DMT DSL (which is the DSL modulation used pretty much everywhere) is most definitely a modulated analog signal. DMT is 256 concurrent narrow-band QAM (Quadrature amplitude modulation) channels with variable constellation sizes (up to 15 bit constellations) with a 4KHz (data-)symbol rate. Cable modems use a wide-band QAM channel with a fixed constellation size and a much higher symbol rate.
Building any QAM device without a DAC/ADC and a DSP is a hopeless task. It is most definitely a modem.
QAM, BTW, is what was used for V.34 (33.6Kb/s) voice modems (more-or-less, anyway - there some pretty major differences in implementation but the idea is the same)
"All the world is *not* a PC. All programs are *not* applications. Some programemrs have *zero interest* in writing code that isn't assembly language for small systems, to them C code and operating systems just get in the way"
I happen to be one of these people. Applications hold very little interest for me. I like programming hyper-optimized aeembly for DSPs, data framers, network aggregators, embedded control systems, etc. To me, if I'm not programming to the bare-metal, its an uncomfortable experience.
Needless to say, there isn't a whole lot of hobbiest stuff to do in these areas within the means of anybody who doesn't have a trust-fund and tons of spare time. Just to do my job effectively at work, I have over $300K of scopes and analyzers of various sorts at my bench - needless to say, open source is not gonna happen.
Well, at my company several engineers interview each candidate and everyone has there own interview style.
For myself, I like to pick a DSP subject on their resume that I only have a passing familiarity with and teach me as much as possible about the subject during our hour or so together. Filter design is my favorite, as I do very little of it myself, but I know enough to be able to pick out bullshit. Generally, I look for how much I learn and how enthusiastically its presented - enthusiastic, competent people that can explain complex subjects to others, well those people rock:)
But, in the end, you're right. I may have gotten bad vibe off of some competent people because they weren't as hyper as I am about the subject. On the otherhand, I don't know if I want to work with people who don't totally dig their work...
According to the G.DMT standard (for all annexes), 8 Mbs is required to be considered standard compliant.
However, the standard allows optionally reconfiguring the Reed Solomon codec to S=1/2 (that is, 2 RS codewords per data symbol) which means support up to 16 Mbs. Of course, the actual channel capacity will top out somewhere in the neighborhood of 12-13 Mbs (assuming a short, clean loop) due to Shannon's Law and the realities of current DAC and signal processing limitations. But really, anywhere up to about 7 Kft, 10 Mbs should be no problem assuming the line is reasonably clean.
S=1/2 is a bit tricky to implement and while most vendors have it, it tends not to be interoperable with the implementations of other vendors, so its not widespread since in order to work your chipset at home would have to be the same brand as the one at the CO. In Japan, this is being solved by the LECs standardizing, so 12 Mbs is being rolled out there.
When you're learning the math, I agree, calculators will inhibit your learning to some degree. If you're learning to do something with the math, like in an engineering or physics course, perhaps working through it a few times could be useful. But, when you're trying to actually accomplish a useful goal with the math, its virutally always better to let the computer do it for you.
I work as a DSP engineer and I honestly don't know if I could so a Fourier transform anymore without a boatload of reference material. I just don't deal with the math at that level - I work with people who are much much better at it than I. To me, to getting filter coefficients from time-domain to frequency-domain is going to Matlab and typing:
load stuff plot (20*log10(fft(stuff, 512)))
Now, my complaint about my education is that learning to do cool stuff with Matlab or Mathematica or a calculator was actively discouraged - everything was to be worked out by hand. I just don't have the time for that now, so Matlab it is, but I feel that my horizons could've been expanded (so to speak) by using the most effective tools while I was still in college and learning to use them to their full potential by solving much more interesting problems with them, rather than 'here, solve this one with Maple/Matlab/Mathematica but do the rest of the drudgery by hand'. Rather, I feel that, for instance, my signals courses (besides the first one, where the math is important) could've covered a lot more ground by explaining the concepts and being Matlab-centric on the implementation. In the real world, the guys who really, really know FFT algorithms well tend to be math-majors or other hard-on-for-numbers types anyway, so they'd probably would've learned it regardless of the tools, but us schmoes could've been saved the headache and learned something useful to boot.
