Minutes Cooling Working Group (CWG) #5

Date of the meeting:
19/03/97
Place of the meeting:
112 R-018
Present:
P. Bonneau, M. Bosteels, A. Carraro, G. Dumont, J. Godliewski, H.P. Gugerli (Advanced Ceramics Corp.), M. Hatch, E. Hodin, M.J. Montesano (K. Technology / ACC), A. Onnela, P. Petagna, W. van Sprolant.

Agenda:

Short Summary:

Stabilised regions at different temperatures will adjoin one another in the future detectors; the lack of space and the low permissible gradients necessitate the use of active screens and their technologies were discussed.
A system for measuring low flow rates developed during direct electronic rack cooling tests was presented together with an idea for the acquisition of multiple temperature and flow rate channels while limiting the extent of the wiring.
Advanced Ceramics Corporation presented its very highly conductive materials and its thermal management technology intended for the electronic boards.

Detailed Minutes:
Comments from M. Hatch for ATLAS:
- Concerning the presentation by G. Hallewel. The work on the evaporative cooling method, as performed in Marseilles was presented. However, it should be made clear that cooling with Binary ice is the baseline solution for both the Pixel and SCT detectors.
-Safety considerations on coolant. The actual figure of 600 kg of methanol would need to be more accurately determined if indeed methanol was an acceptable additive (which it is probably not).
The pipework shown on the current layout drawings is based on an ethylene glycol/water mixture at -15°C and not methanol/water as stated. However, if methanol could be used (not probable) then these pipes could be reduced in diameter.
Moreover, M. Edwards is a member of the RAL ATLAS group and not of its Health & Safety Group.
These comments gave rise to a fresh discussion on the operating temperature of the silicon detectors and the effects on the materials budget of the use of a mixture based on glycol rather than methanol. It was found that a temperature of -7°C on the silicon would be a good compromise and, in order to keep within an acceptable materials budget, one of the items to be dealt with was the ÆT between the -7°C and the temperature of the coolant. The nearer that temperature was to -7°C, the smaller would be the pipe cross-sections (there was a considerable difference between a fluid at -10 and one at -15°C). Stress should therefore be laid on the transfer of heat between the Si and the heat-conveying fluid.
Leaving cryogenics aside, the LHC experiments would have at least two areas involving heat: the silicon environment at about -10°C and the other detectors at ambient temperature, i.e. about 20°C. To that was added the temporary problem of baking out the vacuum chamber. The temperature gradients in those areas had to be kept to the minimum to attain the mechanical stability required for the resolution of the detectors. In order, therefore, to make those regions thermally stable it was necessary to separate them by screens within the experiments where there was no room and should be no materials.
An effective passive screen would require several centimetres of insulating material -- which was obviously impossible. Therefore the only really effective means of achieving minimum thickness was the active screen consisting of a barrier cooled on the cold side and heated on the hot side and making it possible to set the two temperatures. Figure 5-1 Heating the hot side might raise some eyebrows, since there were also component to be cooled in those areas (e.g. ATLAS' TRT or the MSGCs of CMS), and some people had put forward the idea of recovering the cold from the screen to thermalise the hot region rather than heating the screen.
Besides the fact that that was not an easy matter at circuit level and a gradient was necessarily created in the region, the problem arose of the independence of the detectors on either side of the screen: what would happen when one was ON and the other OFF? In such a case it would be preferable to have active screens controlled independently of the detectors.
The periodic bake-out of the vacuum chamber raised an even more thorny problem since the gradient was much higher (~ 250°C) and the screen could not remain in place. Its fitting and removal therefore had to be carefully examined. The problem could be approached from the opposite direction, i.e. by baking the vacuum chamber out before installing the Pixel detector, as proposed by CMS.
On the technology side, the screen consisted of a heating film/insulating layer/cooled structure sandwich. Industry offered perfectly suitable heating films consisting of a polyimide (Kapton) substrate and resistive components deposited on it "printed-circuit fashion". Those films were very fine (~ 0.25 mm), flexible and strong. They could easily be made on request and thus made suitable for the desired power. A standard material (conductivity ~0.04 W/mK) was also adequate for the insulating layer, which was about 5 mm thick; a ceramic foam was preferable for safety and flammability reasons; a simple "air blade" could also be considered.
