Well, I find it funny people think it's easy to insert such things into a finished IC design. The first few batches of a large IC (think computer hardware) always tend to fail. While the foundry might run some tests of their own on the device this doesn't happen all too often. It's up to the designer to find the error in the design. Changes like the ones you're suggesting would require extensive modification, addition of extra layers, logic,... It'd be very difficult to add such things afterwards in a large complicated design without breaking it. Resulting in the designer spotting the modification. So unless the designer is in on it you won't see this happening anytime soon. Electron microscopes are a bitch to hide things from, especially if the operator is sufficiently skilled. Even if the required design modifications to be able to make the processor burn on command would be relatively simple it'd still be easy to spot such rough modifications afterwards, even on xrays.
The problem is firmware. If the designer makes it impossible to modify the firmware easily it's near impossible to put a bug in the hardware during manufacturing.
Get the necessary equipment and make your own CPU. Also make the lithography masks yourself to ensure your paranoia score reaches a maximum level! Next proceed to make your own motherboard (making all the components yourself as mentioned earlier). Also you'll have to create your own CRT monitor (imagine if they intercepted the signals between the graphics card and the monitor!!!). And you might want to sit in a faraday cage made out of mu metal with your own personal lemon battery based power supply.
ARM isn't always the right choice and it does have its problems. Additionally if you have to interface with an old system it's often easier to just grab a M68K or an old Intel 8xxx series device. The interface was already designed in a lot of cases for the older devices. And the M68K is advanced and fast enough to easily interface with modern hardware. So it actually does make for a pretty good bridge when you have to make two incompatible systems work together and don't want to go through the trouble of starting from scratch. Not to mention that the M68K is a great device to introduce people to the hardware side of embedded system design. Fairly cheap, comes in easy to solder packages unlike most ARM processors, robust, well documented, loads of software has been written for it,...
Totally worth the effort! It's not because something is old that it's not worth using anymore. Look at the Intel 8051 architecture. You'll find several microcontrollers based on that architecture in your house on this very moment. Sure it's an ancient 8 bit CISC architecture, but most designers are very familiar with it and it's one of the cheapest microcontrollers available so it still sees quite a lot of use. Fun fact is that it's commonly used as USB host controller.
Well, if the provider uses a copy of the blacklist to simply start blocking spammers it is the blacklist who's responsible for blocking people. Your core assumption is that every system in use is well made and properly designed. Reality is that most systems are horrible in structure and very illogical. Problems only get solved if there are sufficient complaints. Funny problem is contacting the target if you've been blacklisted. They often use the same blacklist to filter their own emails so you need to use a gmail account or something similar to get through to them.
Well yes, but you make one assumption: that everybody runs a well built system.
I've found that assumption to be incorrect, even large corporations with huge IT departments often take the lazy way out when it comes to filtering emails and will just load blacklists as block lists.
So you call BS on me cause I use the wrong terminology according to you? They do in fact block people, in many instances the blacklists are automatically loaded and many providers do use them cause of the spam problems they're experiencing. They're a very cheap solution to a major problem. Not everybody wants to dish out a lot of money for the latest in smart anti-spam software or hire somebody on staff to constantly update their own anti-spam rules.
Claim it's not a major problem? Setup a mail server on a new domain, create a random email address and publish it on a site with a fairly page ranking on google. To give you an idea: I have received over 100 spam emails in the last 24 hours on my regular email account, and I don't even spread the address of that one around. Sadly it turns out I'm not very interested in viagra, penis enlargements, huge fake DHL invoices, Nigerian princes, UN funding,... On the other hand I know this nice person who claims to be able to double your money in only 1 month! Oh wait...
By all means, take them to court in Europe. These is unfair trade practice. For that alone you can get pretty severe fines. Get a preliminary injunction as well, if possible with a nice daily fine attached to it. If they want to play it like that you should too. We had the same thing happen to us a while back (large IRC network). They blacklisted our mail server so our services couldn't email the users anymore to verify their email address. We threatened to get a preliminary injunction against them and they backed down very quickly. It took a total of 5 minutes between our lawyer sending an email and us being removed from the blacklist.