I work for a chip company that does somewhat mass-market chips (around 2 million various chips per quarter), and based on what I know (I don't work much with the manufacturing side, but do deal with LSI designers who bring up their concerns), here's some back of the envelope calculations:
Our single-core chips cost around $3 each for a good chip. Die size is around 80 mm^2, yield is ~70%, process is.18micron.
Intel processors are ~400mm^2, so they can get at best 1/5 the chips on the same size wafer. This brings cost to $15 using my starting point. Their yield would be 0.70^5 or 16% - but lets assume that since Intel runs their own foundry, they have better kung-fu than the fabs we use, so maybe 35% yield. That brings us to $30, and I'm talking out my ass at this point.
However, if they didn't have the improved yields I was assuming, you're looking at $60/chip or more just in fabbing cost. Semiconductor companies tend to like at least 60% gross margins, so a $60 chip would sell for $150 minimum, not counting cost of sales.
Heh, I dropped out of college in June 2000 with 17 credits left for a physics degree and 15 for a compsci degree.
Best decision I ever made:)
In June 2000, the hiring boom was still in progress, and I got $60K as a driver programmer after submitting a small resume and a space-invaders game I wrote in 80x86 assembly language. I still have that job, and I make considerably more now. Getting my foot in the door when the gettin' was good was what mattered, and it worked out beautifully.
I no longer work on drivers, I do DSP and network processor programming stuff now. And I love it. And I have no fears of being laid-off - my company has tons of cash and is profitable.
All-in-all, great decision, though I was derided by my parents and lost my girlfriend over it, but it was well worth it. (I've since made-up with my parents and have a new, better gf, so everything worked out there too)
I can't imagine how hard that first job would've been in 2001 for me with my 2.0 GPA. I feel for ya.
Well, I've seen places where management took the brunt of it.
Several years ago, I interned doing IT work at a foundry that was very dependent on the defense industry. Needless to say, it wasn't the roaring 80's for them anymore.
As a result, there were approximately ~20% layoffs of floor-workers, but NONE of the people in director positions or higher were paid a cent for six months. They were effectively laid off, but still expected to show up for work.
Its not always the big-bad management out to get people.
If its just synchronising, why not just run a wire from light-to-light with a reference tone on it and have each timer phase-lock to that? That would eliminate any drift and probably be a heck of a lot cheaper.
If they were doing any sort of analysis, though, I could see why they'd need PC's.
Assuming you wanted the most basic hardware possible, a WinWiFi would be an Antenna, an A-D/D-A and some (somewhat expensive, hardware wise)mod/demod circuitry. The problem with wireless is that the frequencies are way too high to deal with as baseband (you'd have to process 5 billion samples/second to do baseband processing at 2.5GHz), so some hardware is needed to bring the signal to a useable range.
However, I don't think that anybody is going to do that since even after bringing it to baseband, there are still way too many samples to deal with efficiently on a PC.
Rather, I'm under the impression that the hardware will have the analog bits, a DSP, and a basic ethernet framer and let the PC take care of the rest of the ethernet MAC functionality.
Tim
Re:Shooting them down? I think there is a treaty..
on
Space Wars
·
· Score: 1
Yup, and if you could read you'd see that the treaty gives a provision for either party to withdraw with 6 months notice, which the US gave.
Anyway, I seem to remember from QED that electrons exist in the same place at the same time with alarming regularity. It seems that subatmoic particles move sorta-freely in time at very small scales, but their net time vector is "forwards". I think this guy just wants to induce it to happen on a large scale (neutrons time-traveling distances greater than femtoseconds)
Of course, subatomic particles also break the 2nd law of thermodynamics for very short periods of time, but that doesn't mean it can be done on any useful scale or induced to happen at will.
