Careful who you call ignorant there. I realize that there are plenty of other OS's out there. I use QNX once in a while for realtime development, and on my page you can find a bootable image of an OS that I'm writing myself. I had an old DOS machine for microcontroller stuff that I only recently migrated to Linux.
Besides, I only once mentioned Linux specifically, and in passing. I mostly meant tinker-friendly systems in general.
All that I'm saying is that the same spirit that founded Open/Free software has roots in the old-fashioned HAM community. Sharing of designs, ideas, and often even parts were common, and the will to tinker was high.
I was just pointing out that it's not exactly surprising that a community formed on the same foundations as the various open-*nix communities would favor open platforms.
What a wonderful idea! Let's poll a bunch of people who are, in large part, ELECTRONICS HOBBYESTS, and ask them what OS they prefer!
HAM radio types are often some of the most technically knowledgable in the world. I'll bet that even the ones that still use Windows know enough Linux to survive. They may even have legit reasons for using Windows (like certain Windows-based EDA software still beats anything on Linux...)
I have another great idea. Let's poll KERNEL HACKERS and ask what OS they run:P
QNX is a hard realtime embedded operating system. Just because they developed a full GUI system for it doesn't change this. SCADA and similar control applications, even hard ones, are perfectly suited for QNX.
I haven't done anything as critical as SCADA, but I've done some PC/104 with QNX. It's a nice system, for what it does.
This "blanket statement" is more of a best practice than a strict necessity, but it is true that floating-point numbers lose an amount of accuracy (and not even an easily compensated-for amount either) each time an operation is made using them.
Fixed-point math loses accuracy as well, but in a well-defined way that is easy to account for in the algorithm design.
This may not be an enormous risk on a dataset that is calculated once and discarded (like a user display), but if the data is to be stored and manipulated, accumulated round-off errors can add up quickly.
The other real worry with floating-point numbers is that the range they express is such that some operations (like adding) will lose an operand entirely due to problems with the exponents being drastically mismatched.
This doesn't surprise me in the slightest, and it's not as bad as it sounds, either.
8-bit processors still dominate the CPU market in terms of volume, and very nearly in terms of profitability. They are virtually never used as general-purpose computers anymore, but due to low cost of development, deployment and testing, they are ubiquitous in the control systems industry.
Companies like Atmel and Microchip are constantly devising new and better 8-bit microcontroller chips for this market. A lot of them are available in hardened grades for just these uses. A modern one will often bundle the entire machine onto a single chip, with as much IO and analog interfacing as you could ask for.
Reading the ENTIRE assembly dump of a 32K program is rather simple. A team of a dozen engineers can verify it in a matter of a couple months (I mean formal verification here, like you would do for a truly critical system, not just "give it a look over").
While truly using a BBC micro is a little obsolescant, the ideals that caused them to do so are sound.
Of course Amiga is hard to emulate. The machine had a hardware coprocessor for *EVERYTHING*. It had a built-in sound mixer chip, a heavily accelerated graphics chip, and a myriad of others.
Emulating the CPU alone is easy. But even an older system, when employing a lot of coprocessing, can be quite a task to emulate by a strictly serial processor.
We are attacking this on all fronts. Most people focus on reducing feature size, which is a useful thing but isn't everything. It lets us pack more active devices into the same area, so we can get the same yield from chips with more transistors, and it increases speed and decreases power consumption.
Unfortunately, power consumption (and hence dissipation (heat)) is proportional to the clock frequency, so the faster it goes, the hotter it runs.
Semiconductor characteristics are important. Things like how much capacitance there is across an active device make an enormous difference (dynamic power dissipation, which determines the chip's temperature, is proportional to the capacitance of the gates), and silicon is nowhere near the best at this. We already have 10GHz (digital) chips running in Gallium Arsenide or Indium Phosphide processes. Diamond is supposed to be one of the most promising new materials in this area.
Clock distribution is becoming a huge problem, but it's not related to the materials or production capabilities. It's just a limitation of keeping billions of clock signals synchronized across a chip at speeds that cause even the small metal wires on the IC to behave like transmission-lines. On newer chips, up to HALF of the area on the chip is devoted to synchronizing the clock. Advances in asynchronous design are starting to overcome some of these limitations, but that is still a ways away from mainstream.
