Signals propagate differently when wires are set up as transmission lines - they propagate at much closer to the speed of light, because you're actually sending a wave down the line (imagine creating a ripple on a trough of water, instead of actually filling and emptying the trough).
Last time I checked, speed of electron flow is only based on the material around it. Higher dialectric constant = lower speed of propgaition. Transmission lines aren't voodoo science, they are a property of the electrical length of the line and the rate of change of the signal on that line. It does not change the rate of propagation at all. Whether a given wire is 1" long, or 200 miles long, it will not change the speed of propagation.
I recently attended a seminar where the presenter talked about clocking based on LRC oscillations and he had actually fabbed chips that worked. The basic idea was to put an inductor on the die, and set up oscillations between the inductor and the clock load capacitance, which results in a ticking clock. Of course, you get a sinusoidal clock instead of a nice almost-square-wave, so your circuits have to be designed a little bit differently, but the point is, it works and is doable.
Not to be cheeky, but it's quite easy to change a sine wave into a square wave: Schmidt trigger. While I can't rule this out entirely, I would imagine that if it was more economical to produce an LRC resonator, it would be built into devices already. These circuits have been around for decades. It's very difficult to beat quartz crystals in terms of stability, ease of use, and power consumption.
You're half right. You're right that what's going on is a charging and discharging of a cap, but you're wrong that the charge can't be recycled. A conventional clock works by connecting the gates of a bunch of devices (i.e. capacitance) to Vdd, then after a little time connecting it to ground instead. Wait a little bit, then repeat. What effectively happens is that you dump some amount of charge from Vdd to ground each switch, and it's gone (i.e. it's heat now). A water analogy would be a tub of water above you (Vdd), a bucket in your hand (the capacitance), and the ground (gnd). You pour some water from the tub into your bucket (charge the cap), then dump it on the ground.
Wrong. The clock drives into a high impedance node. (The CMOS receivers on the other side of the clock line). CMOS drivers do have the problem of connecting to ground temporarily during switching - more akin to spilling some of the water out of the bucket as you pour it, not pouring it entirely on the ground. This can be overcome using clocks that are 90deg out of phase. And if the cap that you're talking about is the 10pF or so that is on the gate of the reciever CMOS - there are larger fish to fry power wise than this minimal capacitance. Try taking on the bulk leakage at 90nm before taking on this minimal source of power dissipation.
I don't agree that SW guys are OS agnostic, nor are they programming language agnostic. You'd have just as much of a hard time getting an expert in C# and.NET to use Linux as you would an expert in C to use WinCE.
Yes, of course you choose your device as to performace level, but it's not as relevant now. These days, the same $10 will buy you an ARM9(32-bit RISC), or a 68HC12. Moore's law has done amazing things to the low-end processor market. But you might take on the extra hardware cost to reduce your time to market and subsequent maintainence costs by choosing a different OS.
This is what I mean by properly architecting the product.
There's a fundamental misunderstanding here - GOOD Hardware engineers start out by asking "What OS vendor are you going to use?" before buying an eval board and dictating what OS to use by their choice of processor vendors.
It's true that once the OS and eval boards are selected, a BSP has to be created by one of those vendors. This is much further down the line and usually must be well thought out in order for a project to be successful. Working with MontaVista is a pain, though.
Not to be a shill for Cirrus Logic for a moment, but another route that I find interesting is the route that Cirrus is taking with their ARM9 processors. You can download a full BSP for free without having to go through these third party BSP vendors. http://arm.cirrus.com/
Ultimately, a system improperly architected will fail in cost, schedule, and/or reliability. Processor/OS selection is just one of those steps.
Your skin has a "break down voltage", much like a diode has. Past a certain voltage, your skin no longer provides much resistance (I don't have the exact values) and so as voltage increases, your "hand to ground" resistance decreases. This causes the current to increase exponentially, not linearly.
If you're talking about FPGAs from places like Xilinx, ALtera and Lattice, of course you can program them in the field. However, that's not the case with all FPGA's and all mfgrs. Actel has an entire line of FPGAs that you can only program once, just like the poster a couple of levels up was talking about. While the case is true in general it does not hold true across all products.
Of course, even this statement isn't entirely true. Actel is now working on an FPGA family which will be flash based and once they release this product the above statement will no longer be true. The only thing constant is change in the FPGA industry!
Agreed... I was talking about Modelsim for Linux on x86. Big difference.
As I said in a previous post, I'll be getting a demo of one of these in the next couple of weeks. I can post the real costs of all the software (linux or otherwise) then.
Nobody here has mentioned the LOEN progect, which is based on the SPARC V8. This is an open processor core that you can put into any FPGA. Speeds aren't as great as the PowerPC in this desing, but hey, it works!