"One does tend to ask oneself why the world's largest military machine was incapable of developing something like this itself until recently"
Most likely, for two reasons:
1) US air-combat stratgies focus on "beyond-visual-range" enagement. Thus, the research focus has been on better long range radars and tracking systems.
2) The US hasn't put a new fighter-plane into service since the early 80's (F-16 and F-18). And the focus on those two planes was maneuvarability and reduced-cost, not an opportunity to try out new tech. Before that, the F-14 and F-15 were in service by the mid-70s and are still the premier fighter planes in service.
The F-22, being the first new fighter-plane the US has made in a long, long time is the first opportunity to add this type of tech.
Those 100 mile range missiles like the Phoenix are pretty unlikely to hit a competent pilot from any significant distance. All they'd need to do is turn perpindicular to the missile (beaming the radar) and it would most likely be avoided.
The point behind a phoenix missile is to present a credible threat at a distance, forcing the enemy to take these evasive manuveurs (and thus sacrificing tactical advantage) while the F-14 quickly closes the distance and makes a much shorter range kill.
Damn, an S4 weighs that much? Honestly, I never looked up the weight, but just assumed that since it was about the same size as my 323i, it would weigh roughly the same (~2300 lbs), plus a little added weight for AWD and a bigger engine. Where does the extra weight come from? (the engine and AWD can't account for 1500 lbs? or can they?)
Anyway, I don't think it would be reliable at 400 HP. The extra stress on the pushrods alone I would think would significantly shorten the lifespan of the engine. Yeah, replacing a snapped rod isn't that big of a deal, but its still a pain in the ass if you're somewhere inconvenient. Not to mention the other parts that would be running out of spec.
As for Ferraris, I'm not arguing with you. YOu basically made my point: They are engineered with performance and performance alone in mind. If you build a car this way, its just not going to be reliable, but it will be damn fast. Taking a reliable car and trying to turn it into a ferrari-beater is going to give you the same adverse effects on reliability.
In general straight line, though the newest Z06 did pull higher on the skidpad (dry) than any car I've ever seen, well over 1G.
On the wet skidpad, well....
Anyway, a Corvette is definitely a luxury sports car, maybe not quite as nice as a BMW, but quite plush inside and fast and good handling (as long as it doesn't rain).
Good lord dude. What the hell do you need with 400 HP on something as small as an S4? Like changing tires much? How about cylinder heads? I heard those were real cheap now...
sigh
The high HP mainstream luxury sports cars (S4, M3/M5, 911, Corvette, etc), in general are limited at the power that they get because a) Its damn fast as is and b) its actually reliable at those HP/Torque numbers.
There's a very good reason why $150,000 Ferraris are in the shop for serious engine maintenance every 3000 miles: namely that there are physical limits with what can be done with internal combustion engines without sacrificing reliability. Hell, you'd think if they could make one that didn't require massive maintenance on a short schedule, they could sell it for twice as much.
Boosting an S4 to run with Ferraris is counter-productive in the sense that you're likely gonna end up paying the cost of a ferrari in maintenance anyway (well, not quite, but it will be damn expensive and unreliable).
The Bush administration explained its position wrt peacekeepers.
Essentially, it believes that American peacekeepers would destabilize the situation and become targets. As long as Rangers and marines are running around on the ground and B-52s are bombing enemy positions, US peacekeepers would cause a great deal of resentment and be unable to deal with the situation in the neutral manner required by peacekeepers.
Remember, peacekeepers are supposed to be like a militarized police force, not an army.
The conspiracy theory recognizes that the 16th amendment is there, just that its not valid. The wacko theory goes something like this:
Due to a procedural fuck up in admitting Ohio to the Union, Ohio wasn't actually a state until 1953. The sixteenth amendment was passed before 1953. Without Ohio's vote, the 16th amendment wouldn't have passed, since less than 3/4 of the states would have agreed to it. Therefore, the 16th amendment isn't valid.
