PJM and it hasn’t shown up yet. Two subsequent comments showed up right away. I don’t know what the problem is. Does the word “fascism” trigger the spam filter?
Terms like “public-private partnership” and “pay to play” are pretty much euphemisms for fascism. Fascism is a form of socialism where individuals nominally own businesses, but their activities are heavily regulated and controlled by government. It has become the dominant political/economic system worldwide.
In such a situation, the government picks winners and losers. Businessmen who make political donations are essentially paying protection money. “Go harass my competitors, but leave me alone.”
Often the calculus is to donate not to who you actually support ideologically, but to who you expect to win the election. There could be negative consequences if you guess wrong. There is also a lot of hedging of bets with donations to both candidates and parties.
If everybody is playing that game, it’s almost suicidal not to join in.
I’ve had three more subsequent comments show up, and this one still hasn’t. Interesting.
Market Socialism != Fascism. In the current Market Socialist system the main difference is how the State applies the taxes it collects from the market (sales taxes, income taxes, etc).
In Fascism everyone (including the corporations) is subordinate to the State in a highly hierarchical fashion: Strikes are outlawed. Production itself is probably dictated by the state directly (i.e. instead of not giving tax breaks or subsidies if you over produce or under produce a good they will either put you out of business or incarcerate you) as well to control economic flows. In Market Socialism it is not.
You would be hard pressed to find a country today which doesn’t subsidize a given industry. It often used to be food or energy supplies. Then there was transportation and telecoms. If the state grants Bell a monopoly on telephone systems isn’t that picking winners and losers in an anti-market fashion?
The subsidies exist for a reason. If there are major supply disruptions they may cause economic problems which are difficult to recover from. We can see this today with the so called commodities crisis where goods like oil, gas, coal, metals, grains got much more expensive. I still remember in the 90s hearing about mines being closed on a weekly basis because they weren’t profitable while today it is the other way around. Speculators were buying mines around the world for peanuts and closing them. Meanwhile the country has a shortage of strategic materials. One of the reasons Hitler lost WWII was because of strategic material shortages on commodities like oil, tungsten, which the Allies had in plentiful quantities. The DoD used to subsidize such production by preferring to buy from US companies. As such criteria are relaxed any nation eventually has to face the consequences.
“The subsidies exist for a reason.”
Mostly bad reasons. They usually end up as picking favored industries and businesses at the expense of the economy as a whole.
“If there are major supply disruptions …”
I’m seeing major supply disruptions in oil recently exasperated by interference from the government in domestic production. The recent massive disruption in real estate and banking was the result in government policies subsidizing home ownership.
“The DoD used to subsidize such production by preferring to buy from US companies.”
Domestic production is less likely to be interrupted when war breaks out, and less likely to have subtle sabotage from foreign parties. Sufficient cause for such a policy.
If you don’t control your property, do you really own it? In either market socialism or fascism do you control your property, or does the state?
The recent massive disruption in real estate and banking was the result in government policies subsidizing home ownership.
Agreed. There were many tax breaks intended to increase home ownership that eventually led to the housing bubble. Besides that there were historically low interest rates and it was a lot easier to get credit. Too easy.
Oil prices are high for several reasons. One is the dearth of refining capacity. The refining infrastructure in the US, for example, is old and no one wanted to invest on it. Another is lack of crude oil. This means it is necessary to use lower grade hydrocarbons which are much more expensive to process. The US did not invest on local production because it was cheaper to buy elsewhere. The thing is we got to the point where global crude oil extraction just isn’t enough to supply the demand. Thankfully Canada did invest on tar sands processing. Otherwise we would probably have a major world war by now rather than localized conflicts (which by coincidence seem to be mostly in oil rich nations).
When I look at the current stock market I cannot help but shudder at the thought that a company like Facebook gets tens of billions of dollars for selling a sliver of the company while people who actually work in real products like A123 are worth squat in comparison. The stock market is even more bonkers than I ever presumed was possible.
Before WWII US companies worked on polymers and synthetic rubber to reduce reliance on expensive strategic material imports. I do not see it happening to the same degree today. Some things were done in the recent past like replacing copper with fiber optics in telecoms. Or the increasing use of carbon composites in aerospace, Or the shift to digital photography. That reduced the amount of strategic materials necessary to make many products. However much remains to be done and I’m not seeing it happen at any decent depth or breadth.
Regarding Solyndra (which seems to be a name thrown a lot lately) it failed for all sorts of reasons. CIGS materials got more expensive while the Chinese manufacture silicon solar cells for a pittance. There is a sort of Moore’s law for silicon PV costs and the Chinese entry into the market caused the whole predictions to shift. We are getting prices for silicon PV (W/$) that were originally estimated to only happen in the 2020s. No one expected them to cut costs to such a degree. Such things can happen with any investment. IMO giving hundreds of millions to Solyndra vs tens of billions to Facebook makes the US government seem like a prodigious spendthrift investor in comparison to the people who should know better…
When I look at the current stock market I cannot help but shudder at the thought that a company like Facebook gets tens of billions of dollars for selling a sliver of the company while people who actually work in real products like A123 are worth squat in comparison. The stock market is even more bonkers than I ever presumed was possible.
I don’t see this issue myself. Facebook fell after it came out at a particularly unrealistic price. It also had a profit of a billion dollars last year. While it’s current P/E ratio still looks unrealistic, it is at least near the ballpark for a company earning a billion dollars in a year and expected to earn more that than in the future. Keep in mind that a large part of the reason Facebook stock can stay so high is because very little of it is on the market.
A123 has the problem that it’s going to get dropkicked by Chinese industry. That depresses its stock price.
IMO giving hundreds of millions to Solyndra vs tens of billions to Facebook makes the US government seem like a prodigious spendthrift investor in comparison to the people who should know better…
No it doesn’t. Facebook is a far more serious company than the remarkably poorly run Solyndra. And I think you’d find that the people who invested in Facebook, especially after it came on the market were a combination of suckers, banks running the IPO (who were obligated to buy stock to keep the price up), or smart players operating with Other Peoples’ Money and a conflict of interest. That’s not your typical “know better” crowd.
The refineries we have are old because of NIMBY. There is almost as much Luddite opposition to new refineries as there is to nuclear power. The refineries that we have are in great shape and continue to be modernized, de-bottlenecked and upgraded on a constant basis, simply because it’s all we have to work with.