Condensation
Condensation
That was a word that I didn't have to deal with much until recently. With summer in full force in Virginia and me running some cold processors, I became painfully aware of condensation's effects. Water was condensing on the SEPP through gaps in the insulation and dripped down on to my AGP video card. After a couple of drops, I started getting some interesting effects on the screen. Luckily I didn't toast anything, but I came close enough that steps had to be taken.
First, I had to revisit some middle-school science to remember what it was that I was dealing with and why it formed. For those of you who actually remember those classes, you can skip the next section, but I needed a refresher course.
First, I needed a definition of the terms.
Humidity
Humidity is a measure of the amount of water vapor in the air. It can be defined as the ratio of water vapor "mixed with" each sample of air. It is usually expressed as the number of grams of water vapor in each kilogram of air. In the atmosphere, the ratio can vary from nearly zero (in deserts, polar regions, and at high altitudes) to as much as 30 grams per kilogram (in warm, moist, tropical regions).
Relative Humidity
Relative humidity (RH) is a measure of how much water vapor is actually in the air relative to (a percentage of) the air's maximum capacity to hold water vapor. The relative humidity of a sample of air can be expressed as:
amount of water vapor in the air
RH = ------------------------------------------------------- x 100
air's capacity to hold water vapor
By this definition, RH can change in two ways. It can change if the actual amount of water vapor in the air changes; and/or if the maximum capacity of the air to hold water vapor changes. Since the capacity of air to hold water vapor changes when the temperature changes, it follows that relative humidity depends both on the amount of water vapor in the air and on the temperature of the air.
Dew Point Temperature
The Dew Point Temperature is the temperature at which condensation first starts to form. If air that was not saturated with moisture started to cool, its capacity to hold water vapor would decrease but the actual amount of water vapor in the air wouldn't change. If the cooling continued, the capacity of the air to hold water vapor would continue decreasing until it could hold no more moisture. At that point, the air would then (by definition) be saturated. The temperature at which this occurs is called the dew point temperature.
At a particular pressure, the dew point temperature (the temperature at which a parcel of air would become saturated if it were cooled) depends only on how much water vapor is actually in the air. The more water vapor there is in the air, the higher the dew point temperature is. If saturated air cools further (below its dew point) then the air cannot hold as much vapor as it actually contains, and the excess vapor must condense to become liquid water.
If air at 70°F and 50% relative humidity is cooled to 52°F, the relative humidity will reach 100% and condensation will begin.
Condensation
Condensation is the forming of water from water vapor. It takes place when warm, moisture-laden air comes in contact with a cold surface. In fact, condensation can only take place on a cold surface. Cold, of course, is a relative term.
Controlling Condensation
There are a few methods for stopping condensation on a cold surface. Some are more useful for the overclocker than others. Here are the four I've considered.
Heating
This method relies on raising the temperature of the cold surface so that it stays above the dew point temperature. Electrical wraps, inferred heaters and warm air blowers are sometimes used for this purpose. A good method, but not for my application.
Dehumidification
This method relies on reducing the relative humidity of the air by absorbing the moisture out of it. As the relative humidity drops, the dew point temperature is lowered. Surfaces that were previously below the dew point temperature are now above it and condensation does not occur. Desiccants are used for this and work well in a contained area. Very useful for those of us considering installing their motherboard in a Frigidaire. Less useful when using the typical computer case.
Air Circulation
This method relies on the continuous supply of air of the same or lower relative humidity than the air in the vicinity of the cold surface. This is a good method to use in conjunction with other methods.
Thermal Insulation
This method relies on placing sufficient thickness of insulation over a cold surface so that the exterior surface of the insulation exposed to the moist air stays above the Dew Point Temperature. This method works on the principle that if the moist air can't get to the cold surface, condensation can't occur. This seems to be the best method for extremely cooled processors.
Sealing the Processor
Because of the fact that it usually takes me a few tries to get things working the way they should, and because of my close call with dripping water, I decided that I would "waterproof" the Celeron. I didn't want toasted celery if my insulation proved lacking.
