Fixing Battery Corrosion in a Mini Maglight

About twenty years ago or so, I bought a Mini Maglight to supplement one I had bought about twenty-seven years ago. Both of these use the itty bitty incandescent bulbs, neither one being anywhere near new enough for the fancy modern white LEDs. They work, anyway. Or, they did – until the batteries in the newer one leaked.

Normally, this wouldn’t be a big deal. You replace the batteries and get on with your life.  In this case, though, there was a complication. The leaking battery was completely stuck inside the flashlight barrel. I tried banging the open end against the back of legal pad, and I tried that a lot. No good.  Evidently it was time to take it apart and push out the dead cell.

There are at least two versions of the incandescent Mini Maglight running around, and my two were of different generations. I had taken the old one apart before, but the new one was a bit befuddling. First, field strip the easy parts – take out the bulb and then just pull on the “do not remove” outside bulb holder until it pops off.


This leaves you with the inside bulb holder, the barrel of the flashlight, and one immovable battery. It’s time to move the immovable. I didn’t have a pin punch the right size, so I grabbed a cheap, junky Torx screwdriver that would fit in the access holes at the very back of the inside bulb holder and whacked it with a hammer.


This resulted in squat. I was deforming and slowly hammering my way into the battery instead of pushing it out. Judo is too subtle, I’m switching to Kyokushin: hit it harder.  About ten good whacks with a small hammer and I started to actually move the battery.  A bunch more and it finally came out.  Finally, switch to a larger drift (in this case a, uh, Uniball Vision ballpoint pen…) and gently tap on it until the inside bulb holder falls out of the threaded end.


Inspection of the flashlight barrel revealed the reason the battery wouldn’t come out – when the cell leaked, it dumped potassium hydroxide into the bore.  This reacted with the aluminum barrel forming hydrogen gas (always a crowd pleaser) and potassium aluminum tetrahydroxide – aka “crud”.  Oh, and it also formed a corrosion pit in the aluminum – luckily not all the way through the metal.

Cleaning out the corrosion is done with a 20ga shotgun bore brush threaded into a cleaning rod and the tool spun with a drill.  Wear a dust mask and eye protection.  Bear in mind that the friction will heat up the aluminum, so pay attention and stop to let it cool at intervals.


Assembly is the reverse of dismantling.

Controlling a Dollar Store Solar Dancing Elephant

By now you’ve almost certainly seen those little dancing toys that move when their solar cell gets illuminated. The things fairly well litter the “everything in here is yours for just one stinking dollar” stores. I’m not going to disparage them as toys – to be honest with you, I think some of them are pretty entertaining.

Then again, when have I ever not dismantled one of my toys?

The goal here is to take a solar dancing elephant and hook it up to a computer. When something interesting happens in reality, the elephant starts wagging its ears up and down. I had thought about using it as an email alarm, but that’s just depressing. Instead, I think I’ll scrape the current weather off of the NWS and make it dance in response to something like temperature or rainfall rate. Elephants should get excited when I rains, I guess.

Step one is to crack the elephant open. What really bothers me is that some day, someone is going to arrive on this page after they googled exactly that phrase. Crack the Elephant Open. The mind boggles.


Anyway, with your elephant nicely bisected, note that the solar cell and some electronics are in the top half of the shell, and the lower half of the widget is all moving parts. The ears are pretty finely counterweighted.  It doesn’t take much force to get them moving. This is a good thing because that solar cell isn’t going to make a lot of power (milliwatts, maybe? ) and the electromagnet that it operates isn’t very big. Incidentally, the astute observer will note that there is a tiny, tiny bare die chip under epoxy on the circuit board, and that the coil is connected to it through something that looks like 40ga wire – aka “man! that’s some awfully fine wire there!” So do be careful.  If you break that wire, you’re pretty much out your one dollar.


Next step – clip the wires right at the solar cell (red one is positive, natch) and connect a long-ish two conductor wire to each of the leads going to the board. This is what you need to power your elephant. The cell on there makes about half a volt, but I’ve put five volts across the circuit with no ill effects. Evidently, the design has some wide tolerances. :-)

Final step – connect the wires to the ground (pin 18 was as convenient as any) and to a data pin (pin 2 is the least significant bit) on a connector for a parallel port. I added a parallel 100 ohm resistor that I really didn’t need, but I figured it might help the elephant live longer. Then finally all you have to do is find an old-school laptop with a parallel port and plug it in. A few lines of C, and voila!  You’re (announcer voice) controlling elephants from Linux.  And seriously, once you’ve done that, the rest of your day is going to be all downhill.


Controlling things comes down to manipulating the parallel port (and no, this will not work through a USB-to-Parallel converter).


void main(int argc, char **argv){
unsigned int value=0;
if (argc > 1) {


int port=/*  0x378 */ 0x3bc;
int res = ioperm(port,5,1);  /* allow access to io port plus next 5 addrs */

And that’s all there is to it.

