This page contains old entries from my blog.
I'm building a server for volunteer computing. I also used some 3D printed parts. (Yes, I put the heatsinks on backwards in this photo).
More random 3D prints...
I found something interesting. I've had two vials on my shelf for about eight months: both are solutions of distilled water, hydrochloric acid, and stannous chloride. One of them has a small amount of gold in it. The one without the gold now has a precipitate. I'm not sure if it's just stannous chloride or something else. The one with the gold, however, did not form a precipitate. Curious...
Since replacing the extruder motor on my 3D printer, I've noticed the new motor gets pretty hot. I decided to replace the second extruder, which I never use, with a cooling system. I stuck a heatsink to the side of the motor and designed a cowl to put a fan in the rear. The reduced weight of the carriage also reduced wobble. My print quality is better than ever and it's a bit more reliable.
I found a shelf by the dumpsters. It was just a pale, unfinished wood. I decided to sand it and stain it. I think it looks pretty good. I recently bought more than 50 new books, so it came at just the right time.
I bought a 3D printer and I've been printing constantly. It's a CTC Replicator Dual: a cheap knock off of a knock off. It was only about $350.
When I was shopping around, I didn't know if I wanted to buy a kit and build the printer myself. If I did, I would be forced to learn a lot about 3D printing, and that's what I wanted, but it would be a lot of trial and error before I could get around to printing some stuff. On the other hand, if I bought a pre-assembled printer, it would work out of the box and I could get to printing right away, but then I would just have a "product" and I wouldn't gain the knowledge and skill that comes with building it yourself. It turns out that the CTC is the perfect compromise: it comes pre-assembled, but you can't expect to get good prints out of it without tinkering for a few days, and you can't expect it to work for very long without upgrading it. The reviews I read all said once you get it working it can make some quality prints comparable to a $1500 printer, however it takes a lot of work.
I've been using it for about a month. So far, my experience has been positive. There has been some hair pulling (well, beard pulling; I don't have hair), but the successful prints came out beautifully. I discovered that some of the first things you should print are upgrades to the machine! That's right, start printing parts for the printer so that when it breaks down, you're ready to replace the parts. Upgrades can also improve the quality and reliability of the prints.
I recommend printing these parts in this order:
- Extruder drive block (spring loaded)
- Y-axis bearing brackets
- Filament guide for top of extruder
- Filament spool holder that mounts to top of chassis
- Filament filter
- Replacements for wooden brackets (for guide rails)
- Replacements for limit switch brackets
- Replacement Y-axis motor bracket
Anyway, I will go into more detail about my printing adventures in another section.
It's been a long time since I've updated. I've been working hard and I never really feel caught up. Everything I do in my free time is work as well. It's been weeks since I've relaxed, and many years since I've felt okay with relaxing, even for a moment. If I'm not working feverishly on a project, I feel like I'm wasting time.
I finished this little "bits-and-pieces" jar rack earlier today. The funny part is that after I took this photo, I actually found something I was looking for by looking at this photo! Ha! That tiny little heatsink in the second jar is perfect for a Raspberry Pi!
This is a pretty old picture, but it shows an experiment I did creating a Beofwulf Cluster in a virtual environment. It used OpenMPI running on CentOS. For those who are not familiar, a Beowulf cluster (or just a cluster in general) is a way of distributing computational work over multiple systems to allow massively parallel computing (when scaled up). Some of the biggest supercomputers in the world use OpenMPI and it was the first Message Passing Interface (MPI) to break the petaFLOP/sec barrier (one quadrillion float operations per second). The image below shows the CPU usage on four different nodes as I ran a parallelized program. It was simply squaring a number if I remember correctly. Of course, because the cluster was virtualized, there was no performance gain. In fact, there was probably a performance loss. And the program was redundant across nodes (all nodes were doing the same task). This was just a test, however. I later did this on a rack of decommissioned servers and ran instances for the Folding@Home distributed computing project.
One day as I was doing dishes I discovered something strange on one of my cups, some kind of growth. I thought it was a fungus at first, but then I realized it might be crystals of sodium bicarbonate. I had previously used the cup to make a saturated solution of baking soda and water. The inside of the cup had an intricate crystalline crust on the bottom, but how did some of it get on the outside? IT EXTRUDED THROUGH THE CUP!!
The cup itself had tiny hair-line fractures in it. They're almost invisible to the naked eye, or at least difficult to see if you're not looking very closely. As the crystals grew, they pushed their way through the side of the cup.
At least that was my theory. I still didn't know if the crystals on the outside of the cup were sodium bicarbonate. A litmus test confirmed the alkalinity.
I looked at the crystals under a microscope, but they were a bit underwhelming. My microscope is pretty low quality.
They were much more exciting to look at under a jeweler's glass. These are the crystals from inside the cup.
I had a funny thought: maybe if I submerged the cup in vinegar, the sodium bicarbonate inside the tiny cracks in the cup would react and produce gas, thus expanding the cracks and weakening the structure of the cup. I put the cup in vinegar and let it sit for a couple hours. I then struck the side of the cup with a hammer, but the cup didn't seem to be any weaker. I compared it to an identical cup from the cupboard. Oh well. A non-result is still a result.
