Development of a custom on-line ultrasonic vapour analyzer and flow meter for the ATLAS inner detector, with application to Cherenkov and gaseous charged particle detectors

Precision sound velocity measurements can simultaneously determine binary gas composition and flow. We have developed an analyzer with custom microcontroller-based electronics, currently used in the ATLAS Detector Control System, with numerous potential applications. Three instruments monitor C3F8 and CO2 coolant leak rates into the nitrogen envelopes of the ATLAS silicon microstrip and Pixel detectors. Two further instruments will aid operation of the new thermosiphon coolant recirculator: one of these will monitor air leaks into the low pressure condenser while the other will measure return vapour flow along with C3F8/C2F6 blend composition, should blend operation be necessary to protect the ATLAS silicon tracker under increasing LHC luminosity. We describe these instruments and their electronics.


Introduction
We describe a microcontroller-based combined ultrasonic gas mixture analyzer and flow meter for continuous real-time composition analysis of binary gas mixtures. Five instruments of this type will be used within the ATLAS Detector Control System (DCS) for flowmetry, coolant composition determination, and leak detection in the evaporative cooling systems of the ATLAS silicon tracker [1,2].
The instrument exploits the phenomenon whereby the sound velocity in a binary gas mixture at known temperature and pressure is a unique function of the molar concentration of the two components of differing molecular weight. The combined flow measurement and mixture analysis algorithm is implemented together with the user interface in WINCC-OA R [3] running on a Dell Poweredge R610 under Linux. Measurements of temperature, pressure, and sound transit times in opposite directions are used with a velocity-composition database for on-line determination of binary gas mixtures. The database is created from prior measurements in calibration mixtures or from theoretical derivations made with an appropriate equation of state. Figure 1 illustrates the connection architecture of the five ultrasonic instruments being incorporated into the ATLAS DCS. One analyzer has been used since January 2010 to monitor C 3 F 8 coolant leaks into the N 2 envelope of the Pixel detector. Individual leaking circuits have been identified through C 3 F 8 concentration changes of < 5 · 10 −5 . Similar analyzers will monitor the N 2 envelopes of the silicon microstrip tracker and the new Pixel B layer, respectively for C 3 F 8 and CO 2 coolant leaks.
The compressor-based C 3 F 8 evaporative cooling system of the ATLAS silicon tracker will be replaced with a thermosiphon [2] exploiting the hydrostatics of the 92 m shaft to the experimental cavern. At 60 kW dissipation 1.2 kg·s −1 (0.4 m 3 ·s −1 ) of vapour will return to the surface-level condenser. The flow will be calculated from sound transit time differences in opposite directions angled at 45 • to the stream. Although the thermosiphon will initially operate with C 3 F 8 , the coolant may be changed to a blend containing 20-25% C 2 F 6 to better protect the silicon against thermal runaway under increasing LHC luminosity [4]. The same instrument will be able to measure C 2 F 6 /C 3 F 8 blend composition with a precision of ∼ 3 · 10 −3 [5]. Another instrument will monitor possible air ingress to the condenser, which will operate below atmospheric pressure. When air concentration -1 -exceeds a predetermined threshold it will be automatically vented via a vacuum system to maintain condenser performance.
