The MVD liquid cooling system cools 3 rows of low-voltage regulator chips (LDO) on each of the 4 endcap motherboards of the MVD. An aluminum tube (1) is glued on the board (2) with thermally conducting glue directly opposite of the LDO chips. Because of the severe mass constraints, and heat conductivity requirements, aluminum was chosen over copper stainless steel or plastic. Two short tube stubs, 0.25"OD protrude from the MVD. Custom-made couplings fit over these tube ends to make a water-tight connection and to make an immediate 90° turn; the space available between the tube ends and the magnet pole face is about 1cm. This is too short for any flexible tubing to make a 90° turn without folding. The cooling circuit is a closed-loop, no-pressure system, consisting of a circulating pump, a heat exchanger, a can to trap air bubbles, and an air bubble exhaust line. The plumbing is done in copper and flexible plastic tubing. postscript

Therefore, we would run the system initially with ordinary water, and after no-spill filling procedures were worked out, we would do the paperwork and switch to FC75.

The custom couplings that connect the cooling tube to the supply and return lines consisted of a a 90° piece, with a threaded part, over which a cap can be screwed. These parts were made of brass. The aluminum tube fits loosely through the cap and into the 90° piece. A small O-ring gets compressed between the thread and the bottom of the cap, providing a seal. In the picture, one such fitting is shown disassembled, one is assembled, and a spare cooling tube is also shown. bigger -->

On November 26, work started in the hall in order to mount the NTC counters, which would fit between the MVD and the pole face. Some of the work involved replacing the couplings with different ones, which would be less prone to leaks during installation - the little O-rings had given us trouble in the past. Also, the couplings and tubes needed to be rearranged to clear the NTC. It became apparent that some of the couplings had become frozen to the aluminum tubes. Only with considerable effort could all the couplings be removed. They had become welded together with oxides or other compounds (white, gray, black, green). Some of this is visible in the photo above. Clearly these encrustations had kept the system solid and leakproof, but cleaning this stuff off showed that a lot of the aluminum had been eaten away.

	bigger
	bigger
	bigger

(all photos)

Obviously, this was caused by galvanic corrosion between the aluminum tube and the brass coupling. They were probably touching, or at least very close in all 8 locations (one supply and one return per quadant).
A measurement of the resistivity of a sample of the water retrieved from the system was ~150 micromhos/cm. A sample of the local tap water tested at 100 micromhos/cm.

The new fittings were actually able to take care of the holes that are visible in the photos. Two quadrants were reconnected and were not leaking when water circulation was turned oback n.

What is not visible in these photographs is that in two locations, a hole was found at the point where the tube enters the MVD, and there is nothing that can be done to fix that on a short time scale. Thus two quadrant could not be operated.

With half of the MVD out of commission, and the remaining tubing in unknown condition, and perhaps not able to withstand the jostling of the NTC installation plus 2 months of running, it was decided to pull the detector out.

• Gluing a tube onto the motherboard (4)
• Making the endplate stack (4)
• Making new inner enclosures (2)
• Building the new couplings in all copper/brass (8)
• Making an insulating feedthrough in the aluminum endplate (8)
• Making a no-spill filling system (2)