I don't have a problem with burn-in- but I use the screen saver- more important is to shut off your backlight- they have a limited lifetime, and will lose intensity over time.
It sounds like something's not quite right in your setup. Good cards, like the Acenic cards, can easily max out the PCI bus when you're testing *only* the network, and the disk subsystem isn't involved. That means 80-100 MB/s (640-800 mb/s) throughput on a 32/33 bus, and they hit as closely to a gigabit as can be had on a 64-bit bus. If you're only getting a max of 500 mbits/sec, then something's off.
If you're only testing the NIC and get anything but wire speed, your NIC is garbage. I'm talking about using real world measurement tools like Chariot, from multiple clients, pounding on a server. To approach wire speed with decent benchmarking tools, you need a fast processor and a 64 bit bus, or a 32/66 PCI bus. There is just too much overhead.
Luckily, pretty much ever chipset manufacturer has gotten away from using a 32/33 PCI bus to talk to the northbridge, so that's not a problem.
But how many PCI busses do you have feeding into the chipset? Again, that's your bottleneck. From my experience, PCI is limiting. Sure, if every device has a direct bridge to the chipset, no problem- but thats not the way I've seen any PCs built. I think our arguement here is about overhead and delays- my experience has shown them to be much more limiting that what you've seen.
2 gigabit controllers is 4 gigabits? Assuming that you're really doing a full gigabit in *each* direction on both NICs (and with some gigabit nics, you simply won't), and that you aren't being swamped by interrupts, that's still only ~500 MB/s. barely more than a 64/66 PCI bus. Easily within the reach of a 64/100 bus.
Yep- Gigabit is full duplex. I work in a lab where we test gigabit controllers every day. It isn't easy to choke a gigabit link, but it can be done. The controllers that are on 32/33 PCI busses typically top out somewhere in the 300-500 Mbit region (total throughput, ins and outs).
Sure, I did. I never said differently. Perhaps you're forgetting that a lot of motherboards now have more than one PCI controller, and hence, more than one PCI bus? I've got a 4-year old machine that has 4 seperate PCI controllers, of varying speeds and widths. That means that you don't need to put all of your high-bandwidth devices on the same bus, giving you more total bandwidth, and less overhead/contention from sharing.
You're moving the choke point around, but not eliminating it. If I've got a series of PCI bridges (careful about your terminology- controller is generally used to refer to an end-point, not the bridge), everything still needs to go through the north-bridge, and if you are hooked up to that with a 32/33 PCI bus- you're majorly choked right there.
One of us must be incredibly mistaken. $1500 for optical interconnects? Optical interconnects are getting cheap enough that there's even talk of building them right into processers.
Talk, maybe, but no processor manufacturer has even put anything like that on their roadmap. Optical transcievers are *big* compared to copper, and difficult to connect to. Switching our PCB technology to optical will also be really expensive. At this point, you can't economically just "draw" optical traces. $1500 was my rough estimate for the whole package, not for an infividual transciever. Copper still has a long way to go, in particular for short interconnects like you see on motherboards.
Yes. It takes some work, but it does. It takes some high power hardware, but you can choke a PCI bus.
Let's put some numbers to that. On a 32/33 bus, you're looking at a maximum real-world, sustained throughput of about 100 megabytes/second. Double the width, that's 200 megabytes/second. Double the frequency, that's 400 megabytes/second.
You're being awfully optimistic here. One thing you aren't taking into account is that PCI is a bus. One talker, one listener, and there is a good deal of overhead. If anyone else needs to talk on the same bus- that is a slice taken out of your bandwidth pie.
Alrighty, then. Nearly a half of a gigabyte per second. That's awfully tough to fill. That will handle two gigabit ethernet controllers running full-tilt, and still have enough bandwidth left over that you'd need at least an INCREDIBLY fast RAID array to fill it.
Not quite. To get two full gigabit controllers working full tilt, you will need two ports off of the bridge, and even then, the bridge has to talk to the CPU, so you may be limited by the CPU bus speeds. 2 gigabit controllers running full tilt (full duplex, of course) is 4 gigabit of raw bus bandwidth (not counting overhead, of course).
Now, really... can *anyone* here raise their hand and say that they could actually *utilize* 2200 megabytes/second of bandwidth to the outside world, either via network or disk?
But PCI is not used solely for outside bandwidth. It is a bus, and shared. Everything uses a little bit- down to your legacy I/O. Now, when it *IS* replaced, I'd much rather see the interconnects being optical, not electrical. Instead of cracking open the case, shutting off the power, and trying to wedge yet another card inside (especially in low-height rackmounts), I'd much rather set the device on a shelf, and run a fiber patch cable over to the computer. No shutting down, and a whole lot more simple.