Although ultrasonic binary vapour analysis was first used for the N 2 /C 5 F 12 radiator of the SLD Cherenkov Ring Imaging Detector [6] and subsequently adopted in all major RICH detectors, including DELPHI, COMPASS, and LHCb, none of these applications exploited simultaneous flowmetry and vapour analysis in the same instrument. The instrument described here has many applications where knowledge of binary gas composition and flow are required. 2 Principle of operation of the electronics Figure 2 illustrates the major elements of the electronics, which is based on Analog Devices ADuC and Microchip R dsPIC33F microcontrollers and communicates using the MBED LPC1768 platform [7] via MODBUS TCP/IP over Ethernet. Custom timing electronics -internally implemented in the dsPIC33F microcontroller -generates 50 kHz ultrasound 'chirps' and synchronously starts a 40 MHz transit time clock, which is stopped by the first pulse crossing the user definable threshold (figure 3). The microcontroller also provides two 4-20 mA DAC signals proportional to transit time for use in external hardwired control systems. Analog inputs from NTC thermistors and pressure sensors are read in by the ADuC microcontroller. Figure 4 illustrates the amplification and biasing scheme, which is adapted to the Sen-sComp Model 600 instrument-grade 50 kHz capacitive-foil ultrasonic transducer [8] operated as -2 -2015 JINST 10 C03045 a transceiver. The DC bias voltage of around +300 V is provided by inexpensive custom DC-DC converters. When transmitting, the transducer foil is excited by down-going (300 V → 0 V) square wave pulses generated from LV precursors in the dsPIC33F microcontroller. For reception, the LV signal induced on the floating metal can of the transducer is DC-coupled to a signal chain containing a differential amplifier followed by a combined 4:1 analog multiplexer and linear programmable gain amplifier (Microchip MCP6S26). The following comparator is implemented in the dsPIC33F microcontroller. This new configuration replaces a previous version, where the foil signal was ACcoupled to the first amplification stage, and allows shorter transit times to be measured through elimination of the waiting time associated with the discharge of a HV DC blocking capacitor. Transit times in both directions are needed to compute the gas flow rate, while the average of the sound velocities in the two directions (or a unidirectional measurement in quasi-static gas) is used together with temperature and pressure to compute the binary gas composition by comparison with stored sound velocity vs. concentration look-up tables.
Various modes for pulse injection in the opposite directions can be selected in the dsPIC33F microcontroller. Pulse injection can be simultaneous, alternate after a 'blanking' delay, or alternate triggered by the reception of above-threshold pulses.
Presently, mixture calculations are made using sound velocity vs. concentration look-up tables stored in the supervisory computer. In future versions of the instrument these calculations may be made in an on-board microcontroller to allow full standalone operation.
Bidirectional transit times, vapour temperature and pressure measurements are buffered, averaged, time-stamped, and pipelined by a FIFO memory implemented in the dsPIC33F microcon--3 -  troller for transmission over Ethernet. For transit time measurements the FIFO buffer depth is 16, presently with median finding, while for temperature and pressure measurements the buffer depth is 4 with arithmetic mean averaging.

Results from various implementations of the instrument
The instrument has been implemented in several geometries, starting with a small "pinched axial" flow meter/analyzer [5] which has demonstrated a linearity of 2% of full scale for flows up to 230 litres per minute with simultaneous measurements of the C 2 F 6 concentration in C 2 F 6 /C 3 F 8 blends to a precision of ∼ 0.3%.
A second instrument has been used for more than a year to detect low levels of C 3 F 8 coolant leaking into the N 2 environmental gas surrounding the ATLAS silicon Pixel detector. Figure 5 illustrates an 18-month continuous log of the C 3 F 8 contamination of the Pixel detector envelope.
-4 -Fluctuations of the measured C 3 F 8 contamination are correlated with the development of leaks in some of the 88 individual cooling circuits that were identified by progressive turn-on or turn-off. A reduction in sound velocity of ∼ 0.86 m·s −1 from that of pure nitrogen is typically observed when all 88 circuits are operating.
The precision of mixture determination, δ(mix), at any concentration of the N 2 and C 3 F 8 components is given by; where m is the local slope of the sound velocity vs. concentration curve (m·s −1 ·[%C 3 F 8 ] −1 ). The average gradient of the sound velocity vs. concentration curve is -12.55 m·s −1 ·[%C 3 F 8 ] −1 for C 3 F 8 concentrations in the range 0%-0.5% in N 2 . The intrinsic sound velocity measurement uncertainty of ±0.05 m·s −1 yields, via eq. (3.1) a corresponding mixture uncertainty of ±0.004%.
The gas extraction and sampling system (figure 6) also allows measurement of C 3 F 8 leak rates into the N 2 gas envelope of the silicon microstrip (SCT) tracker, and CO 2 coolant leaks from the new "Insertable B-Layer" Pixel detector (IBL) into its external N 2 envelope.