I recently did an experiment for Recovering and Casting Indium (click link to see more).
I have come to the conclusion that you should ALWAYS socket your ICs! If at all possible; if you're not using surface-mount parts and don't have space constraints; solder a socket down and pop that chip into the socket. I was working on an audio project, and I must have overheated the op amp while soldering. I damaged it just enough to make the audio quality really terrible. De-soldering it was such a pain that, in my frustration, I tore off one of the copper through-hole pads from the PCB. I rebuilt the whole project and socketed all the ICs; now it works like a dream.
This brings me to another pro tip: always buy at least twice the parts; that way, if you screw up the first one you can build another. It may appear to be twice the cost (well, it is), but when you buy parts in larger quantities, they are cheaper per piece and you'll eventually need those parts for something else anyway. So when you start another project, and you happen to have all the parts you need, you realize the value of having parts on-hand, especially when you save yourself a trip to RadioShack (or waiting for parts to arrive in the mail).
What I've done is go on eBay or Amazon and search terms like these:
- resistor assorted pack
- capacitor electrolytic assorted pack
- capacitor ceramic assorted pack
- diodes assorted pack
- 555 timer 50pcs
- transistor 600pcs 15value
- 24awg solid hookup wire
- male and female 2.54mm headers
- prototype board 20pcs
Obviously, there are more parts you might need, but you get the idea. Buying assorted packs removes a lot of hassle and they are cheap.
Serious about data storage? How about a hard drive the size of a rotary engine! This is an IBM 10SR that I saw at Free Geek in Portland.
I setup an encrypted IPSec tunnel between two Raspberry Pi Model B computers. They are running a router OS called IPFire (http://www.ipfire.org/). I tested throughput by watching Blade Runner across the encrypted channel.
This was actually an experimental setup for a client. The question was: Is the Raspberry Pi a viable low-cost VPN appliance for doing nightly off-site backups? (Small business). The answer was "yes, until it dies." It lasted about 6 months. To be fair, this was one of the original models. The new Pi 2 or 3 could probably do a bit better.
A closeup of an Intel processor. The object for comparison is the tip of a sewing needle.
Notice that there are tiny holes on those gold pads in the middle. They're very difficult to see with the naked eye. I didn't know they were there until I took this picture.
My uncle gave me some broken uninterruptable power supplies (UPSs). Three of them were completely useless; the batteries were bulging and utterly ruined. I probably could have collected the lead and sulfuric acid from them, but damn are they hard to open! And they're messy.
One of the PSUs (the one on the far left) had good batteries, but the control unit was dead. So I now have some nice 12V batteries. I kept part of the case to use as a rack to keep them in. Nice!
A friend of mine gave me a TV he found next to a dumpster. It is a Sylvania LC370SS8 1216x676 LCD TV.
It wasn't working properly. It would turn on and then very quickly turn off.
It turned out that the wires for the fluorescent back lights had burned through their insulation and were shorting out. The TV would shut off because the ground fault interrupter (GFI) would trigger to prevent it becoming a fire hazard. Sylvania cheaped out on the wiring, but they can't legally cheap out on the GFI. Hurray for industry standards!
I remedied the problem by cutting the outer jacket from some CAT5 cable and using it to insulate all the wires. The TV then turned on without any problems. It's a cheap TV, but the resolution was higher than my other TV, so this is now my main screen for my living room PC.
Wiping 9 hard drives simultaneously. I found a ton of discarded computers by a dumpster. Most of the other hardware wasn't worth keeping. When you scavenge any kind of storage media, be sure you wipe it clean!!
I'm using Linux and the 'dd' utility.
for x in b c d;do dd if=/dev/urandom of=/dev/sd$x bs=4M conv=notrunc,noerror & done
...where the devices are sdb, sdc, and sdd.
The device /dev/urandom can be used to write the disk with random bits. This is slower than /dev/zero which will write the disk with zeros. For larger disks, like a 3TB drive, the time difference could be huge.
Those familiar with Peter Gutmann might think that one pass isn't enough; it should be 35 passes! Guttman's paper "Secure Deletion of Data from Magnetic and Solid-State Memory" was written in 1996 . His method was designed to take into account that old, low "write-fidelity" hard drives with ancient encoding schemes were still in use in 1996. Basically, if you have a modern hard drive, one pass of random bits is sufficient.
There was a challenge once posed by a group called 16systems (I think) offering a cash reward if anyone in the world could recover data from a hard drive that was wiped with one pass of zeros. No one was able to claim the prize.
I have quite a bit of sheet metal (the kind used for ducting I think). I discovered that treating it with hydrochloric acid (31.45%) gives it a nice finish and makes it a much better surface for paint and adhesives. I recently tested it with JB Weld, and the adhesion is significantly stronger. This is going to be useful.
I recently confirmed that the metal was hot-dip galvanized mild steel. The outer layer is zinc, and when it reacts with the HCl it produces zinc chloride and hydrogen.