The delicate item was the cooled screen, as it had to have a cooling circuit ensuring a uniform temperature over its entire surface with a minimum of material. Use could be made of two Al foils with brazed or adhesively secured embossed circuits. The brazing technique on fine Al foils had to be developed by CERN's Brazing Section and included a few unknown factors regarding corrosion resistance. For adhesive securing the Resin Section was currently using a silicone which exhibited good radiation resistance .An interesting development would be to make the same type of screen using PEEK foils, which would be more advantagous in term of material budget. Such screens were nevertheless still very fragile and difficult to use in the form of half-shells 1 m in diameter and several metres long. One solution would be to make self-supporting half-shells of carbon fibres with cooling channels integrated into the shell thickness. From that aspect, the active heat screen became a highly technologically developed and also very expensive item.
Whichever technology was selected, there could be no hope of reducing the thickness of those screens to under 8 to 10 mm.
In order of power: ATLAS' SCT / TRT screen - thickness 8 mm - 26m2 - DeltaT=25 ==> 5 kW (cooling) + 5 kW (heating).
W. van Sprolant was a member of the ECP team making very precise evaluation tests on a system for cooling electronics crates by direct exchange with water-cooled screens (to be presented shortly). In the context of the thermal balances, and having failed to find a flowmeter on the market with the same order of precision as that for temperature measurement (<1%), the team had built a specific flowmeter with differential water columns. The flow rate detector concerned was simply a regulating valve (a flow rate distributor for a heating radiator) to which a transparent flexible tube was branched on each side. With a length of about 1.5 m, they made it possible to visualise a DeltaH difference between the two columns of water of about 1 m for a nominal flow rate of 1.5 [l/min]. The flow rate/DeltaH ratio adjustment and drawing its calibration curve were achieved by the "water potting" method using a graduated container and a chronometer. The uncertainty on the measured volume was about +1 [ml/l], or +0.1%. It was possible to measure a water level difference between the two columns of the order of a mm in 1 m for a flow rate resolution of the order of (?) [ml/min]. The flow rate measurement precision was then the sum of the calibration uncertainty (0.1%) and that of the DeltaH reading (0.1%), i.e. a total of +0.2%.
In order to improve the performance and regulation facilities, it was intended in future to replace the pressure loss regulating valve by a high-precision needle valve.
Finally, the flowmeter was easy to remove and could be used for other tests.
W. van Sprolant then put forward a technical proposal for a thermo-hydraulic probe for the sequential measurement of the temperature and flow rate of a heat-conveying fluid.
The problem of temperature measurement in the detectors had already been raised at previous meetings and one of the critical points was the wiring of the probes in the Trackers.
While for some, monitoring the temperature of the component to be cooled was sufficient, for others it was also important to know the temperature of the fluid cooling the component, especially to check the cooling system for proper operation in the event of unusual occurrences. On the same theme, there were rarely flowmeters in the installations and it was noticed in the course of shutdowns that those responsible for monitoring cooling systems were installing them nearly everywhere.
Effective monitoring would thus require a knowledge of the temperature and flow rate of each cooling channel and hence 6 to 8 electric wires per channel using traditional methods.
The current proposal concerned a new type of single probe for the sequential measurement of the temperature (T) followed by the flow rate (F) of a heat-conveying liquid. Its principle was based on a double measurement (T + F) of the electrical resistance of a single probe., A different thermodynamic balance between the probe and the fluid concerned would be required depending on the type of parameter of the heat-conveying fluid (T/F) to be detected.
The measurement of the fluid temperature (T) used the conventional method of the inserted electrical resistance. That resistance was 100 ½ at 0°C for a probe on plate PT 100. It involved the injection of a very weak and accurately known current into the probe (~ mA) to measure the voltage drop at its terminals (~ 100 mV), which would be proportional to said resistance, which in turn was linked to the temperature transmitted to the probe by the fluid. During that measurement, the electric power dissipated by the measuring current in the probe was negligible (~ 0.1 mW). After allowing a certain time to attain thermal balance, the temperature of the internal resistor of the probe was equal to that of the fluid.