Pehaps you have heard of a VHDL simulator called Modelsim? They have a Linux version and they have found that through test after test Modelsim runs much faster on Linux than on any other platform. That's why they are targeting Linux.
Actually, the ML300 has the Virtex 2 Pro P7 chip on it, which is only one Power PC core. However, if you wanted to, I'm sure they would sell it to you with the chip that can support the 4 processors, but that chip alone will cost you $6k.
I guess I'm one of the fortunate ones that is actually getting a sample to play with. Xilinx is coming to visit next week and is dropping one off. I'll tell you how it goes.
It's not quite that simple. As has been said before, the majority of the power dissipated in the design is done on the clock edges, not while the transistor is on or off. The physical properties of CMOS, in general, have a very high "static" resistance, but a very low "dynamic" resistance. The "static" resistance (what EE's call DC input impedance) is on the order of 1 megaohm. This would mean at 1.5V, the "static" power dissipation is about 2.25nW (a very, very low number). OTOH, the "dynamic" resistance (what EE's call AC input impedance) is on the order of 50 ohms. At 1.5 V, that's about 50mW or about 20 million times more power dissipation. Now fortunately, the time that it takes to switch the transistor is very short, so the dynamic power dissipation isn't continuous, but it significantly adds to the power dissipation (and that's why Intel's 2.8GHz chips run so hot. So, on to your analogy...
In your example, the synchronous design would use more peak power but, they both use the same amount of total power and the asynchronous design would be faster. Let's say that each of your 4000 transistors switched at 1ps (not physically realistic at this point, but it's your example, not mine!;-) After being "on" for 3ps, your 1000 transistors switched, then after 5 ps, the other 3000 switched. In this case, we go to 6ps, so we can account for the switching time of the second batch of transistors. To make this a bit clearer, the 1000 transistors are "on" for 3ps, switching for 1ps, and then "on" again for 2ps, for a total of 6ps. The 3000 transistors are "on" for 5ps and thesn switching for 1ps. If we calculate the total power for the asynchronous design, it turns out to be: for the batch of 1000 transistors: 1000*2.25nW*3ps + 1000*50mW*1ps + 1000*2.25nW*2ps = 50.00001125W*ps (by the way, it is clear here that most of the power dissipation is the "dynamic" or "switching" kind) and the batch of 3000 transistors: 3000*2.25nW*5ps + 3000*50mW*1ps = 150.00003375W*ps. For a total of 200.000045W*ps. For the synchronous design, it would be: 4000*2.25nW*5ps + 4000*50mW*1ps = 200.000045. However, it is clear that the peak power for the asynchronous design is much lower.
So, to answer your question... No, it doesn't take any more power, and the asynchronous design is faster.
As a person who was laid off last October from a optical networking company, I have seen these types of people before. Here are their misconceptions about finding a job, and how you can keep from falling into the same traps.
First of all, there is no way that you can use the same method of finding a job as you did in the late 90's. You can't just call some recruiters and post your resume on Monster. You have to get out there and talk with people. Yes, it sucks, especially for an engineer like me. However, you have to get you name out there. Call people weekly about positions you know are open. Ask if you can come in to talk to them about the position. If you don't get the job, ask for a meeting with the hiring manager to discuss where you need to develop and what you need to do to get a job in that company in the future. For the companies you'd like to work for, research companies and call them about products that they're working on. Ask to make an appointment to come down and talk to them. Make yourselves some business cards (Avery makes some nice perforationless that almost look professional) and hand them out to everyone you meet.
Look, there are jobs out there, but companies aren't falling all over themselves to find employees anymore. They don't need to print ads or hire recruiters, because the people are already coming to them. I know that you've heard it a hundred times, but only 20% of jobs are found in the clssifieds, 7% on the internet, 5% by recruiters, and only 1% by mass mailing your resume. That means that you are missing out on 2/3 of the jobs by only using these sources.
It's taken me 5 months to realize these truths, but now I'm on my way and I have a job offer. You can, too.
"There are three types of lies: Lies, damned lies, and statistics... " --Mark Twain
Assuming the ASA's report is correct, it is a non-issue. Anyone who was in school during 1990-1996 could tell you that the drop in enrollment was systemic and had nothing to do with high-tech careers.
Check the data yourselves... Enrollment in these classes (especially the class of '96) was at a historical low. The reason? Not as many people. In 1974, the number of people born was less than any time before World War 2. No wonder there wasn't as many students in the high tech fields, there weren't as many STUDENTS.
What does this mean? It means that it's a mere population blip, and that's all. Will there be a shortage? Probably not.:-( Guessing by how skillfully the ASA chose thier data set leads me to believe that they knew the conclusion they wanted before they looked at the data...