PCI *IS* being replaced. PCI-Express is the next stop. It is far better electrical signalling wise, uses a whole lot fewer wires for the same bandwidth. Hot swap already does exist for PCI- you have to go to higher end servers to support it, but it does exist- the biggest limitation is the O/S support. As to external chassis based peripherals you can hot swap- It was tried with Infiniband, but it never really caught on. You can see how far Infiniband has gotten in the past year. (not far) PCI-Express is basically Infiniband-lite. They took away some of the more ambitious goals and made something that was a lot more implementable and realistic.
I'd really like to see more optical interconnects- but the infrastructure is just too expensive. People aren't willing to buy it, and far too expensive to put on a motherboard (Are you willing to pay $1500-$2000 per motherboard?)
PCI-Express is *NOT* compatible (in a hardware sense) with PCI. Never meant to be, will never do it. Different connectors and so forth. From a software sense, it is supposed to look just like PCI.
Accurate intertial sensors are actually really expensive- an inertial sensor doesn't measure speed, it measures acceleration. if you use the simple eqation of d=.5*a*t^2, and say, an error of 1/1000 g, over an hour, you would be 63.5 kilometers off. I didn't think it was that bad- but I double checked my math.
Some of us still run mailing lists to connect a group of friends- who pays then? It is a perfectly legitimate use... but it seems scary if I'm would have to register my mailing list to get an 'exemption'
I think the biggest failing in this is that to tax email would require a massive change to the email infrastructure- just send all email through your government approved relay. Sure- they won't look at it... putting this on top of SMTP- I don't think it would work- what would be the incentive to use it (other than possibly spam free email)?
These devices might not fit into a toaster but I know they could be made smaller.
It's going to be really hard to make a whole jack/magnetics/controller/processor much smaller. You've got the absolute limitation of the RJ45, plus the magnetics- which is absolutely necessary for ethernet. You can get pretty close with capacitive isolation, but you're probably never going to be fully compliant with the IEEE spec. This device has contained in something about the size of a ethernet integrated mag/jack that has the whole shebang- pretty good.
You're going to have a hard time doing a 100Mbit interface that is truly "low power." With 100Mbit, there is always something going over the link, putting +/- 1V over a 100 ohm load, counting inefficiences, you're probably at 40 mA just to support the TX portion of the PHY. Then you have to realize that you need a 125 MHz clock going on inside-and that's all before you have a MAC and a processor going. Ethernet (particularly 100Mb) is not a low power interface.
If you want to keep the customer, the first time it happens, you might want to forgive the excess bandwidth charges (while pointing out the specific clause in the contract that says you have every right to charge them), tell them that it's "for this time only," and make a record of it. This is the type of action that can inspire customer loyalty. If you want to keep customers, you need to find some ways to differentiate yourself from all your competitors. Since you're keeping records, you should be able to tell if a customer is just trying to abuse your policies.
You need to ask yourself- how much did the excess bandwidth really cost, and how much is this customer worth to me in the long run? Probably, keeping that customer will make far more impact on your company in the long term than if you charged them, pissed them off, and inspired them to switch to another ISP.
They have the right, but, by law (in the USA) if they make a negative decision because of the credit report, they have to inform you of that. This is often overlooked. There was a report on this on NPR recently (Jan 31st, All Things Considered).
At the very least, you should check your credit report to make sure it is accurate.
Monitoring is easy, tuning is difficult
on
Cars for Tinkerers?
·
· Score: 4, Informative
While you can monitor your car pretty easy with devices to read things like the OBD, but actually changing how it works is pretty difficult. The car companies have some obligation to the EPA (in terms of making sure their cars are (somewhat) clean, and keeping them that way, by making it difficult for the average joe to twiddle with his ECU like he used to be able to twiddle his carbeurator. From the factories, cars are tuned for the optimum performance/emissions that satisfy the EPA and the marketers, virtually anything you do to improve performance will make emissions worse.
Some carmakers have gone to the extreme to make it difficult to change the ECU maps- Ford's (now obsolete) EEC-IV used a special version of the Intel 8051 (the 8061) and EEPROM that Intel (and their 2nd sources) were *only* authorized to sell to Ford. There were some modifications that could be done to improve performance, but those were a kludge that used a diagnostic port. [I worked as a co-op for Ford in the late 80's]
If you really do want to tinker with your car's engine, look to cars with engines that have been around a very long time- such as the Chevy 350, people have been tuning and playing with the 350 for decades, even as it has gone from carbs to EFI. There is a huge installed base and tinkerers are everywhere.