To measure the flow rate of the fluid (F) in which the probe was immersed, a measuring current of about 100 mA was injected. The electric power dissipated in the probe then became significant and increased exponentially (1 to 2 W), since, for the same resistance, the power was proportional to the square of the electric current and because the increase in the electrical resistance of the probe caused it to heat up (100 to 200 ohm). The voltage drop was then 10 to 20 V. The heating of the probe was counterbalanced by the effectiveness of its cooling and depended on the thermal and hydraulic conditions of the heat-conveying fluid (temperature, flow rate, "boundary layer"). Here, too, a fresh thermodynamic balance was struck by the flow of heat imposed by the electric current in the probe. By suitable calibration and on the basis of the electrical resistance (for F measurement) and the initial temperature of the fluid (for T measurement), it was possible to find the corresponding flow rate of the fluid. The sensitivity of the probe could be adjusted by a retention grid of the "boundary layer" which the fluid formed around the probe and hampered its cooling.
Assembling the thermohydraulic probe (THP) and it electric wiring were relatively simple. Fitted at the outlet of each cooling circuit towards the main manifold, the THP was close to its neighbours, thus permitting short wiring lengths in series with a single current source. That current source powered groups of THPs (e.g. 10 per group) depending on the current level required for the two types of measurement (T and F). A single measuring wire was connected to each interconnection between two adjacent probes (a short, heavy-gauge bridging wire for the electric current). That then made it possible to measure the voltage drop on each THP (the image of its corresponding electrical resistance). In all, the significant wiring was reduced to a single measuring wire per probe (instead of 6 to 8 for the conventional method).
All the hydraulic circuits of the cooling system were then monitored sequentially by the successive read-out from each THP. A given THP was selected via a double multiplexer which selected the wire on either side of the probe concerned. Temperatures were measured on groups of probes, followed by the measurement of the flow rates alone. The level of the current applied by the source thus changed less quickly and the probes were given the time needed to regain their corresponding thermodynamic balance. The measuring time, including the switching time, was then only a few milliseconds and a few seconds for updating the complete image of the state of the detector cooling system.
That proposal of course had to be tested and adjusted before its application.
M.J. Montesano introduced the speaker. Advanced Ceramics Corporation was a young company, but one with considerable experience in the production of ceramic components (Noxide powders, hot pressed, CVD shape), since it was initially a division of Union Carbide. The presentation related to the last-mentioned material, called TC 1050, developed with K Technology Company, consisting of a TPG plate enclosed in a material which may be a metal, a ceramic or a composite. Thus TC 1050 took the form of a fine plate (minimum thickness 1 mm) intended as the substrate for high-density electronic components, with the aim of conducting their heat to the edges of the plate. Several samples were shown.
Thermal Pyrolytic Graphite had very special properties:
Heat conductivity
1700 W/mK (a & b-Axis)
25 W/mK (c-Axis)
Density
2.26 g/cm3
Coefficient of Thermal Expansion
-1.0 ppm/K (a & b-Axis)
25.0 ppm/K (c-Axis)
Thermal Diffusivity
9.8 cm2/s
Specific Heat
0.84 kJ/kg.K
Tensile Strength
1000 Ksi (a & b-Axis)
0 Ksi (c-Axis)

Depending on the material of the envelope, the TC 1050 plates had the (measured) characteristics below:

Encapsulation Material
Thermal Conductivity [W/m.K]
Mass Density [g/cm3]
Aluminium Alloy
1140
2.52
Copper
1142
4.45
AlSiC
1176
2.95
Carbon-Carbon
982
2.02
Pitch Fibre/Al
1020
2.30
Pitch Fibre/Polymer
1180
2.13

It was thus clear that TC 1050 had a much greater heat transmission capacity than anything in existence except diamond.
Plates were mass-produced to VME standard; the locations of the components had to be specified because, as the material was anisotropic, thermal bridges had to be included in the plate at selected points.
The price for short runs varied from $ 2 to 5 /cm2 and ACC could examine any kind.
There were many technical reports available.

The next meeting is to be held on Wednesday 16/04/97, Building 40, Room R-B10 from 10 a.m. to 12 noon.

Agenda:

CERN/P.BONNEAU/30/05/97