Friday, April 01, 2005

my 379k lab partner, who was almost brought to tears a few weeks ago, rallied and produced a working wireless communications link. what this means is that bits leave the computer, enter the air, and return via another antenna (only a few feet away) in the exact same arrangement. i held one of the antennas. the flowchart bryan created to accomplish this looks like the map of a small city. once the bits flew and returned none of his former agony mattered. all of the pain and doubt just vanished as he realized that he grasped something truly difficult. differentiating him as an electrical engineer, and giving him knowledge he might be able to apply for decades. such is the drug.

i've created my own map of a small city, but in a different class, VLSI design. my arithmetic logic unit (ALU) adds, subtracts, increments, or decrements two sixteen bit numbers in 1090 nanoseconds. the core of the ALU, called a kogge stone network, contains several layers of lookahead logic. if at one stage you need to carry a one, it wires this information ahead in parallel rather then slowly ripple the information one carry at a time. for example if your 2nd grade teacher gave you a difficult addition problem like 9997 + 8879 where the numbers were all big and generated carries, the kogge stone network would be like an extra set of hands, doing all of the carries nearly simultaneously rather than one at a time. this is something about 200 ECE graduates learn how to build every year. bryan with his flying bits is in the company of more like 4 or 5 people. i'm beginning to see people punch their tickets out. in fairness to myself, bryan is a graduate student. But I have a tremendous amount of work in front of me and a very short time to do it in. i don't think I'm going to grad school, and for the last few weeks I've been just hanging on.

prof. rappoport gave a brilliant lecture today on the subject of correlation filters and gaussian noise. he showed that with calculus and some basic knowledge of probability theory you can quantify how powerful the distortion is on a filtered channel. it gives noise a bandwidth of a half hertz he said, and indeed you could see on the whiteboard that the effect of a filter on the infinite chaos of random thermal noise is like taking a baby slice of cake out.

alex, my 345L lab partner, perfected his stepper motor program today. my contribution to the lab was to wire all of the hardware together. i used something called an L293 quadruple driver which was a professional kind of touch recommmended by the TA with only 20 minutes to go in lab last week. nobody bothered to try it except me. the thing about this course is that you have to fight for the TAs attention, and when you do get it there are always a few other students hovering around, wanting to take him away. in today's episode two such guys wanted to know how they could calculate the rotational rate of their motor. the lab instructions said that these things turned at 18 degrees per step. we ran our device and counted 20 steps. "count the time duration of 20 steps as the period" we told them. they didn't seem satisfied with this, perhaps because it was OUR device. i multiplied 20 * 18 on my calculator and showed them that the result was 360 degrees. for once i was issuing a smackdown.