/loonie
Of course its a bullshit "The Man is keepin' me down" story that would make a black panther blush, but that's how it goes.
I work on DSPs, and I write maybe 10 lines of production code per *week*, if you were to average it out. Sometimes more, usually less.
Now, I do write lots of test code to find bugs. Debugging communication DSPs is a friggin nightmare: there is little visibility except for ram/register dumps and you can't single-step (for a variety of reasons, but namely the reality of sampling in real-time). So I write a lot of code to trap bugs and test hare-brained hacks to get around chip problems. But that's not production.
I've spent weeks on a problem only to solve it by inserting a single nop:)
Uh no, Saddam agreed to allow in UN weapons inspectors as part of his surrender agreement ending the Gulf War. By ejecting the weapons inspectors, he willfully broke that treaty.
Tim
Re:Well, just look at the technical documentation!
on
Modem Accelerators?
·
· Score: 2, Informative
Nope, a cable modem does plenty of modulation and demodualtion. I'm pretty sure that cable modems use QAM (Quadrature Amplituded Modulation), though they may use PAM (Pulse Amplitude Modulation), which I'm much less familiar with.
Anyway, they most certainly do have an A/D-D/A in them. QAM (which I'm pretty sure cable modems use) is similar to DMT DSL in that DMT is (more or less) 256 concurrent narrow-band QAM channels, while cable modems (presumably, my biz is DSL, I've never read more than the occassional whitepaper on cable modems) use a single wide-band QAM channel for each customer while the head-unit maintains several of these connections (one for each customer).
Here's a suggested course of action, it might suck, but its just a suggestion:
1) Mirror the site privately, no public links. This should fall under fair use as long as nobody is able to access it from the slashdot main page.
2) Send an email to the webmaster stating that you are about to link to his site, thus throwing an ungodly amount of hits his way, and that you can toss up a mirror to reduce the strain on his poor, poor webservers.
3) Wait a hour or two for a reply. Its not like/. reports breaking news; cool stuff yes, breaking and time sensitive, no. If there is no reply, just link the original site. If the webmaster gives permission, link both the mirror and the original, and encourage the original unless the webmaster says otherwise (give this choice in your sp^H^H form letter)
Bandwidth for DSL is indeed limited due to attenuation and xtalk, but the attenuation is correlated to line length. So, if you lived next-door to the CO, you might be able to get 12 Mbs, 9 Mbs if you were a mile away, 7 Mbs at two miles, 2 Mbs at 3 miles, etc.
What they are doing here is two things:
1) The data channel: Taking adavantage of the fact that the line is somewhat capacitive, and thus acts a low-pass filter. This means that lower freqequency bins will have a higher signal-to-noise ratio, meaning that they can carry more data without incurring errors. These bits in these low-freq. bins can be used to form a channel with little or no forward error correction, thus keeping the latency down, making it suitable to use for web-surfing, ICQ, and Counter-Strike.
2) The video channel: Common sense says that latency isn't a concern with video. You're not intertacting with it in bi-directional manner, so it really doesn't matter whether you get the image 16ms or 512ms after it was formed on the other side.
But we do need bandwidth. In this case, the high-frequency bins, which normally couldn't carry (as) much data length without incurring errors, can be used to carry the video channel. Since latency, as explained above, doesn't matter, gobs of forward error correction can be applied. This forward-error correction can be translated into extra dB of SNR, since if there is an error, it can always be corrected so long as certain criteria, adhering to the work of Claude Shannon,are met. With this mystical extra SNR (coding gain), more data than usual can be encoded into these high-frequency bins, witht the expectation that any errors will be corrected and that nobody really cares what the latency of their TV is.
Slashdot Mirror
200 hours per week?!?
:P
I'd be impressed too, but I'd hope I could find a better use for a time machine.
Tim
Bullshit.