I worked on one of these- when I worked for Andrew corporation, I did some hardware and coding for the Extensis (mentioned below). Trying to get it to act like a landline was pretty tough- there is no really set way to determine when the call should actually go through- I ended up with a complex state machine based on a bunch of inputs, like how many numbers dialed, pause between the digits, and what number was dialed. (911 gets you right through, for example) The biggest downside I see is that 911 doesn't give the operator your address instantly in case of an emergency (yet, at least). Also, you're going to be pretty limited in the number of telephones that the device can power- You're probably not going to get one of these cradles that will drive the normal phone company max load of 5B (about 5 normal old telephones with mechanical ringers)
You can't use a digital phone for any sort of modem, but otherwise, it should work great. The first generation I worked on was for a motorola Star-Tac type phone. These are pretty complex little devices.
Re:This is about 5Ghz technology
on
DOD vs. 802.11b
·
· Score: 2
Electromagnetic radiation coming from a discrete source in open space (not a wire that goes to (effective) infinity like a power line) is always proportional to 1/r^2- it may radiate more power, but that's a factor of the radar cross section- remember I said proportional? There may be a constant in front of the 1/r^2. There are other factors like antenna gain and antenna power, but those are part of the constant- Since I have Skolnik's book in front of me: Pr=(Pt*G*Ae*sigma)/((4*pi)^2)*R^4, where Pr is power at the reciever, Pt is transmitted power, G is the antenna gain, Ae is the antenna aperture size, sigma is the radar cross section, and R is the range to the target.
Flipping through the book some more, I think I've found the real source of your inability to swallow- you're thinking in terms of light- where the wavelength is incredibly small in proportion to the size of the target (called the optical reigon in radar cross section-ese). Radars typically operate in the Rayleigh region, where the radar cross section varies as wavelength^-4. In between the Rayleigh and the optical region is the Mie or resonance region- we don't want to go there- it's really hard to work there, because the RCS can vary by 400% with a small change of the target (angle to the radar, typically).
Re:This is about 5Ghz technology
on
DOD vs. 802.11b
·
· Score: 2
Nope, if it's a monostatic radar (TX and RX at the same place)- the path from the radar to target is r, so power incident upon the target is C/r^2, (C is a constant that relates to power, antenna gain, and so forth). You multiply that times the radar cross section to get the amount of power reflected and then multiply that by 1/r^2 again to get the incident power back at the radar. The power at the reciever is proportional to 1/r^4.* The r^4 term is one reason that jammers have such an advantage (their power only drops off at 1/r^2).
If you're really interested in it, check out Merril Skolnik's book, "Introduction to Radar Systems," Chapter 1 (I think) where he generates the "radar equation."
* Of course, this is all assuming that you are far enough away from the antenna to be in the far field, but if you're not, you really don't have to use a radar, you could probably hit the target with a well thrown rock.
Re:This is about 5Ghz technology
on
DOD vs. 802.11b
·
· Score: 2
the return signals are weak and fall off 1/r^2 limiting the range.
Ok, this is nitpicking, and the poster makes a strong point, but I think that there is one point that needs to be corrected, which makes the situation even worse- the transmitted signal falls off at 1/r^2, and the reflected signal falls off at 1/r^2, so the radar return actually falls off at 1/r^4!
When it comes to safety issues, the regulatory bodies are very conservative, and because of the massive amounts of infrastructure it requires, changing things will take a long, long time. There is a finite amount of useful (right now) spectrum, and we have to carefully balance convenience and consumer demand (Wi-Fi) vs. safety.
Fortunately, here in the states, motorcycle drivers are required to take a safety course teaching the proper handling of a motorcycle in difficult situations.
Ummm... motorcycle drivers are not required to take a safety course in the states (that is, if you are referring to the United States of America) That sort of thing is set by the state anyway. Motorcyclists are encouraged to take a safety course, and doing so makes it much easier to get a license. I believe that the laws elsewhere (particularly in Europe) enforce much stricter licensing requirements (such as graduated licensing). I used to be a certified Motorcycle Safety Foundation instructor (I let my certification expire- too much work in miserable conditions for too little pay).
There are some states that require you take a motorcycle safety course if you are under 18 and want to get a motorcycle license, but if you're over 18 in most places, all you have to do is pass the test(s).
I thouroughly recommend the MSF courses for anyone who is even thinking about getting a motorcycle- they are the best form of instruction we have. They have proven techniques and a demonstrated track record. One of the most important concepts stressed is that motorcycles are dangerous and you must accept the risk, or not ride. They teach ways to reduce the risk, but it can't be eliminated.
Sole source parts are always a bad thing to have, but if they are the normal jellybean parts (resistors, capacitors, transistors, etc.) you can usually work around and find an adequate substitute in a reasonable amount of time. On the other hand programmable parts tend to have unique code- that can't easily (or quickly) be transferred from one part to another.