Slashdot Mirror
I love the smell of wrong Physics in the morning.
Signals propagate differently when wires are set up as transmission lines - they propagate at much closer to the speed of light, because you're actually sending a wave down the line (imagine creating a ripple on a trough of water, instead of actually filling and emptying the trough).
Last time I checked, speed of electron flow is only based on the material around it. Higher dialectric constant = lower speed of propgaition. Transmission lines aren't voodoo science, they are a property of the electrical length of the line and the rate of change of the signal on that line. It does not change the rate of propagation at all. Whether a given wire is 1" long, or 200 miles long, it will not change the speed of propagation.
I recently attended a seminar where the presenter talked about clocking based on LRC oscillations and he had actually fabbed chips that worked. The basic idea was to put an inductor on the die, and set up oscillations between the inductor and the clock load capacitance, which results in a ticking clock. Of course, you get a sinusoidal clock instead of a nice almost-square-wave, so your circuits have to be designed a little bit differently, but the point is, it works and is doable.
Not to be cheeky, but it's quite easy to change a sine wave into a square wave: Schmidt trigger. While I can't rule this out entirely, I would imagine that if it was more economical to produce an LRC resonator, it would be built into devices already. These circuits have been around for decades. It's very difficult to beat quartz crystals in terms of stability, ease of use, and power consumption.
You're half right. You're right that what's going on is a charging and discharging of a cap, but you're wrong that the charge can't be recycled. A conventional clock works by connecting the gates of a bunch of devices (i.e. capacitance) to Vdd, then after a little time connecting it to ground instead. Wait a little bit, then repeat. What effectively happens is that you dump some amount of charge from Vdd to ground each switch, and it's gone (i.e. it's heat now). A water analogy would be a tub of water above you (Vdd), a bucket in your hand (the capacitance), and the ground (gnd). You pour some water from the tub into your bucket (charge the cap), then dump it on the ground.
Wrong. The clock drives into a high impedance node. (The CMOS receivers on the other side of the clock line). CMOS drivers do have the problem of connecting to ground temporarily during switching - more akin to spilling some of the water out of the bucket as you pour it, not pouring it entirely on the ground. This can be overcome using clocks that are 90deg out of phase. And if the cap that you're talking about is the 10pF or so that is on the gate of the reciever CMOS - there are larger fish to fry power wise than this minimal capacitance. Try taking on the bulk leakage at 90nm before taking on this minimal source of power dissipation.
DAMN! I missed my queue!
I don't agree that SW guys are OS agnostic, nor are they programming language agnostic. You'd have just as much of a hard time getting an expert in C# and .NET to use Linux as you would an expert in C to use WinCE.
Yes, of course you choose your device as to performace level, but it's not as relevant now. These days, the same $10 will buy you an ARM9(32-bit RISC), or a 68HC12. Moore's law has done amazing things to the low-end processor market. But you might take on the extra hardware cost to reduce your time to market and subsequent maintainence costs by choosing a different OS.
This is what I mean by properly architecting the product.
There's a fundamental misunderstanding here - GOOD Hardware engineers start out by asking "What OS vendor are you going to use?" before buying an eval board and dictating what OS to use by their choice of processor vendors.
It's true that once the OS and eval boards are selected, a BSP has to be created by one of those vendors. This is much further down the line and usually must be well thought out in order for a project to be successful. Working with MontaVista is a pain, though.
Not to be a shill for Cirrus Logic for a moment, but another route that I find interesting is the route that Cirrus is taking with their ARM9 processors. You can download a full BSP for free without having to go through these third party BSP vendors. http://arm.cirrus.com/
Ultimately, a system improperly architected will fail in cost, schedule, and/or reliability. Processor/OS selection is just one of those steps.
This fix works under 1.0 (Other versions not tested)
Install Tabbrowser extensions loacted here:
http://texturizer.net/firefox/extensions// [Texurizer.net]
Look for the "Tabbrowser Extensions" under "Tabs and Windows" (it's about 1/8 down the page) Other extensions may fix the problem as well.
Girmann
Unfotunately, you assume one very wrong thing...
Your skin has a "break down voltage", much like a diode has. Past a certain voltage, your skin no longer provides much resistance (I don't have the exact values) and so as voltage increases, your "hand to ground" resistance decreases. This causes the current to increase exponentially, not linearly.
So you are absolutely not safe in touching 240V.
Girmann
If you're talking about FPGAs from places like Xilinx, ALtera and Lattice, of course you can program them in the field. However, that's not the case with all FPGA's and all mfgrs. Actel has an entire line of FPGAs that you can only program once, just like the poster a couple of levels up was talking about. While the case is true in general it does not hold true across all products.