The RF communications aren't that hard depending on the frequency range you're allowed to work in. Main problem I see is that you'll probably have to go for GaAs based amplifier designs, they tend to perform better in hostile environments like space. But GaAs devices have a few nasty things about them, they tend to oscillate at frequencies you wouldn't expect at the start. Combined with the better characteristics in general they're a good choice for spacecraft. Avago has a few nice GaAs power FETs. Combined with a directional antenna you should have no problem getting your data back to the ground.
Fact is you can get space grade electronics from most manufacturers now, you might want to look into those. They're a bit more expensive but very reliable. Your main problem will be separating essential from non-essential systems. Like it does pay to have a separate microcontroller take care of the commands received and have something like an ARM processor up when you need processing power. For the rest the usual design guidelines: avoid loops, use a large ground plane, etc.
Sure it's possible, it's just not economical. Most SMT pick and place machines can go very precise, but then they slow down to a crawl. That's unacceptable in production, it's cheaper to have humans repair the mistakes the machines make. And it are mechanical limitations, not electrical. The advancements in that field are a lot slower. Not to mention the wear and tear on the machines that screws up the settings. Your time between calibrations will decrease rapidly as you increase precision.
And yes, I am serious. Do you actually do machine vision related things? I've had to do more than I'd like to in that field, and I can tell you it's a mess. It's good enough to say "ok" and "not ok". It'll even tell you where the mistake is, at least for verifying PCBs. But try detecting things like faces, it's a joke to claim it's reliable. Still way too high error margins for production. You don't want to trash correct boards, another problem with verifying a board electrically is the complexity, a device like an iPhone is simply impossible to completely test. You can test a large portion of the connections through JTAG debuggers and probing stations. But probing stations have a limited life time and do break, and they do have their limits, especially with the advent of BGAs. X-Ray doesn't verify electrical properties, SEM is too slow,... It's hard to verify a design. And even if your test sequence detects an error it's often hard to tell where it is. Like in the case of JTAG you're sometimes able to say if analog components are there or not, but you won't be able to verify their value. So the program is incapable of saying where the problem is even if it has the best pattern recognition software in the world. The problem you have is grasping the complexity of the problem. Electrical testing takes valuable time, you can't just test for less than a nanosecond, you need to put the device on the probing station, wait for the contacts to make a good contact, apply values, wait, apply new values, wait... Takes a few seconds for very complicated boards. So even with your pattern recognition software it's still useless. Now lets go to test manufacturing equipment (cause that one illustrates the problems easily). Your network analyser sends out a signal, gets sent through the power splitter, etc... When it's captured it'll go through a few ferrite rings first for filtering. You wrap a coaxial cable through a ferrite ring, you can verify this characteristic with another network analyser, sure. But now comes the problem: something's wrong. Your probing station isn't precise enough or fast enough and your camera's can't see in 3D around the ferrite ring to find the problem. Your pattern recognition software is once again useless because it can't even get the right data. You just need a human. Phones are also very complicated (sensitive high frequency parts) and face many similar challenges, they're just not as profound and clear to explain.
Mass producing complex precision equipment is near impossible, especially if there's a wide range of things that could possibly go wrong. You can never write software that's as reliable as a human to look at weird fault patterns. The human brain is wired to find patterns, a computer isn't. And it's very hard to program a computer for said job.
And in niché markets -like T&M- mass production is impossible, too low quantities to start with, and even if they were big enough you still run into the problem of manual calibration. If you're building something on the cutting edge of what's possible in terms of precision you can't build a device to test it, you need a human operator to verify its function based on very indirect tests in a lot of cases.
Automation, cheaper? You wish...
Most of the cost of electronics comes from testing, and the automated test equipment costs a fortune. Yield drops leading to higher repair costs to avoid wasting too much resources increasing the cost further.
Yeah, people underestimate the cost of proper automatisation. Good test and measurement equipment costs a fortune. Just to test and tune the radio on each of the devices you'll need a spectrum analyser and a network analyser. The former sells for at least 20000 if you need something reliable and quick, the latter doesn't sell for less than 25000. And you'll need several of those as well. People need to realise that most of the cost of electronics actually comes from testing the equipment and parts. The yield figures drop dramatically as complexity increases as well. And all of this needs to be charged for in the end device.
More than that as well.
