Epic Failure 1 Year in the Making

My expression is completely flat right now. I simply cannot believe I’m about to say what I’m preparing to say. I spent nearly a year cracking large prime numbers. In short, I took-on a project I called The Flowering N’th Prime Project, where I used my SheevaPlug to generate a list of every [every millionth] prime number. The current “golden standard” is this page where one can look-up the N’th prime up to 1 trillion. My goal was to reach over 1 trillion, which I did just this morning! I was planning on being the only source on the web to allow lookups of prime numbers greater than 1 trillion. flowering_primes

However, when I went to look at the logs, I realized that the software had a small, fatal bug in it. Apparently every time the program restarted (which happened a few times over the months), although it resumed at its most recent prime number, it erased the previous entries. As a result, I have no logs below N=95 billion. In other words, although I reached my target this morning, it’s completely irrelevant since I don’t have all the previous data to prove it. I’m completely beside myself, and have no idea what I’m going to do. I can start from the beginning again, but that would take another YEAR. [sigh]

So here’s the screw-up. Apparently I coded everything correctly on paper, but due to my lack of experience I overlooked the potential for multiple appends to occur simultaneously. I can only assume that’s what screwed it up, but I cannot be confident. Honestly, I still don’t know specifically what the problem is. All in all, it looks good to me. Here is the relevant Python code.

def add2log(c,v):
	f=open(logfile,'a')
	f.write("%d,%dn"%(c,v))
	f.close()

def resumeFromLog():
	f=open('log.txt')
	raw=f.readlines()[-1]
	f.close()
	return eval("["+raw+"]")

For what it’s worth, this is what remains of the log file:

953238,28546251136703
953239,28546282140203
953240,28546313129849
...
1000772,30020181524029
1000773,30020212566353
1000774,30020243594723

     

Mental Viscosity

A few weeks into dental school I feel I’m fairing decently. I have reached a point where I know everything will be okay, but am still disappointed at the [immense] amount of time it requires. There are so many things I wish I could do, but all of my projects need to be placed on a 4-year hiatus. I can bask in the satisfaction of the few projects I completed this summer, and I only hope that it’s enough to last me for four years. Dentistry, while important, is nothing more than emulation/repetition of what everybody else does. I simply have to satisfy my creative and ingenuitive desires in my hobbies, whatever they may be. For now, this website will cease to grow. Perhaps when I become more in control of my studies I will contribute to it, but in all likelihood I won’t be able to do anything worth writing about until 4 years from now [sigh]. With that being said, adieu, and goodnight.

I realized I never posted video of my finished prime number generator, so here it is. Full details are described on the project page. In brief, the 2-digit display on the left is the last two digits (in base-10, decimal) of a number currently being tested for primeness. This number is also displayed on the bottom red bar above the yellow lights (in base-2, binary). Once proven to be prime (by attempting to divide it by every number between 2 and its square root, every 1000th attempted number shown in yellow lights in binary), it’s loaded onto the top row of red lights (binary) and on the character LCD. N represents the Nth prime, with V representing its value. Half way through the video, the display says that the 16,595,044’th prime (N) equals 306,692,621 (V). Don’t believe it? Check my work.


     

Microcontroller-Powered Prime Calculator is [Mostly] Complete!

My microcontroller-powered prime number generator/calculator is virtually complete! Although I’m planning on improving the software (better menus, the addition of sound, and implementation of a more efficient algorithm) and hardware (a better enclosure would be nice, battery/DC wall power, and a few LEDs on the bottom row are incorrectly wired), this device is currently functional therefore theoretically complete (I met my goal). This entry will serve as the primary reference page for the project, so I will provide a brief description of what it is and what it does. First, here’s a picture of the device in its current state (click to enlarge):
primepic1

BRIEF DESCRIPTION: This device generates large prime numbers (v) while keeping track of how many prime numbers have been identified (N). The 5’th prime number is 11. Therefore, at one time this device displayed N=5 and V=11. N/V values are displayed on the 20×2 LCD. In the photo, the 16,521,486th prime is 305,257,039 (see for yourself!). The LCD had some history. In December, 2003 (6 years ago) I worked with this SAME display, and I even located the blog entry on November 25’th, 2003 where I mentioned I was thinking of buying the LCD (it was $19 at the time). Funny stuff. Okay, fast forward to today. Primes (Ns and Vs) are displayed on the LCD, but what’s with all those other LED lights? I’ll tell you:
primepic2

In short, each row of LEDs displays a number. Each row of 30 LEDs allows me to represent numbers up to 2^31-1 (2,147,483,647, about 2.15 billion) in the binary numeral system. Since there’s no known algorithm to generate prime numbers (especially the Nth prime), the only way to generate large Nth primes is to start small (2) and work up (to 2 billion) testing every number along the way for primeness. The number being tested is displayed on the middle row (Ntest). The last two digits of Ntest are shown on the top left. To test a number (Ntest) for primeness, it is divided by every number from 2 to the square root of Ntest. If any divisor divides evenly (with a remainder of zero) it’s assumed not to be prime, and Ntest is incremented. If it can’t be evenly divided by any number, it’s assumed to be prime and loaded into the top row. In the photo (with the last prime found over 305 million) the device is generating new primes every ~10 seconds. Not bad! Let’s discuss technical details.
primepic3

I’d like to emphasize that this device is not so much technologically innovative as it is creative. I made it because no one’s ever made one. It’s not realistic, practical, or particularly useful. It’s just unique. The brain behind it is an ATMEL ATMega8 AVR microcontroller (What is a microcontroller?), the big 28-pin microchip near the center of the board. (Note: I usually work with ATTiny2313 chips, but for this project I went with the ATMega8 in case I wanted to do analog-to-digital conversions. The fact that the ATMega8 is the heart of the Arduino is coincidental, as I’m not a fan of Arduino for purposes I won’t go into here).