There are some good articles on waterproofing on the web with some different takes on the ways you can go about it. Most people use some type of commercially available conformal coating. However, there's always someone who goes about things in a unique way. The "Vaseline solution" at the Hack Shack is one such unique way.
After reading the ingredients of the different conformal coatings, I found that a number of them were polyurethane based compounds. Having just done a refinish of our hardwood floors a while back, I happened to have some poly made by a company called Deft. I coated an old circuit board with it and not surprisingly it worked great. The board belonged to a digital multimeter with a bad ammeter shunt and it worked as well as I could have asked for while submerged in water up to the display.
Tape it and Coat it
I started by taping off the edge connector and slug. I then coated the front side by applying the poly with a foam brush. For the pedestal with its exposed pins, I worked on one side at a time, forcing the poly under the pedestal and letting it flow. After drying for a few hours, I could apply the next coat and fill the space a little higher. After the fourth coat, I had completely sealed the space.
I coated the entire SEPP a few more times to ensure that I had not missed any spots where moisture could enter. This process took a couple of days to complete, but I figured that the time was worth not having to worry about short circuits.
 |
 |
Insulating the Celeron
Well, I'd done my homework and read everything I could get my hands on about condensation and how various manufacturers and others had dealt with it. I'd played with the "foam in a can" and made a mess all over the workshop without achieving a great deal of success. I'd stuck a thermocouple all over the various parts of my heat sink, backing plate, and insulation and measured temperatures. One thing I saw caused me to re-think some previous notions I had about backing plates and trying to cool the back side of a Celeron.
I stated some time ago that cooling the back side of the Celeron was an exercise in futility. My reasoning was that the fiberboard was an insulator and doesn't transmit heat well. While I still believe that this is not the most effective way to cool your Celeron, I was surprised to find that the Aavid heat sink I used as a backing plate was getting cold enough to build up condensation when used with a peltier. The TEC was actually drawing heat through the slug, wafer inside, fiber circuit board, duct tape (to protect the back of the Cele), and dropping the temperature of the Aavid. Not just dropping it a bit, but by more than 20°F below room temp!
With this in mind, I started designing my new insulation. Undoubtedly this design will be refined through trial and error, but these were my considerations:
• Get rid of the Aavid and insulate the back of the Cele.
• Design the backing plate to fit the Abit SEPP holders on the board to hold the Cele steady.
• Try to make the Celeron as air tight as possible - seal all seams on the insulation.
• Insulate the screws that hold the sandwich together so the heat sink can't transmit heat to the cold plate.
• Try to find an insulation that is superior to egg cartons and meat trays.
I succeeded on all but the last item. I know that there's better insulation available, I just haven't come across it yet. Here's a cut-away view of how the parts stack together.
Here are the parts that make up insulation - from the backing plate to the top of the Celeron.
From the top; fiberglass, insulation, Cele, insulation.
And a couple of shots of the pieces being assembled. The pieces are held tight to each other with waterproof, double-sided tape. Well, it's not really tape, more like the adhesive from tape. Some stuff I had left over from my days with VW. It's used to hold the waterproof door panel liners on. (It's expensive and silicone sealant would probably work as well, but here is the part number if you're interested ... AKL 434 019 25.)
.
The rest of the insulation stacks on top of this unit. It is a bit of a trick to get the right thickness of insulation between each stage so that the plates make good contact while keeping the outside air from penetrating through to the cold pieces. Once you get the proper thickness for each piece, it goes together quickly.
Performance
Adding this insulation dropped the internal diode temp by about 12°F at idle and about 8° when gaming. The best part of it all is that with this new insulation attached to the watercooler/TEC heat sink and running some sub zero temps at idle and in the low 40's when being pushed, I have yet to see even a hint of condensation. This is with room temperatures in the mid-eighties, and with the kind of humidity that soaks you when you step outside. Of course, no sooner did I finish this project when I read ( at Melcor, I think.) that if one type of foam is good, two types are better. One close to the chilled item to insulate it and another outer layer to act as a vapor barrier. Maybe I'll start on that next week.