Even for putting components on a PCB a machine is often insufficient. Our current SMT pick and place machines are fast, and they're very precise considering the speed they work at. But if you go to the small component sizes you either sacrifice yield or production time (increase in cost). You'll need to test every single one of the devices if you sacrifice the yield (Apple already did this). In the latter case the cost will increase by a significant factor. In either case you'll need to employ a lot more humans, and Apple isn't going to suddenly fix the precision vs. speed problem unless they come up with a new way to build servo systems.
Another problem is that machines can't easily put together certain things like lens assemblies. Not to talk about fixing the devices that fail the post-production tests, if you don't fix those you're going to lose a lot of money. So you need to employ people to fix those. A probing station and a computer might be able to tell you where the error is, but desoldering the components is work for a human no matter what you try.
Exactly, all the mandatory diversity rules actually qualify as racism in my opinion. Simply get the best speakers you can get if you're organizing a conference.
I honestly don't care how they look, what their believes are, etc... As long as they have something interesting to say!
You're dead wrong. Most of the things we make are still based on very simple microcontrollers.
There's a 8051 in almost every single device out there as a USB host controller. It's simply a very simple and common design, very useful and easy to modify according to your needs. Simplicity wins over complexity in most applications. Simple systems are less likely to break and easier to debug. We only add such functionality if it's found to be necessary.
You can't, you can model it, you can try to simulate parts of it. But a design with thousands of transistors isn't very feasible to simulate. That's why you model it layer by layer and abstract further and further.
Well, you have to assume the manufacturer made the IC correctly (and this is usually the case, especially for simple things like those microcontrollers).
Simulation software is available for all of those, based on that you can make a model on RTL level that reads a binary memory dump. The model will behave exactly like the microcontroller assuming you write it correctly. And if you resort to the 8051 (like you should in a lot of cases to be frank) you can find hundreds of VHDL models for free on the internet.
Actually I disagree on that, if you're working with digital I/O you can model most of the things in VHDL and Verilog quite well as long as you stay under the 200 MHz line, then you have to start looking at transmission lines and then it does become complicated again.
All models are abstractions and linearizations of reality. Assuming that your simulation software will tell you exactly if it's work is a pipe dream. Even if your math is correct and your simulation says it'll work you'll often run into problem. Even if it's something stupid. Like a few years ago we spent a week debugging a stupid OLED screen. Turns out that we had to move a PCB trace 1mm to make it work, but in that position it was causing too much noise. The manufacturer of the OLED panel never mentioned that in their datasheets or application notes. Nor did they actually know it until we contacted them. Needless to say simulation was useless at finding that problem.
Another area particularly fond to me is MRI coil design. Good luck simulating that. There's so much going on, the system is too complex to model. You can only give a guess at what's going to happen. Yes I can design the coil and transmission system adhering to the necessary specifications. But it turns out if you put it next to a high power transmitter inside a 3T magnetic field you'll run into things behaving not quite as expected. And no amount of simulation ever really catches those weird little quirks (we even modeled the inside of the transistors we used in the pre-amplifier in case you were wondering). There are just too many obscure effects that you can't quite predict until you put it together and try, even when you test it part by part in simulation. It's more than about bandwidth and amplification mind you. The NF has to be correct, you have to avoid coupling, and so on. Though it has made me a better designer by a huge margin in the long run. But that's experience preventing the errors instead of simulation and testing.
I got several of those LED strips and hang them just above my work area from all directions. Pretty much eliminates all shadows, no noise (the power supply is a few metres away), and the only problem I can think of is that it isn't truly white light.
Yes and no. In the case of digital logic you can write a model in VHDL, Verilog or whatever floats your boat. And then do functional verification on that. Dump it in a FPGA or hope your computer is fast enough. But for analog circuits it's not as easy. SPICE can give you a pretty good idea, but it's not perfect. Even more advanced programs like ADS won't give you the full picture. As the final step you will always need to actually make a prototype and just hope it works.
So yes you can test small parts of the design, but a large design can't be simulated without significant computing resources.
Well, obviously you might want to avoid metal. You can get these great plates for lab table surfaces made from some sort of ceramic. It's heat resistant and pretty tough, which is really necessary if a SMPS decides to hit the self destruct button. For soldering, just get a wooden board to protect the surface from direct impact with a soldering iron.
