Wait, I'm confused - digital communications theory is pseudo-mathematics? Are you claiming, for instance, that the nyquist theorem is pure hookum? That orthogonal frequency domain multiplexing doesn't work? That it is impossible to estimate the error rate of channel given its constellation dimensions and SNR? That designing a system that can transmit an outgoing signal and decode an incoming one in a fixed amount of time doesn't have a performance requirement?
If that were true, you wouldn't be posting to slashdot or making cell phone calls. Just because it is done in software doesn't make it not engineering - after all, any software could be committed to silicon but it is just silly to do so some of the time.
Ummm, Verilog is a chip-design language. Or doesn't that count engineering either?
Also, lots of software is based heavily mathematics. I've written gobs of production DSP code for digital communications and I could give you more statistics and models of its performance, reliability, lifetime and failure-modes than anybody would care to read (and that wasn't even a particularly mission-critical system - guys who do flight-control and the like could bury you in statistics and models)
I've worked for 10 years as a DSP engineer in the digital communications field for several large semiconductor companies without a degree (mostly on the metal firmware, a little bit of HW design).
Never really had any problems finding a job, either.
If the price goes low enough, another company will buy a controlling stake and take over their revenue stream, thus it very much does have intrinsic worth.
Annex I in the ADSL2 and ADSL2+ standards allows the upstream channel* to start start at tone 1 (4.3125 KHz) instead of tone 6 (25.875 KHz). Obviously, you don't get POTS in this mode - it's meant as an all-digital design for telco VOIP roll-out. It doesn't buy you all that much - 15 bits/tone * 5 tones * 4Khz data symbol rate = 260 Kbits/second. That's the theoretical maximum and most likely its going to much less than that since most existing modems weren't designed with this in mind and likely have very poor amplifier performance (if not built in high-pass filters) near the DC range and I don't think the market for it is big enough to convince anybody to spin a chip for this feature.
*- and technically the downstream, too, I think, but there are a myriad of issues with doing that revolving around the massive echo rejection you need and dealing with near-end crosstalk at the DSLAM end, so most ADSL designs that are actually deployed are frequency domain duplex (the upstream and downstream use different frequencies).
Yeah, the TI C6xxx compiler does a good job if you use the intrinsics and the correct #pragmas (generally to promise the compiler you aren't doing any pointer aliasing), but otherwise, you're just as hosed as the poor Itanium developer.
This is fine for embedded DSP applications, where a large project is maybe 100KLOC and one guy can appropriately #pragma-bomb the entire codebase in two weeks; much less fine if you're writing a 30MLOC database application or galaxy simulator...
Hey, assuming that's YahooBB/SoftbankBB or Acca, that's my modem:) And if its NTT, go fuck yourself:)
So, how's it working out for you? 50 Mbs was a tall order, but we think we met it pretty good, at least on lab loops - I'm more interested in how your live loops is doing - are you really getting 50 Mbps on a live loop? Do you even get that during peak connect time? Consider this your chance to talk back and vent to a chip vendor:)
So, the individual must sacrifice for the good of the state/human race? Spoken like a true fascist.
So, what do you plan on doing to people like me who don't particualrly give a fuck about global warming and just want to drive a hummer? Shoot me? Imprison me? Send me to a re-education camp until I take public transit? Come you Nazi fuck, let's hear the plan...
I've never understood that "you must work x hours" attitude. Mandatory core hours (say 10-4), for face-time communication, sure, that's reasonable, but otherwise, who gives a fuck if you're getting everything done?
Fortunately, I work at a (relatively) sane company. In general, engineers are allowed to estimate their work and as long as you hit the estimates, its all cool - how you got there isn't terribly important. In fact, by running this system, it has become almost a point of pride among some of us to grossly underestimate the time required and then show off by finishing up a month early:)
Yeah, this sometimes means I'm working a 60 hour week, but more often, even with the gross underestimates, I find that I get in the necessary crunch at the beginning by attacking the hardest part first and finsing nothing but smooth sailing (and 30 hour weeks) after the really hard part of the problem has been licked. This, I believe is a result of being fully accountable for my estimate - management will give a 2 week grace period or thereabouts if you're making progress, but after that, expect to catch hell for missing deadlines.