A case in point from my own experience about 2.5 years ago- flash memory had a huge upswing in demand- I believe it was from cell-phones. It was so lucrative, that Atmel switched its fabs over to producing lots of flash memory, and putting us microcontroller users on allocation- we went from a 6 week lead time on production quantities to a 6 *month* lead time in a matter of weeks, and even then, they wouldn't guarantee us parts- it was more like 6 months to get on the list to maybe get parts. Microcontroller code doesn't port nearly as easily as higher level code- you tend to have to use every last resource.
This caused a good number of manufacturers to biased against Atmel- they definitely have their good points, but if you can't get them, they're useless. Unless you're a really big company, it is hard to get continuity of supply agreements. I know that even now (working for a really big company) I hesitate to specify Atmel micros.
Ok, I've got some experience here- I've been testing and designing Ethernet for one of the major computer manufacturers for a couple years now. Most manufacturers of Ethernet use shielded connectors (jacks). This shield is what connects to your shield on shielded RJ45 cable. On every shielded ethernet port I've seen, the shield connects to chassis ground- which is the same ground as the ground pin on the power cord. The differential pairs themselves (according to the IEEE spec) should be electrically isolated with a transformer to take up to 2250 VDC or 1500 VAC.
In a situation where you're not sure if you're running the two ends off of a common ground, I definitely would not recommend using shielded cable. You could end up sending that cable up in smoke if your grounds shift.
I've worked for two different government funded labs, so I'm a bit biased, but I think that government funded research is vital for the long term health of our country.
The great thing about working in government funded research is that you have the ability to fail. Failure can be good. Unfortunately, in the commercial world, failure is bad, and must be avoided at all costs- if you fail, you go out of business. But if you don't fail sometimes, you're not pushing the envelope hard enough, not taking any risks. Unfortunately, there has been much movement in the government to divert resources to private industry- to people looking for short term profit.
Private industry looks for short term gains- long term is 5 years, 10 years (or more) out is just too far. The government can afford to look that far out, or farther. That is where the neat stuff happens.
Someday, I'll be back there, back to making cool stuff, and trying to avoid the politics as much as possible.
Really, the thing that needs to be done is improve the *protocol* that the communciation rides upon. If you're using gigabit, usually the limiting factor is the processor- you max the processor (doing the overhead for TCP/IP) before you even come close to filling up the gigabit pipe. There are technologies like Infiniband which try to address that, but progress has been very slow.
There is a big push for TCP offload- to take the effort for TCP and put it into an ASIC- check out this paper. TCP offload is absolutely necessary if we want to go beyond 1 gigabit ethernet any time soon, because the rate that we can communicate is beating the rate of increase in processing. IIRC Moore's law is an 18 month doubling time, there is a similar law for communications speed that says it doubles at a 12 month rate.
In addition, the engineering hassles of making a device with so many paralell channels is pretty hard- When you start doing high speed communications, you have to take a *lot* of care to make sure that the signals get there as you want them. This is why Intel appears to be in the process of moving away from PCI (a paralell bus) and moving to PCI-Express/3GIO, a combination paralell/serial bus, with a variable number of serial channels. PCI-Express is designed to carry on 16 wires what PCI-X carries on 80+ wires. When you actually have to make hardware- this makes a *HUGE* amount of difference.
I read some technical literature on code-hopping remotes- you can have multiple remotes and so forth. A code hopping remote is a one-way device- it only transmits, and it transmits a different code each time you press the button. The reciever knows what the code was the last time you pressed the button, so it knows what the next code should be (they use very, very long pseudo-random sequences).
So what happens if you press the remote when you're too far away, or your 3 year old finds it in the car and presses it merrily for a few miles until he gets bored? If the current code is code N, the next to be transmitted is code N+1, and the next that the reciever will key on will be N+1 to some range of N+X where X is pretty large, but still far smaller than Y, the total number of different codes.
In addition to keying on something in the next group of X codes, another safeguard thats used is to key on two successive remote-keys- lets say my transmitter is out of sequence, say by A - the reciever is looking for a code between N+1 and N+X- it recieves the code N+A (where N+1N+AN+X) It won't key on that, but if the next code it recieves is N+A+1, it knows that it got the right sequence, and will only open after the next keypress.
For multiple remotes, it adds a separate domain to key upon(say, Y to Y+N)- slightly less secure, but of course the more keys there are for any door, the less secure it is.
May seem kinda silly, but an old refrigerator door would work- magnetic, (relatively) airtight, positive seal. Positively pressurize the inside, with a few stages of HEPA filtering. You'll need to be pretty rigorous about filter maintenance. I heard of someone using refrigerator doors in an airlock to seal off a home recording studio. Even if it doesn't work, it would look pretty cool.