Miniature drawer cabinets are important and actually rather expensive, especially those that can be stacked.
Good soldering irons (more than one!) are a must obviously. Get both an analog and digital scope with at least 2 channels each. More is better. Personally I like putting a computer near my electronics workbench to view schematics, considered investing in a large TV for that but I'm a bit short on cash for that.
You want several supplies, current limited and not. Isolation transformers, a good variac, signal/function generators.
Good to have as well are an impedance meter/Q meter, network analyser, spectrum analyser and logic analyser. Especially the latter is worth considering, you can get pretty cheap versions these days that you hook up to a computer. For the other devices I advice stand alone versions cause it's really a lot easier while measuring if you can play with the knobs to home in on what you actually need. If you have more than enough money also get one of those microcontroller programmers with several sockets, that thing has saved my life more often than not.
Anyway, good luck!
It seems people miss the point of calculus quite badly. Calculus allows you to prove or simplify quite a lot of trigonometry. Numerical calculation methods also depend heavily on it. Your entire computer is the work of calculus being applied to physics. So do pay some attention to it and you might learn a useful thing or two. On top of that your game engines also need realistic physics. Good luck doing that well without a three dimensional field of vectors.
But where advanced mathematics really show their real value in my personal opinion is in electronics. Analog electronics, especially high frequency and RF systems can't be described or even approached without it. All the "rules of thumb" find their backing in advanced math. Even something as simple as the Fourier transform requires calculus to explain correctly. And lets not forget about our friend Maxwell...
In short, in the current society where electronics play a major role in almost any device it might be a good idea to be able to approach the subject. But you need calculus for that. A nice bonus is that calculus has helped me the past few years in reducing the amount of taxes I pay. Cause calculating an optimal solution to a problem is also something calculus is an excellent tool for.
I won't even bother responding to the first remark considering how Slashdot has become. You know very well what I mean.
And you can claim what you want, I have had to work on designing photovoltaic devices myself on several occasions, hated every minute of it; I got so fed up with them that I quit my job.
So first of all, where did I say separately? It's simply a well known fact that semiconductor properties depend on environmental factors and those do include temperature, air pressure, humidity,... Obviously temperature is the most important factor, and even that one is often left out of the equation when they do these sort of calculations. And GaAs does come into play when you want efficiency. All the multi-wavelength/junction designs are based on gallium-arsenide or germanium substrata as far as I'm aware of. You do have some variation in dopant materials but I fear I'd be going a bit off-topic with this.
I'll grant you that the first link does have some points. But I do not agree with their calculations.
This being the article they seem to get most of their data from: http://www.bnl.gov/pv/files/pdf/abs_193.pdf
The actual cost of manufacturing the substrata are very hard to define. Due to the increasing demand the solar panel industry has in fact started producing its own substrates as well. So the actual energy usage per square metre increased. The 14% efficiency as it should be around 10% for the panels they mention. The insolation values used are questionable. And I have yet to see an inverter last for more than 5 years without any sort of defect. Both the inverters I have owned always failed after roughly 4 years of service. The one I built myself (bit over designed mind you) is now up and running for 6 years and 2 months. Additionally their assumption about the main source of power is questionable as well as they modelled it based on the European power grid. Most solar panels are actually made in South-East Asia where dirty coal plants without air filters are still in common use. And the efficiency of the manufacturing process is also significantly lower due to older machines. This study on the other hand assumes machines of which some are still cutting edge at this point. Combine all these factors and I don't see it getting anywhere close to the 4 years they wish to claim. Last estimate I've seen that was trustworthy in my opinion was around 7 years on average assuming half of the panels were put at the equator and the other half put at the latitude of Berlin. It might have dropped a bit, but that study didn't really include the costs of the support equipment (mounting frames, inverters, modifications to the power grid for load balancing,...).
The second link mainly seems to rehash what the first link said judging from the references at the bottom so not going to bother.
And we're talking about replacing the main power supply for the power grid. So yes, you will need a large amount of engineering. I doubt anybody has actually ever published a reliable article discussing the full environmental costs of dedicated solar plants.