DMT DSL (which is the DSL modulation used pretty much everywhere) is most definitely a modulated analog signal. DMT is 256 concurrent narrow-band QAM (Quadrature amplitude modulation) channels with variable constellation sizes (up to 15 bit constellations) with a 4KHz (data-)symbol rate. Cable modems use a wide-band QAM channel with a fixed constellation size and a much higher symbol rate.
Building any QAM device without a DAC/ADC and a DSP is a hopeless task. It is most definitely a modem.
QAM, BTW, is what was used for V.34 (33.6Kb/s) voice modems (more-or-less, anyway - there some pretty major differences in implementation but the idea is the same)
Tim
Lookie here, and repeat after me:
"All the world is *not* a PC. All programs are *not* applications. Some programemrs have *zero interest* in writing code that isn't assembly language for small systems, to them C code and operating systems just get in the way"
I happen to be one of these people. Applications hold very little interest for me. I like programming hyper-optimized aeembly for DSPs, data framers, network aggregators, embedded control systems, etc. To me, if I'm not programming to the bare-metal, its an uncomfortable experience.
Needless to say, there isn't a whole lot of hobbiest stuff to do in these areas within the means of anybody who doesn't have a trust-fund and tons of spare time. Just to do my job effectively at work, I have over $300K of scopes and analyzers of various sorts at my bench - needless to say, open source is not gonna happen.
Tim
Well, at my company several engineers interview each candidate and everyone has there own interview style.
:)
For myself, I like to pick a DSP subject on their resume that I only have a passing familiarity with and teach me as much as possible about the subject during our hour or so together. Filter design is my favorite, as I do very little of it myself, but I know enough to be able to pick out bullshit. Generally, I look for how much I learn and how enthusiastically its presented - enthusiastic, competent people that can explain complex subjects to others, well those people rock
But, in the end, you're right. I may have gotten bad vibe off of some competent people because they weren't as hyper as I am about the subject. On the otherhand, I don't know if I want to work with people who don't totally dig their work...
Tim
According to the G.DMT standard (for all annexes), 8 Mbs is required to be considered standard compliant.
However, the standard allows optionally reconfiguring the Reed Solomon codec to S=1/2 (that is, 2 RS codewords per data symbol) which means support up to 16 Mbs. Of course, the actual channel capacity will top out somewhere in the neighborhood of 12-13 Mbs (assuming a short, clean loop) due to Shannon's Law and the realities of current DAC and signal processing limitations. But really, anywhere up to about 7 Kft, 10 Mbs should be no problem assuming the line is reasonably clean.
S=1/2 is a bit tricky to implement and while most vendors have it, it tends not to be interoperable with the implementations of other vendors, so its not widespread since in order to work your chipset at home would have to be the same brand as the one at the CO. In Japan, this is being solved by the LECs standardizing, so 12 Mbs is being rolled out there.
Tim
Well, maybe.
When you're learning the math, I agree, calculators will inhibit your learning to some degree. If you're learning to do something with the math, like in an engineering or physics course, perhaps working through it a few times could be useful. But, when you're trying to actually accomplish a useful goal with the math, its virutally always better to let the computer do it for you.
I work as a DSP engineer and I honestly don't know if I could so a Fourier transform anymore without a boatload of reference material. I just don't deal with the math at that level - I work with people who are much much better at it than I. To me, to getting filter coefficients from time-domain to frequency-domain is going to Matlab and typing:
load stuff
plot (20*log10(fft(stuff, 512)))
Now, my complaint about my education is that learning to do cool stuff with Matlab or Mathematica or a calculator was actively discouraged - everything was to be worked out by hand. I just don't have the time for that now, so Matlab it is, but I feel that my horizons could've been expanded (so to speak) by using the most effective tools while I was still in college and learning to use them to their full potential by solving much more interesting problems with them, rather than 'here, solve this one with Maple/Matlab/Mathematica but do the rest of the drudgery by hand'. Rather, I feel that, for instance, my signals courses (besides the first one, where the math is important) could've covered a lot more ground by explaining the concepts and being Matlab-centric on the implementation. In the real world, the guys who really, really know FFT algorithms well tend to be math-majors or other hard-on-for-numbers types anyway, so they'd probably would've learned it regardless of the tools, but us schmoes could've been saved the headache and learned something useful to boot.
