Fridgehack, part I

I had planned to go to Micro Center on Friday afternoon seeking some hardware upgrades for the Chuxx0rb0x. The weekend was here, and I had to get something done besides sit around on Facebook while sipping fine Mountain Dew and doing problem sets. Since the magnets and bearings for Snuffles Reloaded hadn’t arrived yet, there wasn’t anything to do on that. Then I realized that the mutant minifridge plan won’t take that long and would actually be mildly useful, given that said minifridge in its stock condition wasn’t exactly very cold. I grabbed my calipers, some paper, and a pen, and zipped off to Micro Center, 20 minutes away by foot (there’s one hell of an incentive to get SR done).

One look at the colorful selection of CPU temperature maintenance products told me that no, I was not going to afford at least two of the biggest, flashiest, most obnoxious cooling rigs, which I had planned for the hack. Therefore, I had to spend an hour balancing practicality with obnoxious flair. In the end, practicality and utility won, like it does most of the time with me.

But, it only wins until I can get some of these suckers (blowers?) and upgrade everything else to go with it. It’s alot more expensive in-store. In addition, one of the qualifications is that the bottom contact plate must be at least 40x40mm in order to fit over the entire Peltier device on the refrigerator – or be easily wrangled to do so in the form of adaptor plates or machining. Many of the large heatsink/fan assemblies had very small contact plates, meant for the CPU only, which would require intensive modification to work with the Peltier device.

The looks I got from staff while huddled over big CPU coolers with a caliper measuring things and jotting notes were… well, I suppose priceless doesn’t work here, since most of them cost $50 or so.

“Sir, may I help you?”
“Thanks, but I’m collecting engineering data. I’ll put everything away when I’m through.”
“…….”

After browsing, I settled on a new plan. I wanted to see what could be done with minimal additional hardware.

This could have been as simple as getting higher-speed fans and modding the power supply, but I did want to replace one thing.

That’s the tiny aluminum structure on the inside which is the “coldsink” of the heat pump assembly. It attaches to the Peltier device, sandwiching it between the outside heatsink and itself through an aluminum block spacer of sorts. The two beige things are plastic-mounted nuts. This was just so disproportionately small compared to the outside heatsink that I though an upgrade was necessary.

I ended up leaving Micro Center with a 120mm fan, 80mm fan, and a conservatively-sized Pentium 4 cooler which happened to be all-copper with a flat copper slab as a base, perfect for this application.

Taking everything apart for chopping and screwing. The Peltier device is a common enough one such that I could get more on the surplus channels and “stack” them for more temperature differential later on.

I also had a look at the temperature adjust switch. The pot and thermistor board were firmly glued to the plastic covering, so I couldn’t look directly at the circuit. Staring at the PSU board long enough with a voltmeter let me guess that the power output was controlled by voltage through an op-amp on the board itself. The NTC thermistor and “temperature adjust” pot worked together on opposites of a voltage divider circuit. Lower output voltage (higher ambient temperature on the thermistor, or higher resistance on the pot) drove the PSU harder, increasing heat throughput. The voltage seemed to vary from 1.2 to 1.9 volts (outside, room temperature), and in normal operation with the pot set to “max cooling”, was about 1.4.

For the sake of overdriving, I switched out a 20K resistor attached to the pot to 12K. This wasn’t very scientific, and the value was selected based on guess-and-hope. However, experiments found that the output voltage at max setting was now around 1.0 volts (room temperature). This would be higher if the ambient temperature was colder.

While I was playing with the power supply, I took the opportunity to remove several power resistors that were planted right by the outputs, presumably as a form of current limiting. This means the PSU was even closer to meltdown than before, since it now had the opportunity to deliver to the components without the crippling effects of P=I²R on the power resistors. It didn’t concern me that much, since the only major power-sucking things on these are the fans – the Peltier device is directly attached to a 12 volt output.