May not be good enough for silicon fab, but it probably would be good for people with serious allergies.
Also known as PCI-Express- but these have the difficulty of not being designed for long distances. PCI-express does have some advantages, like being able to choose the amount of bandwidth you want by choosing the number of pairs of wires, so it is scalable.
Whatever hardware interconnect comes next, it will only catch on if we have a big software effort to support it- I'd like to know which interconnect the software community may rally behind.
Slashdot Mirror
Energy storage is still expensive- which is another technology jump that may be required if we want portable cheap electricity.
Unfortunately, petroleum products remain as one of the densest ways to store energy (with a relatively low cost to convert this into useful work).
I don't have a problem with burn-in- but I use the screen saver- more important is to shut off your backlight- they have a limited lifetime, and will lose intensity over time.
If you're only testing the NIC and get anything but wire speed, your NIC is garbage. I'm talking about using real world measurement tools like Chariot, from multiple clients, pounding on a server. To approach wire speed with decent benchmarking tools, you need a fast processor and a 64 bit bus, or a 32/66 PCI bus. There is just too much overhead.
Luckily, pretty much ever chipset manufacturer has gotten away from using a 32/33 PCI bus to talk to the northbridge, so that's not a problem.
But how many PCI busses do you have feeding into the chipset? Again, that's your bottleneck. From my experience, PCI is limiting. Sure, if every device has a direct bridge to the chipset, no problem- but thats not the way I've seen any PCs built. I think our arguement here is about overhead and delays- my experience has shown them to be much more limiting that what you've seen.
Yep- Gigabit is full duplex. I work in a lab where we test gigabit controllers every day. It isn't easy to choke a gigabit link, but it can be done. The controllers that are on 32/33 PCI busses typically top out somewhere in the 300-500 Mbit region (total throughput, ins and outs).
Sure, I did. I never said differently. Perhaps you're forgetting that a lot of motherboards now have more than one PCI controller, and hence, more than one PCI bus? I've got a 4-year old machine that has 4 seperate PCI controllers, of varying speeds and widths. That means that you don't need to put all of your high-bandwidth devices on the same bus, giving you more total bandwidth, and less overhead/contention from sharing.
You're moving the choke point around, but not eliminating it. If I've got a series of PCI bridges (careful about your terminology- controller is generally used to refer to an end-point, not the bridge), everything still needs to go through the north-bridge, and if you are hooked up to that with a 32/33 PCI bus- you're majorly choked right there.
One of us must be incredibly mistaken. $1500 for optical interconnects? Optical interconnects are getting cheap enough that there's even talk of building them right into processers.
Talk, maybe, but no processor manufacturer has even put anything like that on their roadmap. Optical transcievers are *big* compared to copper, and difficult to connect to. Switching our PCB technology to optical will also be really expensive. At this point, you can't economically just "draw" optical traces. $1500 was my rough estimate for the whole package, not for an infividual transciever. Copper still has a long way to go, in particular for short interconnects like you see on motherboards.
Yes. It takes some work, but it does. It takes some high power hardware, but you can choke a PCI bus.
Let's put some numbers to that. On a 32/33 bus, you're looking at a maximum real-world, sustained throughput of about 100 megabytes/second. Double the width, that's 200 megabytes/second. Double the frequency, that's 400 megabytes/second.
You're being awfully optimistic here. One thing you aren't taking into account is that PCI is a bus. One talker, one listener, and there is a good deal of overhead. If anyone else needs to talk on the same bus- that is a slice taken out of your bandwidth pie.
Alrighty, then. Nearly a half of a gigabyte per second. That's awfully tough to fill. That will handle two gigabit ethernet controllers running full-tilt, and still have enough bandwidth left over that you'd need at least an INCREDIBLY fast RAID array to fill it.
Not quite. To get two full gigabit controllers working full tilt, you will need two ports off of the bridge, and even then, the bridge has to talk to the CPU, so you may be limited by the CPU bus speeds. 2 gigabit controllers running full tilt (full duplex, of course) is 4 gigabit of raw bus bandwidth (not counting overhead, of course).
Now, really... can *anyone* here raise their hand and say that they could actually *utilize* 2200 megabytes/second of bandwidth to the outside world, either via network or disk?
But PCI is not used solely for outside bandwidth. It is a bus, and shared. Everything uses a little bit- down to your legacy I/O.
Now, when it *IS* replaced, I'd much rather see the interconnects being optical, not electrical. Instead of cracking open the case, shutting off the power, and trying to wedge yet another card inside (especially in low-height rackmounts), I'd much rather set the device on a shelf, and run a fiber patch cable over to the computer. No shutting down, and a whole lot more simple.