Finally, temperature. Performance is inversely proportional to the temperature (since, as the temperature rises, so does the propogation delay, and hence the chip has to run at a lower frequency). Since diamond has such a high thermal conductivity, it is possible to mount the wafer upside-down inside the plastic body, with a block of copper (or possibly even more diamond, when the prices drop that much) to draw the heat directly from the core. It's not the ability of diamond to run at higher temperatures that is the important part (although accidents happen, and it's nice to be able to heat your CPU to 300 without killing it). It's the higher thermal conductivity, which makes it easier to dissipate the heat.
Moissanite has the second-highest hardness of any material, at something like a 9.7 (if I remember correctly). For reference, corrundum (sapphire, ruby) is 9, and diamond is 10.
After all, this is silicon-carbide. We use this stuff as an industrial abrasive for a reason.
Re:The reverse IS true!
on
The Diamond Age
·
· Score: 2, Informative
Hrm... While they do have a bit of a yellow or grey tint to them, the refractive index is superior to diamond, and far superior to CZ.
You know, if it weren't for the fact that you were responding to me, I would have modded you up *BEFORE* flaming you.
I'm well aware of the differences in the actual processes, although it's not as simple as it once was either. Thanks to SOI, epitaxy is being used more and more to grow the actual semiconductor layers. In this way, the two processes are strikingly similar.
They won't be producable from the same fab, but the fab to do this shouldn't have to be any more complex.
I admit a little bit of skepticism about their method for growing large crystals, but it sounds like they already have that part under control, if you take their word for it.
Besides, the other method mentioned in the article (you did read it, right?) was a little closer to the way it's done with silicon, although the volumes and qualities achievable this way may be a little bit too low for semiconductor use.
Re:If geeks had girlfriends...
on
The Diamond Age
·
· Score: 1
You sure know how to make a girl swoon. Yummy supercomputing goodness.
The reverse IS true!
on
The Diamond Age
·
· Score: 4, Interesting
The funny part is that silicon carbide crystal, which started off as a semiconductor (especially for use in blue LED's), was later marketed as Moissanite, a gemstone with superior lustre to diamonds.
If you ever see a top white diamond next to a Moissanite, you'd swear the diamond was glass. The Moissanite is almost blinding.
I've never actually read the book, so I might be wrong on this, but I'd suspect that it was literal.
See, the earliest nanites (which we have almost developed now, in fact) are what are called diamandoid nanites... They are made of pure carbon in a lattice. So nanites==diamonds, at least for the beginning of the nanite age.
Read the fine article. These diamonds are grown in much the same way as a silicon wafer. The processes involved don't sound particularly more expensive, and the materials involved are simple methane and hydrogen.
They are listing numbers like $5/carat (1ct should be enough to make a processor chip... certainly 2 or 3 cts is).
If anything, this might actually be cheaper than silicon by the MHz, thanks to its superior semiconductor and insulating properties and higher thermal conductivity.
Gemstones as investments.
on
The Diamond Age
·
· Score: 5, Interesting
Gemstones make *AWEFUL* investments. Changes in the market can cause the loss of the value of anything you have, and seldom do they increase in value.
Diamond is the only gem that's still worth anything (thanks to De Boer's monopoly). With the advent of the internet, virtually anyone can order other gems directly from Thailand and the like. Sapphire and ruby prices have crashed as a result. You can get a 1 carat pigeonblood ruby for just $10 or so nowadays.
And that's not counting advances in synthetic gemstones. Hydrothermal processes for sapphire, ruby and emerald have made it virtually impossible to detect a good quality gem (most synthetic sapphire and ruby is still grown the old way though, which is easier to detect).
I personally have a roughly 10+ carat white sapphire heart and a top blood red ruby of about the same size, both synthetic. I paid about $10 each for them, including the.925 sterling silver pendant setting.
In context, natural gems like these, a few decades back, would be worth tens of thousands of dollars.
Controversial issues? These are on the table NOW up here in Canada. The courts have ruled that gay marriage is the only fair solution, so government can't really do anything about it, and the laws against marijuana possession are no longer in force (I believe this is due to another court ruling, but I'm not really sure... I don't keep up on marijana news).
