The use of a DLP projector as image source elegantly and cheaply solves the problem of multilayer alignment.

Chosen excerpts from the paper:
“This overlay technique is easy to implement by placing the overlay in the master-slide mode of presentation software such as PowerPoint.”

“Creation of exposure slides using presentation software is easy. A typical slide show consists of a sequence of slides, with the design projected first in red for alignment and focusing, followed by a blue exposure slide for a controlled time, followed by a new red image to align the next pattern.”

“Students have used this system for a variety of research projects and upper-division laboratory exercises. These have included making optical diffraction patterns, catalyst pads for the growth of carbon nanotubes, and a variety of micron-scale symbols and signs. We have our used our thin-film evaporator in combination with this process to make metal patterns using the liftoff technique, where metal is deposited on the patterns made in the photoresist, and the remaining photoresist is then dissolved, leaving metal patterns stenciled on the substrates. Such films allow millimeter-scale electrical contact pads to be made for micron-scale objects such as thin-film resistors and long carbon nanotubes. Metal patterns on glass substrates should be suitable for creating custom two-dimensional binary diffractive optical elements similar to those used with inexpensive laser pointers.”

Aurora compressors (photo from the original website)

The setup is impressive, with 4 refrigeration stages equipped with compressors (like those in freezers, but bigger), hand-built heat exchangers (made from annealed copper tubes) and a huge list of various parts and bits (I wonder what the total project cost was, when home improvement retailers sell you a single valve for a dozen euros…)

I am not a big fan of overclocking, but I am excited by the other possibilities that such a project would bring. Do you want to cool down that thermal infrared camera? Ever wanted to play with superconductors? They are not so expensive, if you can cool them… Do you need a high vacuum, for, say, making your own triodes (french), building a cyclotron, making your own Nixie tubes (okay, you do not need that much of a vacuum for that, but why have two pumps) or VFDs, or for doing evaporative vapor deposition (DIY semiconductors, flexible PCBs, DIY OLEDs)? A sorption pump, commonly used as roughing pump in industrial setups and based on a molecular sieve costing a few dozen euros only, might provide you with what you want, but again requiring very low temperatures.

Way to go!

It gives some insight about how a modern processor is designed (entirely in SystemVerilog), how it could be ported to FPGA (including a lot of practical aspects), and about the debugging and testing phases (on real PC hardware).

“Once the netlist is produced by DC-FPGA, Xilinx ISE 10.1.03 is used to produce the final bitfile. However, it turns out that DC-FPGA in rare occasions can produce buggy netlist silently, even for a correctly developed RTL circuit that passes the RTL simulation”

“Though these behavioral RTL models can be naively passed to the logic synthesis tools which may sometimes infer appropriate FPGA-specific memory structures, for most structures we observed an explosion in LUT utilization.”

URL: http://portal.acm.org/citation.cfm?id=1508160

http://lekernel.net/links/

http://www.validlab.com/goldberg/paper.pdf