Slashdot Mirror
Well, I find it funny people think it's easy to insert such things into a finished IC design. The first few batches of a large IC (think computer hardware) always tend to fail. While the foundry might run some tests of their own on the device this doesn't happen all too often. It's up to the designer to find the error in the design. Changes like the ones you're suggesting would require extensive modification, addition of extra layers, logic, ... It'd be very difficult to add such things afterwards in a large complicated design without breaking it. Resulting in the designer spotting the modification. So unless the designer is in on it you won't see this happening anytime soon. Electron microscopes are a bitch to hide things from, especially if the operator is sufficiently skilled. Even if the required design modifications to be able to make the processor burn on command would be relatively simple it'd still be easy to spot such rough modifications afterwards, even on xrays.
The problem is firmware. If the designer makes it impossible to modify the firmware easily it's near impossible to put a bug in the hardware during manufacturing.
Sadly the needs of software designers often differ from what delivers high performing hardware :(
Get the necessary equipment and make your own CPU. Also make the lithography masks yourself to ensure your paranoia score reaches a maximum level! Next proceed to make your own motherboard (making all the components yourself as mentioned earlier). Also you'll have to create your own CRT monitor (imagine if they intercepted the signals between the graphics card and the monitor!!!). And you might want to sit in a faraday cage made out of mu metal with your own personal lemon battery based power supply.
ARM isn't always the right choice and it does have its problems. Additionally if you have to interface with an old system it's often easier to just grab a M68K or an old Intel 8xxx series device. The interface was already designed in a lot of cases for the older devices. And the M68K is advanced and fast enough to easily interface with modern hardware. So it actually does make for a pretty good bridge when you have to make two incompatible systems work together and don't want to go through the trouble of starting from scratch. Not to mention that the M68K is a great device to introduce people to the hardware side of embedded system design. Fairly cheap, comes in easy to solder packages unlike most ARM processors, robust, well documented, loads of software has been written for it, ...
Totally worth the effort! It's not because something is old that it's not worth using anymore. Look at the Intel 8051 architecture. You'll find several microcontrollers based on that architecture in your house on this very moment. Sure it's an ancient 8 bit CISC architecture, but most designers are very familiar with it and it's one of the cheapest microcontrollers available so it still sees quite a lot of use. Fun fact is that it's commonly used as USB host controller.
Well, if the provider uses a copy of the blacklist to simply start blocking spammers it is the blacklist who's responsible for blocking people. Your core assumption is that every system in use is well made and properly designed. Reality is that most systems are horrible in structure and very illogical. Problems only get solved if there are sufficient complaints. Funny problem is contacting the target if you've been blacklisted. They often use the same blacklist to filter their own emails so you need to use a gmail account or something similar to get through to them.
Well yes, but you make one assumption: that everybody runs a well built system.
I've found that assumption to be incorrect, even large corporations with huge IT departments often take the lazy way out when it comes to filtering emails and will just load blacklists as block lists.
So you call BS on me cause I use the wrong terminology according to you? They do in fact block people, in many instances the blacklists are automatically loaded and many providers do use them cause of the spam problems they're experiencing. They're a very cheap solution to a major problem. Not everybody wants to dish out a lot of money for the latest in smart anti-spam software or hire somebody on staff to constantly update their own anti-spam rules. ... On the other hand I know this nice person who claims to be able to double your money in only 1 month! Oh wait...
Claim it's not a major problem? Setup a mail server on a new domain, create a random email address and publish it on a site with a fairly page ranking on google. To give you an idea: I have received over 100 spam emails in the last 24 hours on my regular email account, and I don't even spread the address of that one around. Sadly it turns out I'm not very interested in viagra, penis enlargements, huge fake DHL invoices, Nigerian princes, UN funding,
By all means, take them to court in Europe. These is unfair trade practice. For that alone you can get pretty severe fines. Get a preliminary injunction as well, if possible with a nice daily fine attached to it. If they want to play it like that you should too. We had the same thing happen to us a while back (large IRC network). They blacklisted our mail server so our services couldn't email the users anymore to verify their email address. We threatened to get a preliminary injunction against them and they backed down very quickly. It took a total of 5 minutes between our lawyer sending an email and us being removed from the blacklist.