The fans I got were some Antec things which, curiously enough, were not LED fans even though they were clear. They were, however, 3-speed, selectable by slide switch. These 3-speed non-LED fans were less expensive than the 3-speed LED fans by a fair margin, and had significantly higher throughput than the plain LED single-speed fans. I wasn’t about to pay Antec an extra $10 to wrap some LEDs around it, so I decided to do it myself. Economists would cry at such a waste of labor optimization, and while I could be doing something better besides soldering LEDs, that $10 would go so much better towards lunch for the next few days. I also have something like 500 LEDs.

The 3-speed feature was superfluous and a flagrant violation of my life philosophy – that if something is clear, it must be glowy. I locked the fans to the highest speed setting internally and used the spare clear-insulated wire for glowy hookup.

A little bit of heat helps the adhesives set, so I stuck the fans on top of the radiator for a little while.

Modding up the Pentium 4 cooler. Even though it had a broad copper base, it was just narrow enough to not fit a set of tapped holes for the through-bolts which hold the heat pump assembly together. I had to slam this on the mill at the Media Lab and carefully carve out these indentations to seat the little flanged nuts.

Clamping onto thin sheets of metal and milling using a 1/2″ endmill was an adventure I hope to never repeat, but it was disaster-free.

After wiping the surface down with thermal compound, I stuck it back on the inside and spun up the fan for a test. This one looks fine.

The outside fan is attached with clampy-screw things and now almost covers the entire external heat sink area. I wanted to mill cross-channels into the aluminum heat sink to increase the surface area more, but couldn’t find a narrow endmill. It probably doesn’t matter much in the long run anyway.

Ahh, color over-saturation. At this point in most projects is where I ditch elegant and planned engineering for something spontaneous that works. Here, some duct tape acts as a windguide and directs air from the heat sink over the electronics, which, running everything at over 100% capacity, kept me from getting close out of fear of explosion.

Blocking the side vents means air can only escape from the top vents, right over the electronics. It’s a form of active-passive cooling. The 120mm fan fills the entire indentation in the original sheet metal back like it was supposed to be there or something.

The result looks very much stock, which is what I was going for. Only when the lights turn off does a radioactive blue glow project onto the back wall.

Final cost: $38 for the fans and cooler assembly, throw on a little bit for my time, and add $65 for the original appliance, and I could have just bought a much better performing product, but that’s not fun at all!

Performance is better – it gets down to temperature much quicker for certain, but I haven’t taken a thermometer to it yet. I didn’t take a thermometer to the original either, and so any judgement in temperature difference is purely subjective.

It isnt’ quite at the liquid nitrogen level (that comes later), but it’s enough for now and satisfies my desire to not leave any appliance stock and untouched and any warranty unvoided.

Semi-mutated Minifridge gets its own project page soon!

Map shenanigans!

On a whim, I decided to check out the MIT campus on Windows Live Maps, something which I haven’t ever used before. What amused me the most was the “bird’s eye view” pictures, taken with the camera not pointed straight down, so the buildings had an illusion of depth.

Now, I understand that imaging companies don’t update their archives daily – and this is to be expected, unless someone has corporate-level OCD. Some times, though, the images are from years ago. Google Maps seems to still think the Walmart near where I lived in Atlanta (built in 2002) is still a dirt lot.

When you add in multiple-angle views to the situation, some times things get a little weird.

Let me direct your attention to the Stata Center, one of the examples of innovative architecture (or unarchitecture, depending on who you ask) on campus.

Repeat after me: THIS IS STAAAAAAAAAATAAAAAAAAAAAAAAAAA

This picture was probably taken in late 2003 to early 2004, when the building was in the last phases of construction. This was okay – Google Maps just shows a big construction-equipment-littered mess, even now.

However, as soon as I scooted north a bit, half of it disappeared.

Rotating the map to face west, it seems to be in a different phase of construction.

 And moving a bit eastward (farther away), it seems to be slightly different again.

The observant will also notice something missing in the previous 3 pictures that was present in the first: the large white-draped building across the street. That would be one of the newest buildings, the Brain & Cognitive Sciences center.

The direct overhead view is the most recent, as it shows both buildings completed. The brain-surgery-like covering over the B&CS is, of course, gone.