I’d like to thank my grandmother’s brother and his wife (my great uncle and aunt I guess) for getting me interested in microcontrollers almost 10 years ago when they gave me BASIC Stamp kit (similar to this one) for Christmas. I didn’t fully understand it or grasp its significance at the time, but every few years I broke it out and started working with it, until a few months ago when my working knowledge of circuitry let me plunge way into it. I quickly outgrew it and ventured into directly programming cheaper microcontrollers which were nearly disposable (at $2 a pop, compared to $70 for a BASIC stamp), but that stamp kit was instrumental in my transition from computer programming to microchip programming.

The microcontroller is currently running at 1 MHz, but can be clocked to run faster. The PC I’m writing this entry on is about 2,100 MHz (2.1 GHz) to put it in perspective. This microchip is on par with computers of the 70s that filled up entire rooms. I program it with the C language (a language designed in the 70s with those room-sized computers in mind, perfectly suited for these microchips) and load software onto it through the labeled wires two pictures up. The microcontroller uses my software to bit-bang data through a slew of daisy-chained shift registers (74hc595s, most of the 16-pin microchips), allowing me to control over 100 pin states (on/off) using only 3 pins of the microcontroller. There are also 2 4511-type CMOS chips which convert data from 4 pins (a binary number) into the appropriate signals to illuminate a 7-segment display. Add in a couple switches, buttons, and a speaker, and you’re ready to go!

I’ll post more pictures, videos, and the code behind this device when it’s a little more polished. For now it’s technically complete and functional, and I’m very pleased. I worked on it a little bit every day after work. From its conception on May 27th to completion July 5th (under a month and a half) I learned a heck of a lot, challenged my fine motor skills to complete an impressive and confusing soldering job, and had a lot of fun in the process.

By the way, here’s a simplified schematic:
text5130


     

Summer's End is Nearing

My most glorious summer yet is reaching its end. With about a month and a half before I begin dental school, I pause to reflect on what I’ve done, and what I still plan to do. Unlike previous summers where my time was devoted to academic/thesis requirements, this summer hosted a 9am-5pm job with time to do whatever I want to after. I’ve made great progress in the realm of microcontroller programming, and am nearing the completion of my prime number calculator. I’m very happy with its progress. I think it’s time for some photos.
scott_working

Here I can be seen working on my prime number calculator. The primary display is nearing completion, and now it’s time to start wiring the buttons, switches, speaker, etc. Note the vintage scope in the background. In the photo it’s showing 60Hz (I couldn’t think of anything more profound to display?) which I’ll say is a representation of the fact that your body is continuously bombarded by electromagnetic radiation whenever you set foot in a house.
wiremess

This is the current state of the back panel of the prime number calculator. It’s becoming quite complicated.
lightson

As you can see, most of the LEDs are working but I’m still missing a few 74hc595 shift registers. It’s not that they’re missing, so much as I broke them. (D’oh!) I have to wait for a dozen more to come in the mail so I can continue this project. Shift registers are also responsible for powering the binary-to-7-segment chips on the upper left, whose sockets are currently empty. Since this project is on pause, I began work hacking a VFD I heard about at Skycraft. It’s a 20×2 character display (forgot to photograph the front) and if I can make it light up, it will be gorgeous.
vfd

Here’s a high resolution photo of the back panel of the VFD. I believe it used to belong to an old cash register, and it has some digital interfacing circuitry between the driver chips (the big OKI ones) and the 9-pin input connector. I think my best bet for being able to control this guy as much as I want is to attack those driver chips, with help from the Oki C1162A datasheet. It looks fairly straightforward. As long as I don’t screw up my surface-mount soldering, and assuming that I come up with 65 volts to power the thing (!) I think it’s a doable project.

Update: I found a funny photo from field day. After the tents, antennas, and radios were mostly set up, everyone was exhausted. I was ready to make some contacts! I fired-up my ‘ol netbook and tried communicating over 40m using psk (a digital mode), a mode I’ve never used, with software I’ve never used, on a band I’ve never used. It wasn’t working either. I spent the first several hours in frustration because what I was trying to do wasn’t working, and I couldn’t figure out why. This photo was taken at the height of my frustration =o)
me


     

Prime Schematics

Here’s a rough approximation of the current schematic of the prime number calculator I’m working on. Last night I finished wiring all 12 shift registers for the primary display, so now it’s time to start working on software. Notice that we have a lot of pins free still. This will be advantageous if I decide to go crazy adding extraneous functionality, such as fancy displays (LCD?, 7-segment LEDs?, VFD?, all 3?!) or creative input systems (how about a numerical keypad?). After feeling the stink of paying almost $15 for 100′ of yellow, 24 gauge, solid-core wire from DigiKey I was relieved (and a little embarrassed) to find I could score 1,000′ of yellow, 24 gauge, threaded wire for $10 at Skycraft! Anyway, here’s the current schematic:
text5130