Gas is continuously aspirated from six points (Pixel: 1; SCT: 4; IBL: 1) of the sub-detector environmental gas envelopes, and is sent for sequential analysis -via a matrix of normally-open (NO) and normally-closed (NC) pneumatic valves -in three ultrasonic instruments (see also figure 1). The entire sequence of operations is supervised by the WINCC-OA control system.
Gas exiting the three tubes is vented to the air extraction system for return to the surface. Injection points for automated periodic recalibration of the analysis tubes with pure nitrogen are also incorporated.
In the case of the ATLAS IBL, the smaller difference in molecular weight between CO 2 and N 2 results in a shallower slope to the sound velocity vs. concentration curve. For example, in the (0%-0.1%) molar range, of most interest in leak detection, the slope is -1.12 m·s −1 ·[%CO 2 ] −1 . This slope, taken in combination with a δc of ±0.05 m·s −1 would result in a mixture resolution of ±4.4 · 10 −4 . Efforts are underway to improve the precision of temperature control: the biggest contributor to the sound velocity uncertainty.
Initiated by the plans to replace the present underground compressor-driven C 3 F 8 circulation plant with a thermosiphon [2], two additional instruments have been constructed.
The first of these (figure 7) will monitor and eliminate air leaking into the surface condenser, which operates below atmospheric pressure. Although the difference in molecular weight between C 3 F 8 and air (respectively 188 and 29 units) is very large, this instrument operates well -5 -  -6 -in a mode where a light contaminant leaks into a heavy carrier. The slope for the corresponding end of the sound velocity vs. concentration curve is shallower, e.g., in the (0%-10%) molar range, of most interest in this application, the slope of the sound velocity vs. concentration curve is 0.53 m·s −1 ·[% air] −1 . This slope, taken in combination with the sound velocity measurement error of ±0.05 m·s −1 , results in a mixture resolution of ±9.4 · 10 −4 . The second instrument is a combined ultrasonic gas mixture analyzer and flow meter (figure 8) which has been installed in the 133 mm diameter C 3 F 8 coolant vapour return tube from the silicon tracker to the surface condenser, in which very high flow rates of around 1.2 kg·s −1 (> 400 l·s −1 ) are expected. Computational fluid dynamics studies [5] showed that only an angled crossing geometry, with the transducers not impinging on the gas flow, was suitable. The flow meter has been built with a 45 • crossing angle and calibrated in air against a commercial anemometer (Amprobe model TMA10A: 25 m·s −1 full-scale (FS) with intrinsic accuracy ±2% FS). The rms accuracy of the ultrasonic flow meter is equivalent to ±2% of the full-scale flow, of 10 m·s −1 , achievable with the available air supply. Tests of this instrument with fluorocarbons will be made as soon as the closed thermosiphon circulation system becomes available.

Conclusions and future applications
We have developed a combined, real-time flow meter and binary gas analyzer with custom electronics and dedicated SCADA software running under WINC-OA [3] (a CERN standard). First, a small "pinched axial" flow meter/analyzer has demonstrated a linearity of 2% of full scale for flows up to 230 l·min −1 with simultaneous measurements of C 2 F 6 /C 3 F 8 blend molar ratios to a precision of ∼ 0.3%. Another instrument, in continuous operation over more than 18 months, has demonstrated a sensitivity of < 4 · 10 −5 to C 3 F 8 leaks into the N 2 envelope of the ATLAS Pixel detector. The sampling system also includes analyzers for C 3 F 8 leaks from the SCT detector and CO 2 leaks from the IBL detector.
-7 - Lastly, two other instruments have been developed for the new ATLAS thermosiphon recirculator. One will detect and eliminate ingressed air from the surface condenser, which will operate below atmospheric pressure. The second instrument, with a sound path angled at 45 • to the gas flow, will measure fluorocarbon flows back to the condenser which can be as high as 0.4 m 3 s −1 when the SCT and Pixel detectors operate at their full dissipation of 60 kW. Tests in air have demonstrated a flow measurement precision of ±2% of full scale up to 10 m·s −1 and the instrument will soon be tested with fluorocarbons.
The instruments described in this work have many potential applications where continuous binary gas analysis is required in real-time; these include the analysis and flowmetry of hydrocarbon mixtures, vapour mixtures for semi-conductor manufacturing, and anaesthesia.