We have our pathologies to be sure, mostly related to people working on the wrong things from a priority perspective (there are no less than 7 different priority lists floating around in different departments!), but we do have an uncanny ability to deliver when we say we're going to deliver without having to do a whole lot of overtime.
Where I grew up, I could live quite well on $35K. Maybe not as well as I do now, but I wouldn't exactly be starving either.
What do you think people make in Europe? As far as engineers go, I work with 30+ Europeans and almost every single one of them came to the US because engineers make roughly 2-3x what they do in Europe. In the US (on either coast - in the middle its less), a MSEE is easily $90K/yr. In France (even in Paris), $35-40K Euro if you're lucky enough to find a job.
Quite frankly, I'll take 4 weeks of vacation at twice the pay rate over 9 weeks of vacation and the government taxing the shit out of me to give it to a bunch of lazy slobs who can't be bothered to find a job.
If you go from a 256 bit parallel bus running at 800 MHz to a serial link, your motherboard traces are going to have to carry a signal at something like 200 GHz to get the same bandwidth
You're assuming a 1-bit/symbol serial link using nothing but TTL logic, when really, that's just a horrible way to transport data from a spectral- efficiency standpoint.
By using, say, a differential pair and a 64-QAM modem (in the strict modulator/demodulator sense) we can get 5-bits per clock. But really, we can probably do even better than that, given the short distances involved.
The SNR is there to transport the data very fast in a reasonable physical bandwidth and the cost of the logic to do the transmission and receiving is getting cheaper. I wouldn't be suprised to see this approach take hold in the next 5 years or so.
GSM in Europe works well due to a combination of popualtion density and build-out.
Techwise, though, GSM doesn't hold a candle to CDMA. GSM is just TDMA with a slightly different equalizer training period (in the middle of the symbol, rather than the beginning). CDMA, on the otherhand, is uber-cool technology and the future of wireless communications (everything is going to WCDMA or CDMA2000 eventually). The cool thing about CDMA, BTW, is that it can theoretically support unlimited (yes, unlimited!) calls per cell if we can perfect the power-control on the transmitters such that the received power at the receiver is the same from every transmitter (and provide enough mod/demod functions, but providing unlimited mod/demod-functions isn't that hard)
For me, though, my low-tech TDMA phone works just fine in New Jersey and I have no plans on switching until something services my area better. Verizon CDMA is good, but more expensive and GSM just blows around around here...
That's because you have a set (large) amount of computing resources and your OS has to manage that.
In this case, they're building their system from scratch and will only include as much processing power as they need. So a less resource intensive OS can mean the difference between a 400 MHz ARM or a 200 MHz one, with he resultant power savings being always there.
You're an elitist shithead because you don't want other people to watch "Hannity & Colmes". Why does it offend you so much that other people prefer Fox News to NPR? Can't they just watch what they want without a fucking sermon?
Slashdot Mirror
Even if there is a market for featureless, large TVs, the manufacturing oligopoly will not create them,
What on earth are you talking about? I just bought a 42" panasonic plasma for $550.
6 years ago, I bought a 42" panasonice plasma TV for $3000 (still works great, but I need another TV for the other living room)
Wait, I'm confused - digital communications theory is pseudo-mathematics? Are you claiming, for instance, that the nyquist theorem is pure hookum? That orthogonal frequency domain multiplexing doesn't work? That it is impossible to estimate the error rate of channel given its constellation dimensions and SNR? That designing a system that can transmit an outgoing signal and decode an incoming one in a fixed amount of time doesn't have a performance requirement?
If that were true, you wouldn't be posting to slashdot or making cell phone calls. Just because it is done in software doesn't make it not engineering - after all, any software could be committed to silicon but it is just silly to do so some of the time.
Ummm, Verilog is a chip-design language. Or doesn't that count engineering either?