Installing thick (15mm) 2.5″ drives in Dell drive carriers

I have a bunch of Dell 1955 blades that I’m reconfiguring. They used to be a fairly garden-variety supercomputer – 530 blades, Infiniband network, made it onto the Top500 list in 2006 (UNC’s “Topsail” cluster). Topsail reached the end of its usable life for that sort of workload (how progress does march on, no?) and was headed to surplus. The group I work in at Renci intercepted it and turned it into a Big Data cluster. The first-generation Nehalem cores aren’t so hot anymore, but having 1060 SATA channels for running Hadoop on… dude! You’re resurrecting a dell!

The only real hitch was storage. The blades came with 40 gig disk drives – enough to hold a boot image and not a whole lot more. The machine is long out of maintenance at this point, and the biggest drive Dell ever officially supported in there was probably a 300 gig part, so off to third-party land we go. I scored a couple of Hitachi 750 gig drives a year ago and they worked great. Now, Western Digital has started volume shipments of their 2TB 2.5″ drive. Naturally I ordered a pair of those.

Best laid plans, etc. When I went to install them in the Dell carriers (with the 40gig drives removed), there was a slight clearance problem The metal corner brace gets in the way of letting the drive fully seat, so the fourth mounting hole doesn’t line up with the drive:



…so it’s off to the Secret Underground Laboratory. Fortunately, the carriers are cast out of zinc, rather then being made out of steel, so they’re really quite easy to modify. I cut away some of the corner bracing of the first one with a Dremel tool, but that took about fifteen minutes. I have 530 of these to do, so I want something a lot faster.

Jamie would use a tablesaw and face shield, but I decided to go with something a slightly more subtle. I grabbed my total-piece-o-junk Jorgensen mitre box – never buy one of these, folks. The cast bed comes from the factory out of true, and if you have to pay a machine shop to resurface it, well, it’s cheaper to buy a good one.


Which is why I don’t mind using it to cut metal. I hate that tool, I want it dead so I can throw it away. You understand.


Once you have a nice notch all the way down to the side rail, just grab the rail with one pair of pliers and the excess tab with another, and twist. The zinc will fail right where it should. Clean up the burrs with a file, and you’re in business.

Then install the drive in the carrier, the carrier in the blade, the blade in the bladecenter, and let Rocks reinstall Linux on it over the network.


Now I just need to do this 527 more times…

Bose Headphone Repair

I have a pair of Bose Reality-Cancelling Headphones – not too effective against human voices, but they do a good job of cancelling steady noises like air conditioners and computer fans. I like them well enough, but I wouldn’t recommend them anymore.  Maybe the Panasonic version?

The problem is that after five years or so, the pads that go over the outside of my ear have started to come apart at a seam, exposing the foam. The solution is simple, of course: sew them back together. The hard part is the lack of room to work. I cut off a sewing needle to roughly 3/8″ and ground it back to a point. Surgical forceps work well to hold the seam back together temporarily, then a pair of long-nose pliers is what it takes to jam the short little needle through. I ending up using a second pair of forceps as an anvil and as as extra set of fingernails to grab the needle when it comes through. I didn’t even attempt any kind of running stitch or fancy lacing/trussing/macrame, I just tied off each stitch as I completed it and moved on to the next one.

_DSC0014 _DSC0010

Now, the failure is running around the inside of the cushion – when I repair one section, another one pops loose.  Pretty soon I will have sewn all the way around, and shortly after that I suspect the material will simply fail entirely.  At that point, I don’t know.  I’ve been known to limp stuff along just for fun.

Current score:  Patience vs Rampant Consumerism: 1-0.


The Knucklewarmer

You ever have this happen to you? There you are, working away in your secret underground laboratory, trying to be a good citizen of the earth. You have the heat turned down to an environmentally-responsible way chilly, and your hands are freezing. Happens to me every winter. It also happens to my colleague Dan Bedard, who thought “Gee, I should just make something to redirect all the warm air coming from my computer so it blows over the keyboard!”

Well, that’s a good idea, Dan. So good, in fact, I’m stealing it.

Behold the Knucklewarmer. Made from genuine earth-friendly recycled junk mail, it catches the warm air venting from the side of my rusty-but-trusty Thinkpad and directs it over the keyboard. 2.13 GHz of Pentium M goodness makes enough warmth to keep the knuckles working and to maintain feeling in my fingertips. Not bad for a laptop.


What about those newer laptops? What about even, God forbid, a netbook? No problem. Sure, an Atom N270 is something like a sub-5-watt part, but have no fear. The chipset uses another 9, and together they’ll keep ice crystals from forming. This baby works pretty well on a Dell Latitude 2100 netbook.

Construction is simplicity itself, as should be obvious from the pictures. It’s pretty much (1) locate some junk mail, (2) cut it to shape to form a little piece of ductwork, and (3) tape it up and slide it into position. More complicated versions are left as an exercise for the reader.


Downside? My next main machine was a MacBook Pro, and this didn’t work with the milled aluminum monoblock of desire. The heat comes out of that one from under the display hinge. No problem, though, because after that I switched back to a Thinkpad (both run Linux full time, by the way) and I’ll have heat to spare this winter.



I’ve made a few things and hacked a few things over the years, but my documentation was scattershot at best.  This blog should serve as a central place to record them, mostly for my own reference.  On the other hand, you’re reading it, so “Welcome”.  I hope something will be useful.


Now for a flurry of copying stuff over to here…