And we have real healthcare up here too. I think we should let this woman know that if she fails her bid in the USA, she should run in Canada. Her platform would be very popular up here.
There are ways to overcome this problem. Between using an aerodynamic shape for the gasbag and larger fans, a miniblimp is quite capable of outdoor operation.
The one that I really like is the internal-rotor helicopter-style design. I actually helped someone test some ideas for one of these. We took a plastic toy propeller (from those pull-string toys), and attached it to a Dremel rotary tool. Fire the thing up to 15,000 RPM, and it rapidly lifted the tool, which was being held down by the AC cord.
If we had used a cordless version, it probably would have been capable of independant flight.
The real problem came from the fact that these little flimsy plastic rotors aren't meant for any more than, say, 1000RPM. After 20 or 30 seconds of flight, it promptly exploded.
I saw this nifty R/C blimp at a tradeshow once. It was about 3 feet long. It could be moved in all three axis, and was perfectly content hovering as well. It broadcasted an image from an underbelly camera to a standard TV channel.
It was rather neat watching the blimp flying around the auditorium and spying on things from the air.
Ever since I saw the thing, I've been wanting to build one.
It's funny how they describe introversion. I can understand it readily, as I'm definately introverted. A lot of these things are also signs of autism though (I have high-function autism myself...).
The difficulties with large, noisy groups, dealing poorly with interruptions, the intensity of focus required to just sit for hours and do one thing, these are all autistic traits. Are autism and introversion on the same scale?
Actually, even modern printers can take considerable time for the postscript. It's just a matter of what you print. Regular text/bitmaps won't cause problems.
If you, like I occasionally do, like to print engineering drawings/draftings (which heavily use PostScript's more processing-intensive features), then you may have to wait a little longer on the spool.
This question is covered in any intro engineering law course, and in the government brochures on IP law. (Is this possibly just a canadian thing? I somehow doubt that...)
The layout of the traces isn't just determined by its function, or we would have autorouters that work by now (yes, I know, they work sometimes...).
The PC-board artwork refers specifically to the *SPECIFIC* design on the board. An identically netlisted board (which would look very, very similar but not identical) is not covered by the copyright on the board artwork.
It's perfectly legal to copy the electrical design of someone's product. It's just not legal to desolder all the components and *PHOTOCOPY* the board.
PC Board *DESIGNS* are not copyrightable... However, PC board *ARTWORKS* are.
This means that you can very legally draw a netlist from someone's board, and make a new PC-board layout that's electrically identical, but you can't verbatim copy it.
The odd part about this is that, while I can go buy a TV and do almost anything I want with it, I can't buy it, dismantle it, copy it, and give the copy to a friend legally.
What, you say? I can't?
Nope. Printed Circuit Board artwork is copyrighted material. I might be able to mangle or modify the TV, but the minute I copy it, it's illegal.
Of course, if I lifted the schematic diagram from the PCB and made my own layout, it would again be legal. What a weird world.
Actually, the problem with this is that the pressure is not enough to balance gravity. Gravity is balanced by the inertia of orbitting.
The standard view of navigation doesn't help much when dealing with gravity wells. It's simpler just to think of the orbit around the sun as having a certain "orbital energy". The lower the orbital energy, the closer you are to the sun. This energy is a combination of potential (how far from the sun you are) and kinetic (how fast you are moving).
If you are heading directly away from the sun, you will quickly be pulled straight towards it. If you can avoid being crisped, you'll end up in an extremely elliptical orbit.
To have a non-orbital path that would not fall into the sun would require the craft to have escape energy (be moving at the escape velocity of the sun). Orbiting allows the orbital energy to be slowly increased.
Slashdot Mirror
Besides, I only once mentioned Linux specifically, and in passing. I mostly meant tinker-friendly systems in general.
All that I'm saying is that the same spirit that founded Open/Free software has roots in the old-fashioned HAM community. Sharing of designs, ideas, and often even parts were common, and the will to tinker was high.
I was just pointing out that it's not exactly surprising that a community formed on the same foundations as the various open-*nix communities would favor open platforms.