Also, lots of software is based heavily mathematics. I've written gobs of production DSP code for digital communications and I could give you more statistics and models of its performance, reliability, lifetime and failure-modes than anybody would care to read (and that wasn't even a particularly mission-critical system - guys who do flight-control and the like could bury you in statistics and models)
I've worked for 10 years as a DSP engineer in the digital communications field for several large semiconductor companies without a degree (mostly on the metal firmware, a little bit of HW design).
Never really had any problems finding a job, either.
How much do they pay you to make shit up?
Huntsville, Alabama has more PhDs per capita than any place on earth.
Paying off credit cards fully every month for example is a negative on your credit score.
Bullshit. You're lying and you know it.
If the price goes low enough, another company will buy a controlling stake and take over their revenue stream, thus it very much does have intrinsic worth.
This signal processing engineer and his hundreds of Matlab scripts would like to politely diagree with your assessment of Matlab:)
Ya know what, I'll take my cheap tickets, screw the service. How much did it cost to fly from San Jose to JFK (non-stop even!) in 1976? $1000? $3000?
Today, you can do it for ~$200 round trip. Which is freaking awesome. Its only 5 hours each way, I can suck it up and deal. Quit being such a whiner.
(and the last trip I took was San Francisco to Munich. $710 round trip on United. How much was that one in 1976?)
They're called photons. You probably see trillions every day.
Actually, the answer is yes.
Annex I in the ADSL2 and ADSL2+ standards allows the upstream channel* to start start at tone 1 (4.3125 KHz) instead of tone 6 (25.875 KHz). Obviously, you don't get POTS in this mode - it's meant as an all-digital design for telco VOIP roll-out. It doesn't buy you all that much - 15 bits/tone * 5 tones * 4Khz data symbol rate = 260 Kbits/second. That's the theoretical maximum and most likely its going to much less than that since most existing modems weren't designed with this in mind and likely have very poor amplifier performance (if not built in high-pass filters) near the DC range and I don't think the market for it is big enough to convince anybody to spin a chip for this feature.
*- and technically the downstream, too, I think, but there are a myriad of issues with doing that revolving around the massive echo rejection you need and dealing with near-end crosstalk at the DSLAM end, so most ADSL designs that are actually deployed are frequency domain duplex (the upstream and downstream use different frequencies).
Erm, no, you are actually completely wrong:
http://www.bls.gov/cps/uiclaims.htm
Yeah, the TI C6xxx compiler does a good job if you use the intrinsics and the correct #pragmas (generally to promise the compiler you aren't doing any pointer aliasing), but otherwise, you're just as hosed as the poor Itanium developer.
This is fine for embedded DSP applications, where a large project is maybe 100KLOC and one guy can appropriately #pragma-bomb the entire codebase in two weeks; much less fine if you're writing a 30MLOC database application or galaxy simulator...
Tim
Hey, assuming that's YahooBB/SoftbankBB or Acca, that's my modem :) And if its NTT, go fuck yourself :)
:)
So, how's it working out for you? 50 Mbs was a tall order, but we think we met it pretty good, at least on lab loops - I'm more interested in how your live loops is doing - are you really getting 50 Mbps on a live loop? Do you even get that during peak connect time? Consider this your chance to talk back and vent to a chip vendor
Thanks,
Tim
I. Don't. Want. To.
Clear enough? I don't like being in apartment buildings, I don't like public transit, I don't like bumping into people on the subway, etc, etc.
You live where you want, I live where I want, we're both happy.
At least until some condescending asshole from NYC decides to berate people for living where they want...
Tim
So, the individual must sacrifice for the good of the state/human race? Spoken like a true fascist.
So, what do you plan on doing to people like me who don't particualrly give a fuck about global warming and just want to drive a hummer? Shoot me? Imprison me? Send me to a re-education camp until I take public transit? Come you Nazi fuck, let's hear the plan...