PCI *IS* being replaced. PCI-Express is the next stop. It is far better electrical signalling wise, uses a whole lot fewer wires for the same bandwidth. Hot swap already does exist for PCI- you have to go to higher end servers to support it, but it does exist- the biggest limitation is the O/S support. As to external chassis based peripherals you can hot swap- It was tried with Infiniband, but it never really caught on. You can see how far Infiniband has gotten in the past year. (not far) PCI-Express is basically Infiniband-lite. They took away some of the more ambitious goals and made something that was a lot more implementable and realistic.
I'd really like to see more optical interconnects- but the infrastructure is just too expensive. People aren't willing to buy it, and far too expensive to put on a motherboard (Are you willing to pay $1500-$2000 per motherboard?)
PCI-Express is *NOT* compatible (in a hardware sense) with PCI. Never meant to be, will never do it. Different connectors and so forth. From a software sense, it is supposed to look just like PCI.
Accurate intertial sensors are actually really expensive- an inertial sensor doesn't measure speed, it measures acceleration. if you use the simple eqation of d=.5*a*t^2, and say, an error of 1/1000 g, over an hour, you would be 63.5 kilometers off. I didn't think it was that bad- but I double checked my math.
Some of us still run mailing lists to connect a group of friends- who pays then? It is a perfectly legitimate use... but it seems scary if I'm would have to register my mailing list to get an 'exemption'
I think the biggest failing in this is that to tax email would require a massive change to the email infrastructure- just send all email through your government approved relay. Sure- they won't look at it... putting this on top of SMTP- I don't think it would work- what would be the incentive to use it (other than possibly spam free email)?
It's going to be really hard to make a whole jack/magnetics/controller/processor much smaller. You've got the absolute limitation of the RJ45, plus the magnetics- which is absolutely necessary for ethernet. You can get pretty close with capacitive isolation, but you're probably never going to be fully compliant with the IEEE spec. This device has contained in something about the size of a ethernet integrated mag/jack that has the whole shebang- pretty good.
You're going to have a hard time doing a 100Mbit interface that is truly "low power." With 100Mbit, there is always something going over the link, putting +/- 1V over a 100 ohm load, counting inefficiences, you're probably at 40 mA just to support the TX portion of the PHY. Then you have to realize that you need a 125 MHz clock going on inside-and that's all before you have a MAC and a processor going. Ethernet (particularly 100Mb) is not a low power interface.
If you want to keep the customer, the first time it happens, you might want to forgive the excess bandwidth charges (while pointing out the specific clause in the contract that says you have every right to charge them), tell them that it's "for this time only," and make a record of it. This is the type of action that can inspire customer loyalty. If you want to keep customers, you need to find some ways to differentiate yourself from all your competitors. Since you're keeping records, you should be able to tell if a customer is just trying to abuse your policies.
You need to ask yourself- how much did the excess bandwidth really cost, and how much is this customer worth to me in the long run? Probably, keeping that customer will make far more impact on your company in the long term than if you charged them, pissed them off, and inspired them to switch to another ISP.
They have the right, but, by law (in the USA) if they make a negative decision because of the credit report, they have to inform you of that. This is often overlooked. There was a report on this on NPR recently (Jan 31st, All Things Considered).
At the very least, you should check your credit report to make sure it is accurate.
While you can monitor your car pretty easy with devices to read things like the OBD, but actually changing how it works is pretty difficult. The car companies have some obligation to the EPA (in terms of making sure their cars are (somewhat) clean, and keeping them that way, by making it difficult for the average joe to twiddle with his ECU like he used to be able to twiddle his carbeurator. From the factories, cars are tuned for the optimum performance/emissions that satisfy the EPA and the marketers, virtually anything you do to improve performance will make emissions worse.
Some carmakers have gone to the extreme to make it difficult to change the ECU maps- Ford's (now obsolete) EEC-IV used a special version of the Intel 8051 (the 8061) and EEPROM that Intel (and their 2nd sources) were *only* authorized to sell to Ford. There were some modifications that could be done to improve performance, but those were a kludge that used a diagnostic port. [I worked as a co-op for Ford in the late 80's]
If you really do want to tinker with your car's engine, look to cars with engines that have been around a very long time- such as the Chevy 350, people have been tuning and playing with the 350 for decades, even as it has gone from carbs to EFI. There is a huge installed base and tinkerers are everywhere.
