OpenCores - free hardware

OpenCores - free hardware

Taking the Open Source ideology into literature and other creative areas is an obvious step, but there's yet another interesting question to consider. If software can be produced and distributed using the open model, then how about the hardware? If I can use Linux and other Open Source programs on my computer, when will the computer itself be as open as the programs running on it? In a sense, personal computers are an open architecture in the sense that you can build your own computer by using components that meet certain open standards. But today, the components and their production processes are in no way open.

Offhand, you'd think there's no way the open model could work for computer components - surely, what works for software can't work for hardware. The source code of programs, the text of a book, films and audio files all have one thing in common: information stored in digital form can be copied an unlimited number of times, for free. But computer components are tangible parts of the physical world, and here again we hit the traditional laws of supply and demand. Our empirical knowledge of the world says that one computer is one computer and as such can't simply be distributed by open methods.

But in fact, despite our empirical experience, there are projects working towards the manufacture of Open Source computer components. The most well-known of these, OpenCores.org, actually seems to be doing well and is even mass-producing some products for the computer market!1

So what is this all about?

I don't understand about electronics any more than most people, but to go any further we'll need some knowledge about how integrated circuit cards are made.

Some components of a computer are the micro processor, various RAM memories, a modem or network card for connecting to the Internet, a card to connect the screen and another to bring sound to the speakers. All these various components are attached to the motherboard, which in itself is a brilliant piece of electronics containing several microchips. Consumers assume these bits and pieces all come from a factory somewhere, packed into plastic and cardboard, and are eventually fitted together. To that is added a keyboard, a mouse, speakers, monitor and - Hey, presto! - we have a computer.

But in reality the manufacture of these microchips is a lot like programming - but instead it's called design. Actually, separate programming languages are used for designing microchips. The most popular languages are called VHDL and Verilog.2 These descriptions of integrated circuit cards written in VHDL or Verilog are the really hardcore stuff. To be able to read them, you need to understand both electronics and programming, and all they tend to give a regular Joe like me is a headache. But those who understand this language are the people who make the microchips you and I use every day. And I'm not just talking about computers, because microchips are everywhere: in fridges and cars, in wood pulp factories and elevators.

Having the VHDL language files on a computer allows one to use a more or less automated process to manufacture a so-called gate-level description, then to use these logical gates to create the physical model, or layout, of the integrated circuit. The final stage is to make the mask (a sort of mould) which is used to radiate (or cast) small transistors and wires onto a silicon disk.

In this way the manufacturing line at the microchip factory is relatively automatic for making microchips from semi-conductor materials according to a blueprint. The chips are then packed and sent off to the computer manufacturer, the car plant, fridge factory, or wherever. Naturally, there's an art to building and maintaining a manufacturing line like that, and one essential is that these factories must be cleaner than an surgical operating theatre and be entirely dust free. There's a lot of other magic to the process, but we won't go into all that here. One time-consuming part of the work of designing an integrated circuit is the so-called timing. In this world of physics, one must understand that the electric signal in the circuit only moves with the speed of light - no faster. That's why a logical and accurately designed VHDL file won't generate a microchip that works as it should unless the electrical impulse actually reaches from transistor A to transistor B fast enough but not too fast.

For this general explanation, that is as far as we need to delve into that. We just have to believe that what it comes down to is that the data in a VHDL file is used to make a mould which is then used to make semi-conductor materials become a working microchip.

That means the major single effort in making a microchip goes into designing it. That takes a programmer who can write code in the VHDL language. Then testers are needed to check that the chip works as designed. Naturally, the testing too is done before a single chip is made, by creating a model of the VHDL file inside a computer program to simulate how the chip works. Testing is an extremely important part of the manufacturing of microchips and plays a far larger part in the design work than, for instance, in the writing of software, because it's a lot harder to fix a microchip after it's been made. So, microchips have to be virtually flawless before a manufacturing line is set up to produce them in the thousands.

This simplified explanation doesn't really hold true for all microchips. The fast processors like Pentium, Opteron or Athlon, which are made for new computers almost approach the occult, defying as they seem to do the laws of physics. But for those processors, too, design is the largest part of the job, after which making the fancy microchips is no different from making simpler ones. A great deal of fine-tuning of the production line and other hard work goes into squeezing the last ounce of effect out of these high-speed processors. Again, without going into detail, the challenge is to get as many transistors and wires to fit onto the semi-conductor as possible. And once the production line can be made to manufacture small enough transistors, the next task is to get the electrons that transmit the electrical signal to stick to the right wires. If the wires are too small and too close together, the electrical signals will hop from one wire to the next, everything gets confused, and the processor is useless.

That's the case for the newest and fanciest microprocessors. In contrast, as long as you're not out to set a speed record the manufacture of other processors is - so I'm told - very nearly as easy as I've described above. You take the finished VHDL file to a subcontractor who manufactures microchips, and moments later finished processors start popping out the other end.

OpenCores received a lot of publicity in December 2003, when a genuine integrated circuit was made from its Open Source design.3 Several manufacturers have now made the same chip, but the first, historical, Open Source chip was made by Flextronics Semiconductor, or the same company that makes chips for Nokia, of cell phone fame, and Cisco, which makes most of the routers for the Internet. This certainly proved that Open Source can work even in the components part of the computer business - in hardware as well as in software.

It's not as if the OpenCores chips would not have been realized much earlier as so-called FPGA chips.4 Widely used in electronics because they are flexible, FPGA chips are also popular with OpenCores and others who design chips as a hobby or for teaching purposes. It's a sort of general-purpose microchip. It comes empty from the factory and users can "load' it with whatever features they like, which they've first written in VHDL, for instance. The P in the acronym FPGA means "programmable'. In making prototypes the FPGA chips are invaluable, because they are usually reprogrammable. Which means, once a user has tested their chip they can tweak their own VHDL code and reuse the same chip to try the new version.

To "load' - or program, as the professionals say - an FPGA chip, a separate device is connected to a computer for the empty FPGA circuit to be plugged into. However, some FPGA circuits forget all they had contained every time the computer is turned off, and the chips have to reload each time it's turned on again. Actually, FPGA circuits could be compared to CD burners, which are now an integral part of all new computers. We all used to think CDs could only be made at a CD factory, but then somebody invented the CD burner and now anybody can burn a CD with their home computer. Naturally, a home-burned CD won't be of equal quality or as durable as those made in factories, so there's no point in making multiple copies of CDs at home. But thanks to CD burners any music lover can release their own CD if they want. Similarly, your average electrical engineer can make their own processors using FPGA chips.

And although you won't get the effect of a Pentium or Opteron using these general-purpose chips, a modern FPGA chip can already be used to create a microprocessor corresponding to the ancient 486dx. In plain language that means a computer as powerful as those used in the early nineties could be built by a handy electrical engineer entirely from Open Source components! The equipment necessary to program an FPGA chip isn't even expensive. For those interested, various development kits are available from a number of companies, such as Altera and Xilinx, and prices range from around €80 to several thousand. Of course, a really handy electrical engineer could also build the circuit board that is used to program the chip, in which case the whole process would cost next to nothing.

hingo Sat, 2006-09-02 08:12
  • 17490 views

Comments