Capn's Blog

Monday, April 17, 2006

I fixed my computer with glue and newspaper!

My Windows computer at home has three fans in it - one for the power supply, one for the CPU, and one for the video card. Amongst owners of this model computer, the fan on the video card has a bad reputation for failing after about a year. If it completely stops, the video card overheats and sometimes dies. Since a replacement is about US$250, it's an expensive failure.

The video card fan in my computer has been noisy for many months. Sometimes it spins faster, sometimes slower, but always it's groaning and making "I don't want to do this" noises. I knew it was only a matter of time before it failed.

Last wednesday night, I was using the computer and I heard the fan slow and finally stop. To the friends I was chatting with at the time, I had to say "bye, I must go before my computer catches fire!". Stretching the truth a little, but it sounds funny :-)

Three-legged fanI pulled the computer to bits and removed the graphics card. The fan was very hard to turn, and wasn't fit for duty anymore. Time to replace it!

The only problem is that the fan doesn't have the standard square cage. Instead, it 's fastened to the heatsink by three tiny screws at the end of three legs.

Well, here's the space the card goes in. There's nothing next to it, which means I have room to make an airpipe to carry air from another fan.

My idea was to mount a new fan at right angles to the old fan, above the Wifi card (that tan-coloured rectangle). Then duct the air from the new fan onto the heatsink of the video card.

First step was to get a block of polystyrene foam. I carved the polystyrene into a shoe shape, being careful not to touch the foam onto any of the circuit boards because of static electricity.

After making the shoe shape, I got a page from an old newspaper and tore it into squares of about 1 inch. Using water-based paste glue, I started gluing the torn-up squares onto the foam, covering it on every side. After applying several layers, I dried the glue in the oven at 60°C until all the paper was dry and hard.

Once it was hard, I cut the square "heel" out of the shoe for the fan, and the "sole" out of the shoe near the toes for the heatsink. I then extracted all the polystyrene, leaving an empty shoe-shaped shell.

The paper had become surprisingly stiff cardboard. It held its shape very well and knocking on it made a very pleasing rigid hollow sound.

I then took some solvent-based craft glue and glued the fan onto the end of the shoe. When I applied power to the fan, there was an impressive amount of air coming out the bottom of the shoe.

Finally, I glued the shoe onto the heatsink of the graphics card. After drying in the oven, the whole card-shoe assembly became very solid, with the new fan really feeling part of the card. I used an elastic band when drying it, but I decided to leave it on permanently.

Here's a closeup of where the pipe attaches to the heatsink. The glue forms a seam around the top of where the fan used to be.

For power, I used the old trick of running the fan under the rated 12V. This is much quieter, if you don't need full airflow. In this case, I run the fan at 7V by connecting it between the 12V and 5V wires on an unused power connector. You can find out a lot more at 7 volts, a first-rate site.

Finally, I plugged the card into the computer and connected everything up. Fortunately, everything seems to have survived the static electricity, ovens, glues and everything else, and is now working fine.

The video card fan is still the noisiest in the computer (the CPU fan in this computer is wonderfully quiet) but it's a lot quieter than it used to be, and will do the job for another year or two.

Technical details:
  • Medion MD8083 3GHz P4.
  • ATI Radeon 9800XXL graphics card.

Sunday, April 02, 2006

"Flexible working hours"

It's half-past-11 on Sunday night, and I'm at work. Why? What would possess someone to leave their family and come to work?

My present company (OPEN Networks) produces ADSL modems. In particular, the modems are manufactured in Asia but we put our own software on them. The special software makes the modems substantially faster and more reliable that anyone else's, and it's because we have a team here (which I'm part of) that has worked hard to improve the software and fix so many bugs.

Some bugs are easier to fix than others. Usually a customer or our testing department tells us what to do to cause a problem. We can do that ourselves, find out what's happening inside, then fix it.

Some bugs are much harder to find. We may not know of a way to make the bug happen, and we will have to wait months to see the problem again.

We have one such fault that was reported nearly a year ago. It happens to quite a few customers but I've only seen it once before. And that means it's not possible to find and fix it.

On Friday, someone in the office had this particular problem with one of our modems. So I want to investigate the problem - haha, finally we caught a "live one"! And I have to do it before someone reboots the box, otherwise the problem will go away.

Sadly, I didn't get a chance to look at it on Friday. And I don't like doing it with people looking over my shoulder. So I'm at work tonight to investigate the problem.

I had one hypothesis for what's causing the problem, but I did some tests tonight and my hypothesis is wrong. Now I have to do some more thinking, imagine what else it might be. Then I have to work out how I can prove or disprove my theory. I'm going to keep doing that until I find the cause of the problem.

I always imagine customers using our products, and how frustrated they must feel when it doesn't work properly. If I can find and fix the problem, I will feel a great sense of satisfaction because I've made something that people can really use in their lives, and the product works well. That for me is why I like being a software engineer.

Which doesn't explain why I'm at work on a Sunday night. Well, it's give and take. I'll be getting in late to work on Tuesday morning and I have to leave early on Tuesday afternoon. My company is flexible about this. As long as I'm more or less spending enough time in the office and enough time working on problems, I can choose my own hours.

If I come in tonight, I have the hours to make up for Tuesday, I can work in peace, I'll look good to my peers tomorrow, and maybe I can even find the problem and get a buzz!