HAM radio types are often some of the most technically knowledgable in the world. I'll bet that even the ones that still use Windows know enough Linux to survive. They may even have legit reasons for using Windows (like certain Windows-based EDA software still beats anything on Linux...)
I have another great idea. Let's poll KERNEL HACKERS and ask what OS they run :P
I haven't done anything as critical as SCADA, but I've done some PC/104 with QNX. It's a nice system, for what it does.
Fixed-point math loses accuracy as well, but in a well-defined way that is easy to account for in the algorithm design.
This may not be an enormous risk on a dataset that is calculated once and discarded (like a user display), but if the data is to be stored and manipulated, accumulated round-off errors can add up quickly.
The other real worry with floating-point numbers is that the range they express is such that some operations (like adding) will lose an operand entirely due to problems with the exponents being drastically mismatched.
8-bit processors still dominate the CPU market in terms of volume, and very nearly in terms of profitability. They are virtually never used as general-purpose computers anymore, but due to low cost of development, deployment and testing, they are ubiquitous in the control systems industry.
Companies like Atmel and Microchip are constantly devising new and better 8-bit microcontroller chips for this market. A lot of them are available in hardened grades for just these uses. A modern one will often bundle the entire machine onto a single chip, with as much IO and analog interfacing as you could ask for.
Reading the ENTIRE assembly dump of a 32K program is rather simple. A team of a dozen engineers can verify it in a matter of a couple months (I mean formal verification here, like you would do for a truly critical system, not just "give it a look over").
While truly using a BBC micro is a little obsolescant, the ideals that caused them to do so are sound.
Emulating the CPU alone is easy. But even an older system, when employing a lot of coprocessing, can be quite a task to emulate by a strictly serial processor.
- Feature Size
- Semiconductor Characteristics (bandgap, intrinsic capacitance etc)
- Clock Distribution
- Temperature
We are attacking this on all fronts. Most people focus on reducing feature size, which is a useful thing but isn't everything. It lets us pack more active devices into the same area, so we can get the same yield from chips with more transistors, and it increases speed and decreases power consumption.Unfortunately, power consumption (and hence dissipation (heat)) is proportional to the clock frequency, so the faster it goes, the hotter it runs.
Semiconductor characteristics are important. Things like how much capacitance there is across an active device make an enormous difference (dynamic power dissipation, which determines the chip's temperature, is proportional to the capacitance of the gates), and silicon is nowhere near the best at this. We already have 10GHz (digital) chips running in Gallium Arsenide or Indium Phosphide processes. Diamond is supposed to be one of the most promising new materials in this area.
Clock distribution is becoming a huge problem, but it's not related to the materials or production capabilities. It's just a limitation of keeping billions of clock signals synchronized across a chip at speeds that cause even the small metal wires on the IC to behave like transmission-lines. On newer chips, up to HALF of the area on the chip is devoted to synchronizing the clock. Advances in asynchronous design are starting to overcome some of these limitations, but that is still a ways away from mainstream.
Finally, temperature. Performance is inversely proportional to the temperature (since, as the temperature rises, so does the propogation delay, and hence the chip has to run at a lower frequency). Since diamond has such a high thermal conductivity, it is possible to mount the wafer upside-down inside the plastic body, with a block of copper (or possibly even more diamond, when the prices drop that much) to draw the heat directly from the core. It's not the ability of diamond to run at higher temperatures that is the important part (although accidents happen, and it's nice to be able to heat your CPU to 300 without killing it). It's the higher thermal conductivity, which makes it easier to dissipate the heat.
After all, this is silicon-carbide. We use this stuff as an industrial abrasive for a reason.
Hrm... While they do have a bit of a yellow or grey tint to them, the refractive index is superior to diamond, and far superior to CZ.
I'm well aware of the differences in the actual processes, although it's not as simple as it once was either. Thanks to SOI, epitaxy is being used more and more to grow the actual semiconductor layers. In this way, the two processes are strikingly similar.
They won't be producable from the same fab, but the fab to do this shouldn't have to be any more complex.
I admit a little bit of skepticism about their method for growing large crystals, but it sounds like they already have that part under control, if you take their word for it.