Of course, even this statement isn't entirely true. Actel is now working on an FPGA family which will be flash based and once they release this product the above statement will no longer be true. The only thing constant is change in the FPGA industry!
Agreed... I was talking about Modelsim for Linux on x86. Big difference.
As I said in a previous post, I'll be getting a demo of one of these in the next couple of weeks. I can post the real costs of all the software (linux or otherwise) then.
Nobody here has mentioned the LOEN progect, which is based on the SPARC V8. This is an open processor core that you can put into any FPGA. Speeds aren't as great as the PowerPC in this desing, but hey, it works!
Hmm... I beg to differ.
Pehaps you have heard of a VHDL simulator called Modelsim? They have a Linux version and they have found that through test after test Modelsim runs much faster on Linux than on any other platform. That's why they are targeting Linux.
4 Power PC cores? Well, not quite.
Actually, the ML300 has the Virtex 2 Pro P7 chip on it, which is only one Power PC core. However, if you wanted to, I'm sure they would sell it to you with the chip that can support the 4 processors, but that chip alone will cost you $6k.
I guess I'm one of the fortunate ones that is actually getting a sample to play with. Xilinx is coming to visit next week and is dropping one off. I'll tell you how it goes.
In your example, the synchronous design would use more peak power but, they both use the same amount of total power and the asynchronous design would be faster. Let's say that each of your 4000 transistors switched at 1ps (not physically realistic at this point, but it's your example, not mine! ;-) After being "on" for 3ps, your 1000 transistors switched, then after 5 ps, the other 3000 switched. In this case, we go to 6ps, so we can account for the switching time of the second batch of transistors. To make this a bit clearer, the 1000 transistors are "on" for 3ps, switching for 1ps, and then "on" again for 2ps, for a total of 6ps. The 3000 transistors are "on" for 5ps and thesn switching for 1ps. If we calculate the total power for the asynchronous design, it turns out to be: for the batch of 1000 transistors: 1000*2.25nW*3ps + 1000*50mW*1ps + 1000*2.25nW*2ps = 50.00001125W*ps (by the way, it is clear here that most of the power dissipation is the "dynamic" or "switching" kind) and the batch of 3000 transistors: 3000*2.25nW*5ps + 3000*50mW*1ps = 150.00003375W*ps. For a total of 200.000045W*ps. For the synchronous design, it would be: 4000*2.25nW*5ps + 4000*50mW*1ps = 200.000045. However, it is clear that the peak power for the asynchronous design is much lower.
So, to answer your question... No, it doesn't take any more power, and the asynchronous design is faster.
--Girmann
As a person who was laid off last October from a optical networking company, I have seen these types of people before. Here are their misconceptions about finding a job, and how you can keep from falling into the same traps.
First of all, there is no way that you can use the same method of finding a job as you did in the late 90's. You can't just call some recruiters and post your resume on Monster. You have to get out there and talk with people. Yes, it sucks, especially for an engineer like me. However, you have to get you name out there. Call people weekly about positions you know are open. Ask if you can come in to talk to them about the position. If you don't get the job, ask for a meeting with the hiring manager to discuss where you need to develop and what you need to do to get a job in that company in the future. For the companies you'd like to work for, research companies and call them about products that they're working on. Ask to make an appointment to come down and talk to them. Make yourselves some business cards (Avery makes some nice perforationless that almost look professional) and hand them out to everyone you meet.
Look, there are jobs out there, but companies aren't falling all over themselves to find employees anymore. They don't need to print ads or hire recruiters, because the people are already coming to them. I know that you've heard it a hundred times, but only 20% of jobs are found in the clssifieds, 7% on the internet, 5% by recruiters, and only 1% by mass mailing your resume. That means that you are missing out on 2/3 of the jobs by only using these sources.
It's taken me 5 months to realize these truths, but now I'm on my way and I have a job offer. You can, too.
-=-Girmann-=-
Something which all of you /.ers have missed...
:-( Guessing by how skillfully the ASA chose thier data set leads me to believe that they knew the conclusion they wanted before they looked at the data...
"There are three types of lies: Lies, damned lies, and statistics... " --Mark Twain
Assuming the ASA's report is correct, it is a non-issue. Anyone who was in school during 1990-1996 could tell you that the drop in enrollment was systemic and had nothing to do with high-tech careers.
Check the data yourselves... Enrollment in these classes (especially the class of '96) was at a historical low. The reason? Not as many people. In 1974, the number of people born was less than any time before World War 2. No wonder there wasn't as many students in the high tech fields, there weren't as many STUDENTS.
What does this mean? It means that it's a mere population blip, and that's all. Will there be a shortage? Probably not.
--Girmann