The RF communications aren't that hard depending on the frequency range you're allowed to work in. Main problem I see is that you'll probably have to go for GaAs based amplifier designs, they tend to perform better in hostile environments like space. But GaAs devices have a few nasty things about them, they tend to oscillate at frequencies you wouldn't expect at the start. Combined with the better characteristics in general they're a good choice for spacecraft. Avago has a few nice GaAs power FETs. Combined with a directional antenna you should have no problem getting your data back to the ground.
Fact is you can get space grade electronics from most manufacturers now, you might want to look into those. They're a bit more expensive but very reliable. Your main problem will be separating essential from non-essential systems. Like it does pay to have a separate microcontroller take care of the commands received and have something like an ARM processor up when you need processing power. For the rest the usual design guidelines: avoid loops, use a large ground plane, etc.
Sure it's possible, it's just not economical. Most SMT pick and place machines can go very precise, but then they slow down to a crawl. That's unacceptable in production, it's cheaper to have humans repair the mistakes the machines make. And it are mechanical limitations, not electrical. The advancements in that field are a lot slower. Not to mention the wear and tear on the machines that screws up the settings. Your time between calibrations will decrease rapidly as you increase precision.
... It's hard to verify a design. And even if your test sequence detects an error it's often hard to tell where it is. Like in the case of JTAG you're sometimes able to say if analog components are there or not, but you won't be able to verify their value. So the program is incapable of saying where the problem is even if it has the best pattern recognition software in the world. The problem you have is grasping the complexity of the problem. Electrical testing takes valuable time, you can't just test for less than a nanosecond, you need to put the device on the probing station, wait for the contacts to make a good contact, apply values, wait, apply new values, wait... Takes a few seconds for very complicated boards. So even with your pattern recognition software it's still useless. Now lets go to test manufacturing equipment (cause that one illustrates the problems easily). Your network analyser sends out a signal, gets sent through the power splitter, etc... When it's captured it'll go through a few ferrite rings first for filtering. You wrap a coaxial cable through a ferrite ring, you can verify this characteristic with another network analyser, sure. But now comes the problem: something's wrong. Your probing station isn't precise enough or fast enough and your camera's can't see in 3D around the ferrite ring to find the problem. Your pattern recognition software is once again useless because it can't even get the right data. You just need a human. Phones are also very complicated (sensitive high frequency parts) and face many similar challenges, they're just not as profound and clear to explain.
And yes, I am serious. Do you actually do machine vision related things? I've had to do more than I'd like to in that field, and I can tell you it's a mess. It's good enough to say "ok" and "not ok". It'll even tell you where the mistake is, at least for verifying PCBs. But try detecting things like faces, it's a joke to claim it's reliable. Still way too high error margins for production. You don't want to trash correct boards, another problem with verifying a board electrically is the complexity, a device like an iPhone is simply impossible to completely test. You can test a large portion of the connections through JTAG debuggers and probing stations. But probing stations have a limited life time and do break, and they do have their limits, especially with the advent of BGAs. X-Ray doesn't verify electrical properties, SEM is too slow,
Mass producing complex precision equipment is near impossible, especially if there's a wide range of things that could possibly go wrong. You can never write software that's as reliable as a human to look at weird fault patterns. The human brain is wired to find patterns, a computer isn't. And it's very hard to program a computer for said job.
And in niché markets -like T&M- mass production is impossible, too low quantities to start with, and even if they were big enough you still run into the problem of manual calibration. If you're building something on the cutting edge of what's possible in terms of precision you can't build a device to test it, you need a human operator to verify its function based on very indirect tests in a lot of cases.
Automation, cheaper? You wish...
Most of the cost of electronics comes from testing, and the automated test equipment costs a fortune. Yield drops leading to higher repair costs to avoid wasting too much resources increasing the cost further.
Yeah, people underestimate the cost of proper automatisation. Good test and measurement equipment costs a fortune. Just to test and tune the radio on each of the devices you'll need a spectrum analyser and a network analyser. The former sells for at least 20000 if you need something reliable and quick, the latter doesn't sell for less than 25000. And you'll need several of those as well. People need to realise that most of the cost of electronics actually comes from testing the equipment and parts. The yield figures drop dramatically as complexity increases as well. And all of this needs to be charged for in the end device.