Tim
I work for a chip company that does somewhat mass-market chips (around 2 million various chips per quarter), and based on what I know (I don't work much with the manufacturing side, but do deal with LSI designers who bring up their concerns), here's some back of the envelope calculations:
.18micron.
Our single-core chips cost around $3 each for a good chip. Die size is around 80 mm^2, yield is ~70%, process is
Intel processors are ~400mm^2, so they can get at best 1/5 the chips on the same size wafer. This brings cost to $15 using my starting point. Their yield would be 0.70^5 or 16% - but lets assume that since Intel runs their own foundry, they have better kung-fu than the fabs we use, so maybe 35% yield. That brings us to $30, and I'm talking out my ass at this point.
However, if they didn't have the improved yields I was assuming, you're looking at $60/chip or more just in fabbing cost. Semiconductor companies tend to like at least 60% gross margins, so a $60 chip would sell for $150 minimum, not counting cost of sales.
Tim
Heh, I dropped out of college in June 2000 with 17 credits left for a physics degree and 15 for a compsci degree.
:)
Best decision I ever made
In June 2000, the hiring boom was still in progress, and I got $60K as a driver programmer after submitting a small resume and a space-invaders game I wrote in 80x86 assembly language. I still have that job, and I make considerably more now. Getting my foot in the door when the gettin' was good was what mattered, and it worked out beautifully.
I no longer work on drivers, I do DSP and network processor programming stuff now. And I love it. And I have no fears of being laid-off - my company has tons of cash and is profitable.
All-in-all, great decision, though I was derided by my parents and lost my girlfriend over it, but it was well worth it. (I've since made-up with my parents and have a new, better gf, so everything worked out there too)
I can't imagine how hard that first job would've been in 2001 for me with my 2.0 GPA. I feel for ya.
Tim
Well, I've seen places where management took the brunt of it.
Several years ago, I interned doing IT work at a foundry that was very dependent on the defense industry. Needless to say, it wasn't the roaring 80's for them anymore.
As a result, there were approximately ~20% layoffs of floor-workers, but NONE of the people in director positions or higher were paid a cent for six months. They were effectively laid off, but still expected to show up for work.
Its not always the big-bad management out to get people.
Tim
If its just synchronising, why not just run a wire from light-to-light with a reference tone on it and have each timer phase-lock to that? That would eliminate any drift and probably be a heck of a lot cheaper.
If they were doing any sort of analysis, though, I could see why they'd need PC's.
Tim
Well...
Assuming you wanted the most basic hardware possible, a WinWiFi would be an Antenna, an A-D/D-A and some (somewhat expensive, hardware wise)mod/demod circuitry. The problem with wireless is that the frequencies are way too high to deal with as baseband (you'd have to process 5 billion samples/second to do baseband processing at 2.5GHz), so some hardware is needed to bring the signal to a useable range.
However, I don't think that anybody is going to do that since even after bringing it to baseband, there are still way too many samples to deal with efficiently on a PC.
Rather, I'm under the impression that the hardware will have the analog bits, a DSP, and a basic ethernet framer and let the PC take care of the rest of the ethernet MAC functionality.
Tim
Yup, and if you could read you'd see that the treaty gives a provision for either party to withdraw with 6 months notice, which the US gave.
Tim
IANAP (but I majored in Physics as an undergrad).
Me too.
Anyway, I seem to remember from QED that electrons exist in the same place at the same time with alarming regularity. It seems that subatmoic particles move sorta-freely in time at very small scales, but their net time vector is "forwards". I think this guy just wants to induce it to happen on a large scale (neutrons time-traveling distances greater than femtoseconds)
Of course, subatomic particles also break the 2nd law of thermodynamics for very short periods of time, but that doesn't mean it can be done on any useful scale or induced to happen at will.