Tim
I've never understood that "you must work x hours" attitude. Mandatory core hours (say 10-4), for face-time communication, sure, that's reasonable, but otherwise, who gives a fuck if you're getting everything done?
:)
Fortunately, I work at a (relatively) sane company. In general, engineers are allowed to estimate their work and as long as you hit the estimates, its all cool - how you got there isn't terribly important. In fact, by running this system, it has become almost a point of pride among some of us to grossly underestimate the time required and then show off by finishing up a month early
Yeah, this sometimes means I'm working a 60 hour week, but more often, even with the gross underestimates, I find that I get in the necessary crunch at the beginning by attacking the hardest part first and finsing nothing but smooth sailing (and 30 hour weeks) after the really hard part of the problem has been licked. This, I believe is a result of being fully accountable for my estimate - management will give a 2 week grace period or thereabouts if you're making progress, but after that, expect to catch hell for missing deadlines.
We have our pathologies to be sure, mostly related to people working on the wrong things from a priority perspective (there are no less than 7 different priority lists floating around in different departments!), but we do have an uncanny ability to deliver when we say we're going to deliver without having to do a whole lot of overtime.
Tim
GlobespanVirata merged with Conexant and now goes by Conexant. Didn't notice if they were on the list, though.
Tim
Well, as Werner von Braun said: "I aim for the moon, but sometimes I hit London".
Tim
And...?
Where I grew up, I could live quite well on $35K. Maybe not as well as I do now, but I wouldn't exactly be starving either.
What do you think people make in Europe? As far as engineers go, I work with 30+ Europeans and almost every single one of them came to the US because engineers make roughly 2-3x what they do in Europe. In the US (on either coast - in the middle its less), a MSEE is easily $90K/yr. In France (even in Paris), $35-40K Euro if you're lucky enough to find a job.
Quite frankly, I'll take 4 weeks of vacation at twice the pay rate over 9 weeks of vacation and the government taxing the shit out of me to give it to a bunch of lazy slobs who can't be bothered to find a job.
Tim
If you go from a 256 bit parallel bus running at 800 MHz to a serial link, your motherboard traces are going to have to carry a signal at something like 200 GHz to get the same bandwidth
You're assuming a 1-bit/symbol serial link using nothing but TTL logic, when really, that's just a horrible way to transport data from a spectral- efficiency standpoint.
By using, say, a differential pair and a 64-QAM modem (in the strict modulator/demodulator sense) we can get 5-bits per clock. But really, we can probably do even better than that, given the short distances involved.
The SNR is there to transport the data very fast in a reasonable physical bandwidth and the cost of the logic to do the transmission and receiving is getting cheaper. I wouldn't be suprised to see this approach take hold in the next 5 years or so.
Tim
GSM in Europe works well due to a combination of popualtion density and build-out.
Techwise, though, GSM doesn't hold a candle to CDMA. GSM is just TDMA with a slightly different equalizer training period (in the middle of the symbol, rather than the beginning). CDMA, on the otherhand, is uber-cool technology and the future of wireless communications (everything is going to WCDMA or CDMA2000 eventually). The cool thing about CDMA, BTW, is that it can theoretically support unlimited (yes, unlimited!) calls per cell if we can perfect the power-control on the transmitters such that the received power at the receiver is the same from every transmitter (and provide enough mod/demod functions, but providing unlimited mod/demod-functions isn't that hard)
For me, though, my low-tech TDMA phone works just fine in New Jersey and I have no plans on switching until something services my area better. Verizon CDMA is good, but more expensive and GSM just blows around around here...
Tim
That's because you have a set (large) amount of computing resources and your OS has to manage that.
In this case, they're building their system from scratch and will only include as much processing power as they need. So a less resource intensive OS can mean the difference between a 400 MHz ARM or a 200 MHz one, with he resultant power savings being always there.
Tim
I don't care waht you watch of don't watch.
You're an elitist shithead because you don't want other people to watch "Hannity & Colmes". Why does it offend you so much that other people prefer Fox News to NPR? Can't they just watch what they want without a fucking sermon?
Tim