Besides, the other method mentioned in the article (you did read it, right?) was a little closer to the way it's done with silicon, although the volumes and qualities achievable this way may be a little bit too low for semiconductor use.
You sure know how to make a girl swoon. Yummy supercomputing goodness.
If you ever see a top white diamond next to a Moissanite, you'd swear the diamond was glass. The Moissanite is almost blinding.
See, the earliest nanites (which we have almost developed now, in fact) are what are called diamandoid nanites... They are made of pure carbon in a lattice. So nanites==diamonds, at least for the beginning of the nanite age.
They are listing numbers like $5/carat (1ct should be enough to make a processor chip... certainly 2 or 3 cts is).
If anything, this might actually be cheaper than silicon by the MHz, thanks to its superior semiconductor and insulating properties and higher thermal conductivity.
Diamond is the only gem that's still worth anything (thanks to De Boer's monopoly). With the advent of the internet, virtually anyone can order other gems directly from Thailand and the like. Sapphire and ruby prices have crashed as a result. You can get a 1 carat pigeonblood ruby for just $10 or so nowadays.
And that's not counting advances in synthetic gemstones. Hydrothermal processes for sapphire, ruby and emerald have made it virtually impossible to detect a good quality gem (most synthetic sapphire and ruby is still grown the old way though, which is easier to detect).
I personally have a roughly 10+ carat white sapphire heart and a top blood red ruby of about the same size, both synthetic. I paid about $10 each for them, including the .925 sterling silver pendant setting.
In context, natural gems like these, a few decades back, would be worth tens of thousands of dollars.
And we have real healthcare up here too. I think we should let this woman know that if she fails her bid in the USA, she should run in Canada. Her platform would be very popular up here.
The one that I really like is the internal-rotor helicopter-style design. I actually helped someone test some ideas for one of these. We took a plastic toy propeller (from those pull-string toys), and attached it to a Dremel rotary tool. Fire the thing up to 15,000 RPM, and it rapidly lifted the tool, which was being held down by the AC cord.
If we had used a cordless version, it probably would have been capable of independant flight.
The real problem came from the fact that these little flimsy plastic rotors aren't meant for any more than, say, 1000RPM. After 20 or 30 seconds of flight, it promptly exploded.
It was rather neat watching the blimp flying around the auditorium and spying on things from the air.
Ever since I saw the thing, I've been wanting to build one.
True... But we can learn how to do so. It's just not instinctual. We can also interact with other autistics without trouble.
The difficulties with large, noisy groups, dealing poorly with interruptions, the intensity of focus required to just sit for hours and do one thing, these are all autistic traits. Are autism and introversion on the same scale?
If you, like I occasionally do, like to print engineering drawings/draftings (which heavily use PostScript's more processing-intensive features), then you may have to wait a little longer on the spool.
The layout of the traces isn't just determined by its function, or we would have autorouters that work by now (yes, I know, they work sometimes...).
The PC-board artwork refers specifically to the *SPECIFIC* design on the board. An identically netlisted board (which would look very, very similar but not identical) is not covered by the copyright on the board artwork.
It's perfectly legal to copy the electrical design of someone's product. It's just not legal to desolder all the components and *PHOTOCOPY* the board.
This means that you can very legally draw a netlist from someone's board, and make a new PC-board layout that's electrically identical, but you can't verbatim copy it.
What, you say? I can't?
Nope. Printed Circuit Board artwork is copyrighted material. I might be able to mangle or modify the TV, but the minute I copy it, it's illegal.
Of course, if I lifted the schematic diagram from the PCB and made my own layout, it would again be legal. What a weird world.
The standard view of navigation doesn't help much when dealing with gravity wells. It's simpler just to think of the orbit around the sun as having a certain "orbital energy". The lower the orbital energy, the closer you are to the sun. This energy is a combination of potential (how far from the sun you are) and kinetic (how fast you are moving).
If you are heading directly away from the sun, you will quickly be pulled straight towards it. If you can avoid being crisped, you'll end up in an extremely elliptical orbit.
To have a non-orbital path that would not fall into the sun would require the craft to have escape energy (be moving at the escape velocity of the sun). Orbiting allows the orbital energy to be slowly increased.