Tim
"One does tend to ask oneself why the world's largest military machine was incapable of developing something like this itself until recently"
Most likely, for two reasons:
1) US air-combat stratgies focus on "beyond-visual-range" enagement. Thus, the research focus has been on better long range radars and tracking systems.
2) The US hasn't put a new fighter-plane into service since the early 80's (F-16 and F-18). And the focus on those two planes was maneuvarability and reduced-cost, not an opportunity to try out new tech. Before that, the F-14 and F-15 were in service by the mid-70s and are still the premier fighter planes in service.
The F-22, being the first new fighter-plane the US has made in a long, long time is the first opportunity to add this type of tech.
Tim
Those 100 mile range missiles like the Phoenix are pretty unlikely to hit a competent pilot from any significant distance. All they'd need to do is turn perpindicular to the missile (beaming the radar) and it would most likely be avoided.
The point behind a phoenix missile is to present a credible threat at a distance, forcing the enemy to take these evasive manuveurs (and thus sacrificing tactical advantage) while the F-14 quickly closes the distance and makes a much shorter range kill.
Tim
Damn, an S4 weighs that much? Honestly, I never looked up the weight, but just assumed that since it was about the same size as my 323i, it would weigh roughly the same (~2300 lbs), plus a little added weight for AWD and a bigger engine. Where does the extra weight come from? (the engine and AWD can't account for 1500 lbs? or can they?)
Anyway, I don't think it would be reliable at 400 HP. The extra stress on the pushrods alone I would think would significantly shorten the lifespan of the engine. Yeah, replacing a snapped rod isn't that big of a deal, but its still a pain in the ass if you're somewhere inconvenient. Not to mention the other parts that would be running out of spec.
As for Ferraris, I'm not arguing with you. YOu basically made my point: They are engineered with performance and performance alone in mind. If you build a car this way, its just not going to be reliable, but it will be damn fast. Taking a reliable car and trying to turn it into a ferrari-beater is going to give you the same adverse effects on reliability.
Tim
As per the corvette:
In general straight line, though the newest Z06 did pull higher on the skidpad (dry) than any car I've ever seen, well over 1G.
On the wet skidpad, well....
Anyway, a Corvette is definitely a luxury sports car, maybe not quite as nice as a BMW, but quite plush inside and fast and good handling (as long as it doesn't rain).
Tim
Good lord dude. What the hell do you need with 400 HP on something as small as an S4? Like changing tires much? How about cylinder heads? I heard those were real cheap now...
sigh
The high HP mainstream luxury sports cars (S4, M3/M5, 911, Corvette, etc), in general are limited at the power that they get because a) Its damn fast as is and b) its actually reliable at those HP/Torque numbers.
There's a very good reason why $150,000 Ferraris are in the shop for serious engine maintenance every 3000 miles: namely that there are physical limits with what can be done with internal combustion engines without sacrificing reliability. Hell, you'd think if they could make one that didn't require massive maintenance on a short schedule, they could sell it for twice as much.
Boosting an S4 to run with Ferraris is counter-productive in the sense that you're likely gonna end up paying the cost of a ferrari in maintenance anyway (well, not quite, but it will be damn expensive and unreliable).
Tim
The Bush administration explained its position wrt peacekeepers.
Essentially, it believes that American peacekeepers would destabilize the situation and become targets. As long as Rangers and marines are running around on the ground and B-52s are bombing enemy positions, US peacekeepers would cause a great deal of resentment and be unable to deal with the situation in the neutral manner required by peacekeepers.
Remember, peacekeepers are supposed to be like a militarized police force, not an army.