More than that as well.
Even for putting components on a PCB a machine is often insufficient. Our current SMT pick and place machines are fast, and they're very precise considering the speed they work at. But if you go to the small component sizes you either sacrifice yield or production time (increase in cost). You'll need to test every single one of the devices if you sacrifice the yield (Apple already did this). In the latter case the cost will increase by a significant factor. In either case you'll need to employ a lot more humans, and Apple isn't going to suddenly fix the precision vs. speed problem unless they come up with a new way to build servo systems.
Another problem is that machines can't easily put together certain things like lens assemblies. Not to talk about fixing the devices that fail the post-production tests, if you don't fix those you're going to lose a lot of money. So you need to employ people to fix those. A probing station and a computer might be able to tell you where the error is, but desoldering the components is work for a human no matter what you try.
Exactly, all the mandatory diversity rules actually qualify as racism in my opinion. Simply get the best speakers you can get if you're organizing a conference.
I honestly don't care how they look, what their believes are, etc... As long as they have something interesting to say!
You're dead wrong. Most of the things we make are still based on very simple microcontrollers.
There's a 8051 in almost every single device out there as a USB host controller. It's simply a very simple and common design, very useful and easy to modify according to your needs. Simplicity wins over complexity in most applications. Simple systems are less likely to break and easier to debug. We only add such functionality if it's found to be necessary.
You can't, you can model it, you can try to simulate parts of it. But a design with thousands of transistors isn't very feasible to simulate. That's why you model it layer by layer and abstract further and further.
Well, you have to assume the manufacturer made the IC correctly (and this is usually the case, especially for simple things like those microcontrollers).
Simulation software is available for all of those, based on that you can make a model on RTL level that reads a binary memory dump. The model will behave exactly like the microcontroller assuming you write it correctly. And if you resort to the 8051 (like you should in a lot of cases to be frank) you can find hundreds of VHDL models for free on the internet.
Actually I disagree on that, if you're working with digital I/O you can model most of the things in VHDL and Verilog quite well as long as you stay under the 200 MHz line, then you have to start looking at transmission lines and then it does become complicated again.
All models are abstractions and linearizations of reality. Assuming that your simulation software will tell you exactly if it's work is a pipe dream. Even if your math is correct and your simulation says it'll work you'll often run into problem. Even if it's something stupid. Like a few years ago we spent a week debugging a stupid OLED screen. Turns out that we had to move a PCB trace 1mm to make it work, but in that position it was causing too much noise. The manufacturer of the OLED panel never mentioned that in their datasheets or application notes. Nor did they actually know it until we contacted them. Needless to say simulation was useless at finding that problem.
Another area particularly fond to me is MRI coil design. Good luck simulating that. There's so much going on, the system is too complex to model. You can only give a guess at what's going to happen. Yes I can design the coil and transmission system adhering to the necessary specifications. But it turns out if you put it next to a high power transmitter inside a 3T magnetic field you'll run into things behaving not quite as expected. And no amount of simulation ever really catches those weird little quirks (we even modeled the inside of the transistors we used in the pre-amplifier in case you were wondering). There are just too many obscure effects that you can't quite predict until you put it together and try, even when you test it part by part in simulation. It's more than about bandwidth and amplification mind you. The NF has to be correct, you have to avoid coupling, and so on. Though it has made me a better designer by a huge margin in the long run. But that's experience preventing the errors instead of simulation and testing.
I got several of those LED strips and hang them just above my work area from all directions. Pretty much eliminates all shadows, no noise (the power supply is a few metres away), and the only problem I can think of is that it isn't truly white light.
Yes and no. In the case of digital logic you can write a model in VHDL, Verilog or whatever floats your boat. And then do functional verification on that. Dump it in a FPGA or hope your computer is fast enough. But for analog circuits it's not as easy. SPICE can give you a pretty good idea, but it's not perfect. Even more advanced programs like ADS won't give you the full picture. As the final step you will always need to actually make a prototype and just hope it works.
So yes you can test small parts of the design, but a large design can't be simulated without significant computing resources.
Well, obviously you might want to avoid metal. You can get these great plates for lab table surfaces made from some sort of ceramic. It's heat resistant and pretty tough, which is really necessary if a SMPS decides to hit the self destruct button. For soldering, just get a wooden board to protect the surface from direct impact with a soldering iron.