I worked on one of these- when I worked for Andrew corporation, I did some hardware and coding for the Extensis (mentioned below). Trying to get it to act like a landline was pretty tough- there is no really set way to determine when the call should actually go through- I ended up with a complex state machine based on a bunch of inputs, like how many numbers dialed, pause between the digits, and what number was dialed. (911 gets you right through, for example) The biggest downside I see is that 911 doesn't give the operator your address instantly in case of an emergency (yet, at least). Also, you're going to be pretty limited in the number of telephones that the device can power- You're probably not going to get one of these cradles that will drive the normal phone company max load of 5B (about 5 normal old telephones with mechanical ringers)
You can't use a digital phone for any sort of modem, but otherwise, it should work great. The first generation I worked on was for a motorola Star-Tac type phone. These are pretty complex little devices.
Electromagnetic radiation coming from a discrete source in open space (not a wire that goes to (effective) infinity like a power line) is always proportional to 1/r^2- it may radiate more power, but that's a factor of the radar cross section- remember I said proportional? There may be a constant in front of the 1/r^2. There are other factors like antenna gain and antenna power, but those are part of the constant- Since I have Skolnik's book in front of me: Pr=(Pt*G*Ae*sigma)/((4*pi)^2)*R^4, where Pr is power at the reciever, Pt is transmitted power, G is the antenna gain, Ae is the antenna aperture size, sigma is the radar cross section, and R is the range to the target.
Flipping through the book some more, I think I've found the real source of your inability to swallow- you're thinking in terms of light- where the wavelength is incredibly small in proportion to the size of the target (called the optical reigon in radar cross section-ese). Radars typically operate in the Rayleigh region, where the radar cross section varies as wavelength^-4. In between the Rayleigh and the optical region is the Mie or resonance region- we don't want to go there- it's really hard to work there, because the RCS can vary by 400% with a small change of the target (angle to the radar, typically).
Nope, if it's a monostatic radar (TX and RX at the same place)- the path from the radar to target is r, so power incident upon the target is C/r^2, (C is a constant that relates to power, antenna gain, and so forth). You multiply that times the radar cross section to get the amount of power reflected and then multiply that by 1/r^2 again to get the incident power back at the radar. The power at the reciever is proportional to 1/r^4.* The r^4 term is one reason that jammers have such an advantage (their power only drops off at 1/r^2).
If you're really interested in it, check out Merril Skolnik's book, "Introduction to Radar Systems," Chapter 1 (I think) where he generates the "radar equation."
* Of course, this is all assuming that you are far enough away from the antenna to be in the far field, but if you're not, you really don't have to use a radar, you could probably hit the target with a well thrown rock.
the return signals are weak and fall off 1/r^2 limiting the range.
Ok, this is nitpicking, and the poster makes a strong point, but I think that there is one point that needs to be corrected, which makes the situation even worse- the transmitted signal falls off at 1/r^2, and the reflected signal falls off at 1/r^2, so the radar return actually falls off at 1/r^4!
When it comes to safety issues, the regulatory bodies are very conservative, and because of the massive amounts of infrastructure it requires, changing things will take a long, long time. There is a finite amount of useful (right now) spectrum, and we have to carefully balance convenience and consumer demand (Wi-Fi) vs. safety.
Fortunately, here in the states, motorcycle drivers are required to take a safety course teaching the proper handling of a motorcycle in difficult situations.
Ummm... motorcycle drivers are not required to take a safety course in the states (that is, if you are referring to the United States of America) That sort of thing is set by the state anyway. Motorcyclists are encouraged to take a safety course, and doing so makes it much easier to get a license. I believe that the laws elsewhere (particularly in Europe) enforce much stricter licensing requirements (such as graduated licensing). I used to be a certified Motorcycle Safety Foundation instructor (I let my certification expire- too much work in miserable conditions for too little pay).
There are some states that require you take a motorcycle safety course if you are under 18 and want to get a motorcycle license, but if you're over 18 in most places, all you have to do is pass the test(s).
I thouroughly recommend the MSF courses for anyone who is even thinking about getting a motorcycle- they are the best form of instruction we have. They have proven techniques and a demonstrated track record. One of the most important concepts stressed is that motorcycles are dangerous and you must accept the risk, or not ride. They teach ways to reduce the risk, but it can't be eliminated.
Sole source parts are always a bad thing to have, but if they are the normal jellybean parts (resistors, capacitors, transistors, etc.) you can usually work around and find an adequate substitute in a reasonable amount of time. On the other hand programmable parts tend to have unique code- that can't easily (or quickly) be transferred from one part to another.
A case in point from my own experience about 2.5 years ago- flash memory had a huge upswing in demand- I believe it was from cell-phones. It was so lucrative, that Atmel switched its fabs over to producing lots of flash memory, and putting us microcontroller users on allocation- we went from a 6 week lead time on production quantities to a 6 *month* lead time in a matter of weeks, and even then, they wouldn't guarantee us parts- it was more like 6 months to get on the list to maybe get parts. Microcontroller code doesn't port nearly as easily as higher level code- you tend to have to use every last resource.
