Last week, the newly-built a/c hoses went into the shark, providing the Interstate highway connecting the cities of the compressor, evaporator, condenser, and drier. The next step is leak-testing that highway, because—well, actually, that’s a terrible metaphor. Forget I mentioned it.

Seriously, though, leak-testing a system is quite possibly the most important part of the a/c repair, rejuvenation, or retrofitting process. If you are repairing a system, the most common reason for repair is that the system has no refrigerant in it, meaning it leaked out, so you want to be sure that it doesn’t happen again. If you are upgrading or performing a from-scratch installation, after going to all that work you want the system to be tight and leak-free.

In either case, you leak-test. Get it right, and you’re just an evacuation and a recharge away from having a car that reliably gets cold for years. Get it wrong, and you’re just an evacuation and a recharge away from chasing leaks for years.

To be clear, when I say “leak-testing,” I’m talking about making sure that the system is tight by pumping it full of nitrogen or some other inert gas before you evacuate it and charge it with refrigerant. (Once it’s charged, if it seems like it’s leaking, there are leak-detection tools such as chemical sniffers and dye; that’s an overlapping but separate issue. It’s really best to do this prophylactically.)

There’s a school of thought that says pressurizing the system isn’t necessary at all, and that instead you can find out if it’s leaking when you perform the necessary evacuation, which uses a vacuum pump to draw a deep vacuum and boil off any moisture (we’ll describe this in the final installment of this series).

Doing it that way is better than nothing, but pressure-testing is preferable. Here’s why.

From a practical standpoint, leaks fall into three broad categories. The first is leaks that are so large that the system won’t hold pressure for more than a few seconds. In this case, the leak should be so loud that you can find it by simply listening while also feeling for it with your hands.

The second is medium-sized leaks that you may be able to hear and feel, but which may require the use of soap solution to be found.

The third is small, silent leaks that can only be found with a leak-detection aid, be it soap solution or something else.

In all three cases, pressurizing works better than vacuum-testing because you can leave the pressure in the gas bottle turned on, which generates evidence of the leak. In contrast, when you hook up a vacuum pump, it only draws the vacuum while the pump is running, and the pump is loud, making it impossible to localize the leak by ear. If you shut it off, the vacuum quickly goes away. And you can’t feel a vacuum leak with your hands as easily as an under-pressure leak. And you really don’t want to spray soapy solution when the system is under vacuum; the last thing you want is soap and water drawn into the system.

Put another way, pressure-testing the system provides both leak testing and a means of leak detection. Vacuum-testing does not—or doesn’t do it as well. Now, you can pull a vacuum, wait overnight, and see if the needle on the gauge has moved. That’s fine, but it’s just not as accurate as pressure-testing. When you pull a vacuum, you’re drawing the system down as close to 30 inHg (inches of mercury) as possible in order to boil off any moisture in the system. This uses the small negative part of the display scale on the blue pressure gauge as shown below. Because the range on the gauge between 30 inHg and zero is so small—less than an inch on the scale—a small drop in vacuum is difficult to see on the gauge.

In contrast, when you put the system under 120 psi of pressure, you’re using the full range of the blue gauge, which makes it easy to see a small drop in pressure.

In order to pressure-test an a/c system, you need to buy or rent a nitrogen bottle and a regulator from a welding-supply or other store. Nitrogen is used because it’s a dry gas, and you really don’t want to run the risk of introducing moisture into the system, because moisture can react with the refrigerant and oil to produce acids. Some folks pressure-test using a standard shop air compressor with an in-line moisture separator, but I’ve never wanted to run the risk. I’ve read that other folks pressurize with a mix of CO2 and a gas like argon, and then use a chemical sniffer to check for leaks, but I’ve always used just nitrogen. You also need a ½"-flare-to-¼"-NPT adapter to connect the regulator on the bottle to a standard a/c manifold-gauge set.

Once you have the necessary pieces, you use the yellow hose to connect the tank output to the gauge set (in the picture above, the curly blue hose acts as an extension between the regulator and the yellow hose, with the adapter between them), and connect the red (high pressure) and blue (low pressure) hoses to the fittings on your a/c system. If for some reason you can only access one of the fittings but not both, that’s fine; the pressure will very quickly equalize on both sides of the system. On the shark, I’m using the charging fittings on the back of the compressor, so that’s where the red and blue hoses go.

Set the regulator on the nitrogen bottle to the full-range reading on the blue gauge. It’s typically 120 psi. It doesn’t matter if the regulator pressure is exactly 120 psi; in this first step, you’re just checking for large leaks. The main thing is to be sure that the regulator isn’t set to deliver hundreds of psi that can blow the system up. If you’re not sure what the setting is on the regulator, unscrew the T valve on the regulator all the way, partially crack open the main valve on the bottle, and slowly screw the T in until it shows about 120 psi on the gauge on the bottle.