Miniature drawer cabinets are important and actually rather expensive, especially those that can be stacked.
Good soldering irons (more than one!) are a must obviously. Get both an analog and digital scope with at least 2 channels each. More is better. Personally I like putting a computer near my electronics workbench to view schematics, considered investing in a large TV for that but I'm a bit short on cash for that.
You want several supplies, current limited and not. Isolation transformers, a good variac, signal/function generators.
Good to have as well are an impedance meter/Q meter, network analyser, spectrum analyser and logic analyser. Especially the latter is worth considering, you can get pretty cheap versions these days that you hook up to a computer. For the other devices I advice stand alone versions cause it's really a lot easier while measuring if you can play with the knobs to home in on what you actually need. If you have more than enough money also get one of those microcontroller programmers with several sockets, that thing has saved my life more often than not.
Anyway, good luck!
It seems people miss the point of calculus quite badly. Calculus allows you to prove or simplify quite a lot of trigonometry. Numerical calculation methods also depend heavily on it. Your entire computer is the work of calculus being applied to physics. So do pay some attention to it and you might learn a useful thing or two. On top of that your game engines also need realistic physics. Good luck doing that well without a three dimensional field of vectors.
But where advanced mathematics really show their real value in my personal opinion is in electronics. Analog electronics, especially high frequency and RF systems can't be described or even approached without it. All the "rules of thumb" find their backing in advanced math. Even something as simple as the Fourier transform requires calculus to explain correctly. And lets not forget about our friend Maxwell...
In short, in the current society where electronics play a major role in almost any device it might be a good idea to be able to approach the subject. But you need calculus for that. A nice bonus is that calculus has helped me the past few years in reducing the amount of taxes I pay. Cause calculating an optimal solution to a problem is also something calculus is an excellent tool for.
I won't even bother responding to the first remark considering how Slashdot has become. You know very well what I mean.
... Obviously temperature is the most important factor, and even that one is often left out of the equation when they do these sort of calculations. And GaAs does come into play when you want efficiency. All the multi-wavelength/junction designs are based on gallium-arsenide or germanium substrata as far as I'm aware of. You do have some variation in dopant materials but I fear I'd be going a bit off-topic with this.
...).
And you can claim what you want, I have had to work on designing photovoltaic devices myself on several occasions, hated every minute of it; I got so fed up with them that I quit my job.
So first of all, where did I say separately? It's simply a well known fact that semiconductor properties depend on environmental factors and those do include temperature, air pressure, humidity,
I'll grant you that the first link does have some points. But I do not agree with their calculations.
This being the article they seem to get most of their data from: http://www.bnl.gov/pv/files/pdf/abs_193.pdf
The actual cost of manufacturing the substrata are very hard to define. Due to the increasing demand the solar panel industry has in fact started producing its own substrates as well. So the actual energy usage per square metre increased. The 14% efficiency as it should be around 10% for the panels they mention. The insolation values used are questionable. And I have yet to see an inverter last for more than 5 years without any sort of defect. Both the inverters I have owned always failed after roughly 4 years of service. The one I built myself (bit over designed mind you) is now up and running for 6 years and 2 months. Additionally their assumption about the main source of power is questionable as well as they modelled it based on the European power grid. Most solar panels are actually made in South-East Asia where dirty coal plants without air filters are still in common use. And the efficiency of the manufacturing process is also significantly lower due to older machines. This study on the other hand assumes machines of which some are still cutting edge at this point. Combine all these factors and I don't see it getting anywhere close to the 4 years they wish to claim. Last estimate I've seen that was trustworthy in my opinion was around 7 years on average assuming half of the panels were put at the equator and the other half put at the latitude of Berlin. It might have dropped a bit, but that study didn't really include the costs of the support equipment (mounting frames, inverters, modifications to the power grid for load balancing,
The second link mainly seems to rehash what the first link said judging from the references at the bottom so not going to bother.
And we're talking about replacing the main power supply for the power grid. So yes, you will need a large amount of engineering. I doubt anybody has actually ever published a reliable article discussing the full environmental costs of dedicated solar plants.