This caused a good number of manufacturers to biased against Atmel- they definitely have their good points, but if you can't get them, they're useless. Unless you're a really big company, it is hard to get continuity of supply agreements. I know that even now (working for a really big company) I hesitate to specify Atmel micros.
Ok, I've got some experience here- I've been testing and designing Ethernet for one of the major computer manufacturers for a couple years now. Most manufacturers of Ethernet use shielded connectors (jacks). This shield is what connects to your shield on shielded RJ45 cable. On every shielded ethernet port I've seen, the shield connects to chassis ground- which is the same ground as the ground pin on the power cord. The differential pairs themselves (according to the IEEE spec) should be electrically isolated with a transformer to take up to 2250 VDC or 1500 VAC.
In a situation where you're not sure if you're running the two ends off of a common ground, I definitely would not recommend using shielded cable. You could end up sending that cable up in smoke if your grounds shift.
I've worked for two different government funded labs, so I'm a bit biased, but I think that government funded research is vital for the long term health of our country.
The great thing about working in government funded research is that you have the ability to fail. Failure can be good. Unfortunately, in the commercial world, failure is bad, and must be avoided at all costs- if you fail, you go out of business. But if you don't fail sometimes, you're not pushing the envelope hard enough, not taking any risks. Unfortunately, there has been much movement in the government to divert resources to private industry- to people looking for short term profit.
Private industry looks for short term gains- long term is 5 years, 10 years (or more) out is just too far. The government can afford to look that far out, or farther. That is where the neat stuff happens.
Someday, I'll be back there, back to making cool stuff, and trying to avoid the politics as much as possible.
Really, the thing that needs to be done is improve the *protocol* that the communciation rides upon. If you're using gigabit, usually the limiting factor is the processor- you max the processor (doing the overhead for TCP/IP) before you even come close to filling up the gigabit pipe. There are technologies like Infiniband which try to address that, but progress has been very slow.
There is a big push for TCP offload- to take the effort for TCP and put it into an ASIC- check out this paper. TCP offload is absolutely necessary if we want to go beyond 1 gigabit ethernet any time soon, because the rate that we can communicate is beating the rate of increase in processing. IIRC Moore's law is an 18 month doubling time, there is a similar law for communications speed that says it doubles at a 12 month rate.
In addition, the engineering hassles of making a device with so many paralell channels is pretty hard- When you start doing high speed communications, you have to take a *lot* of care to make sure that the signals get there as you want them. This is why Intel appears to be in the process of moving away from PCI (a paralell bus) and moving to PCI-Express/3GIO, a combination paralell/serial bus, with a variable number of serial channels. PCI-Express is designed to carry on 16 wires what PCI-X carries on 80+ wires. When you actually have to make hardware- this makes a *HUGE* amount of difference.
What if you have multiple remotes?
I read some technical literature on code-hopping remotes- you can have multiple remotes and so forth. A code hopping remote is a one-way device- it only transmits, and it transmits a different code each time you press the button. The reciever knows what the code was the last time you pressed the button, so it knows what the next code should be (they use very, very long pseudo-random sequences).
So what happens if you press the remote when you're too far away, or your 3 year old finds it in the car and presses it merrily for a few miles until he gets bored? If the current code is code N, the next to be transmitted is code N+1, and the next that the reciever will key on will be N+1 to some range of N+X where X is pretty large, but still far smaller than Y, the total number of different codes.
In addition to keying on something in the next group of X codes, another safeguard thats used is to key on two successive remote-keys- lets say my transmitter is out of sequence, say by A - the reciever is looking for a code between N+1 and N+X- it recieves the code N+A (where N+1N+AN+X) It won't key on that, but if the next code it recieves is N+A+1, it knows that it got the right sequence, and will only open after the next keypress.
For multiple remotes, it adds a separate domain to key upon(say, Y to Y+N)- slightly less secure, but of course the more keys there are for any door, the less secure it is.
May seem kinda silly, but an old refrigerator door would work- magnetic, (relatively) airtight, positive seal. Positively pressurize the inside, with a few stages of HEPA filtering. You'll need to be pretty rigorous about filter maintenance. I heard of someone using refrigerator doors in an airlock to seal off a home recording studio. Even if it doesn't work, it would look pretty cool.
May not be good enough for silicon fab, but it probably would be good for people with serious allergies.
Also known as PCI-Express- but these have the difficulty of not being designed for long distances. PCI-express does have some advantages, like being able to choose the amount of bandwidth you want by choosing the number of pairs of wires, so it is scalable.
Whatever hardware interconnect comes next, it will only catch on if we have a big software effort to support it- I'd like to know which interconnect the software community may rally behind.