Tim
The conspiracy theory recognizes that the 16th amendment is there, just that its not valid. The wacko theory goes something like this:
Due to a procedural fuck up in admitting Ohio to the Union, Ohio wasn't actually a state until 1953. The sixteenth amendment was passed before 1953. Without Ohio's vote, the 16th amendment wouldn't have passed, since less than 3/4 of the states would have agreed to it. Therefore, the 16th amendment isn't valid.
/loonie
Of course its a bullshit "The Man is keepin' me down" story that would make a black panther blush, but that's how it goes.
Tim
I work on DSPs, and I write maybe 10 lines of production code per *week*, if you were to average it out. Sometimes more, usually less.
:)
Now, I do write lots of test code to find bugs. Debugging communication DSPs is a friggin nightmare: there is little visibility except for ram/register dumps and you can't single-step (for a variety of reasons, but namely the reality of sampling in real-time). So I write a lot of code to trap bugs and test hare-brained hacks to get around chip problems. But that's not production.
I've spent weeks on a problem only to solve it by inserting a single nop
Tim
Uh no, Saddam agreed to allow in UN weapons inspectors as part of his surrender agreement ending the Gulf War. By ejecting the weapons inspectors, he willfully broke that treaty.
Tim
Nope, a cable modem does plenty of modulation and demodualtion. I'm pretty sure that cable modems use QAM (Quadrature Amplituded Modulation), though they may use PAM (Pulse Amplitude Modulation), which I'm much less familiar with.
Anyway, they most certainly do have an A/D-D/A in them. QAM (which I'm pretty sure cable modems use) is similar to DMT DSL in that DMT is (more or less) 256 concurrent narrow-band QAM channels, while cable modems (presumably, my biz is DSL, I've never read more than the occassional whitepaper on cable modems) use a single wide-band QAM channel for each customer while the head-unit maintains several of these connections (one for each customer).
Tim
Here's a suggested course of action, it might suck, but its just a suggestion:
/. reports breaking news; cool stuff yes, breaking and time sensitive, no. If there is no reply, just link the original site. If the webmaster gives permission, link both the mirror and the original, and encourage the original unless the webmaster says otherwise (give this choice in your sp^H^H form letter)
1) Mirror the site privately, no public links. This should fall under fair use as long as nobody is able to access it from the slashdot main page.
2) Send an email to the webmaster stating that you are about to link to his site, thus throwing an ungodly amount of hits his way, and that you can toss up a mirror to reduce the strain on his poor, poor webservers.
3) Wait a hour or two for a reply. Its not like
Tim
Bandwidth for DSL is indeed limited due to attenuation and xtalk, but the attenuation is correlated to line length. So, if you lived next-door to the CO, you might be able to get 12 Mbs, 9 Mbs if you were a mile away, 7 Mbs at two miles, 2 Mbs at 3 miles, etc.
,are met. With this mystical extra SNR (coding gain), more data than usual can be encoded into these high-frequency bins, witht the expectation that any errors will be corrected and that nobody really cares what the latency of their TV is.
:)
What they are doing here is two things:
1) The data channel: Taking adavantage of the fact that the line is somewhat capacitive, and thus acts a low-pass filter. This means that lower freqequency bins will have a higher signal-to-noise ratio, meaning that they can carry more data without incurring errors. These bits in these low-freq. bins can be used to form a channel with little or no forward error correction, thus keeping the latency down, making it suitable to use for web-surfing, ICQ, and Counter-Strike.
2) The video channel: Common sense says that latency isn't a concern with video. You're not intertacting with it in bi-directional manner, so it really doesn't matter whether you get the image 16ms or 512ms after it was formed on the other side.
But we do need bandwidth. In this case, the high-frequency bins, which normally couldn't carry (as) much data length without incurring errors, can be used to carry the video channel. Since latency, as explained above, doesn't matter, gobs of forward error correction can be applied. This forward-error correction can be translated into extra dB of SNR, since if there is an error, it can always be corrected so long as certain criteria, adhering to the work of Claude Shannon
Technical enough?
Tim