Now slowly open the blue and red knobs on the manifold-gauge set (or only one knob if only one hose is connected); let the nitrogen flood in until the pressure stabilizes, watch the manifold-gauge set, and listen. Often, the first time you do this with a newly-repaired a/c system, you’ll hear nitrogen hissing out and see the gauge readings drop like a stone. This is clear evidence of a major leak that you should be able to localize with your ears and your hands. Most of the time a leak of this size will be coming from where a hose fitting is attached to a connection on one of the major components (the compressor, condenser, evaporator, or drier). Check and re-tighten as necessary. Don’t forget to check the connections to the gauge set itself, as they can certainly be a source of leaks. I’ve also found major leaks from hoses that I’d improperly crimped.

If the system passes the gross leak test, you next look for medium-size leaks. Re-check the pressure reading on your gauges; again, you want it to be near the full-range reading on the blue gauge, which is usually 120 psi. It doesn’t need to be exact, but when it’s in the ball park, close both of the knobs on the gauge set and the big knob on the nitrogen bottle, and record what the pressure is on the blue gauge. You can mark the reading with a piece of tape, or write the number down, or take a picture with your phone.

Next, wait. Re-check the gauge at regular intervals. If, after five minutes, the reading is dropping, you have a leak you need to locate; but if it looks stable, I let it sit for at least overnight before declaring the system tight. The readings may drop if the temperature in your garage falls, but they should rebound as the temperature goes back up. If the pressure reading steadily drifts down overnight, and still steadily drifts down as the sun warms your garage the next day, you have a leak—and if you’re trying to convince yourself you don’t, you’re in denial.

If the gauges show that you have a leak, use a soapy leak-detecting solution such as Big Blu from Refrigerant Technologies. These purpose-built solutions are way better at showing up small leaks than dishwasher fluid.

Spray the solution on all of the hose-to-component connections where leaks are most likely to occur. As I said last week, flare fittings sometimes require a lot of torque to get them to seal.

For tiny leaks, the bubbles may take hours to accumulate, but when they do, they often look like small clusters of insect eggs. The picture below is from replacing the condenser on my 3.0CSi, after which I found a leak that turned out to be coming from the joint in the nineteen-year-old crimped-on fitting itself, not from its connection to the drier.

If you’ve examined all of the connections thoroughly and haven’t found the source of the leak, you’ll need to spray the solution on the components themselves. It is possible for compressors, even new ones, to leak at the shaft seal behind the pulley, and for condensers, even new ones, to leak where the tubes are joined.

If you still can’t find the source of the leak, it’s possible that it’s inside the evaporator assembly, in which case you’ll cry and wish you had followed my advice a few weeks back to pressure-test it with nitrogen before it went in the car, probably in that order.

I went through all these procedures with the shark, and the only leak I had was on the flare fitting at the inlet to the drier. This was ironic, since, as I wrote last week, I'd vacillated about whether to use an O-ring or a flare drier, ordered both, and used neither because I’d forgotten that on vintage BMWs the drier inlet and outlet fittings aren’t the same size; then I found a never-used drier with proper-size flare fittings sitting in my basement and installed it.

The leak was quite troublesome. I finally sealed it by using a copper flare washer one size larger than the fitting size, and applying what felt like a fitting-stripping amount of torque—but sometimes that’s what it takes to get a flare fitting sealed up tight.

Note that you can crank the pressure up higher than 120 psi, but be careful. When the a/c system is running, the components conspire to keep the high-side and low-side pressures separate. The high-side pressure may reach as high as 300 psi on a very hot day, but on the low side, which runs from the expansion valve output to the evaporator to the compressor inlet, when the system is running correctly, the pressure is usually in the 20 to 40 psi range. In contrast, during this kind of pressure-testing, the pressure is equalized in both the high and low sides of the system. I’ve always had concerns that evaporators might not be designed to take the kind of pressures that condensers do, so I don’t ever test at pressures above 150 psi. 120 psi works well, since, as I said, it maximizes the use of the scale on the blue gauge.

I now apparently have a leak-free system. Wow! That leaves just the electrical connection, then evacuation and recharge. Think I can get it all done before the snow flies? I’d say it’s going to be tight, but it already is.—Rob Siegel

Rob’s new book, Ran When Parked: How I Resurrected a Decade-Dead 1972 BMW 2002tii and Road-Tripped it a Thousand Miles Back Home, and How You Can, Too, is now available on Amazon. Or you can order personally inscribed copies through Rob’s website: