A Burn-In Apparatus for the ATLAS Tile Calorimeter Phase-II Upgrade Transformer-Coupled Buck Converters

Abstract

The upgrade of the A Toroidal LHC ApparatuS (ATLAS) hadronic Tile Calorimeter (TileCal) Low-Voltage Power Supply (LVPS) forms a part of the Phase-II Upgrade preparations undertaken by the ATLAS experiment for the data taking during the High-Luminosity Large Hadron Collider era. This paper serves to provide a detailed overview of the development of a Burn-in test station for an upgraded LVPS component known as a Brick. The production, quality assurance testing, and all associated apparatus are being jointly undertaken by the University of the Witwatersrand (Wits) and the University of Texas at Arlington (UTA). These Bricks are radiation-hard transformer-coupled buck converters that function to step-down bulk 200 VDC power to the 10 VDC required by the on-detector electronics. To ensure the high reliability of the Bricks, once installed within the TileCal, a Burn-in test station has been designed and built. The Burn-in station functions to implement a Burn-in procedure on eight Bricks simultaneously. This procedure subjects the Bricks to sub-optimal operating conditions, which function to accelerate their ageing, as well as to stimulate failure mechanisms. This results in elements of the Brick that would fail prematurely within the TileCal failing within the Burn-in station or experience performance degradation that can be detected by follow-up testing effectively screening out the non-performative sub-population. The Burn-in station is of fully custom design in both its hardware and software. The development of the test station will be explored in detail; the preliminary Burn-in procedure to be employed will be provided; the preliminary and final commissioning of the test station will be presented. The paper will culminate in the presentation and discussion of the Burn-in of a V8.4.2 Brick and the future outlook of the project.

1. Introduction

In the year 2029, the start of the operation of the High-Luminosity Large Hadron Collider (HL-LHC), located at the European Organization for Nuclear Research (CERN), is planned with a foreseen peak luminosity of $5\times {10}^{34}$ cm⁻² s⁻¹ in order to collect 4000 fb⁻¹ of proton–proton collisions by the end of its operation [1]. The resulting HL-LHC environment has necessitated the Phase-II Upgrade of the A Toroidal LHC ApparatuS (ATLAS) general-purpose detector [2,3]. The ATLAS Hadronic Tile Calorimeter (TileCal) sub-detector will similarly receive an upgrade to both its on- and off-detector electronics to meet the requirements of a 1MHz trigger and higher ambient radiation and to ensure better performance under high pile-up conditions [4,5].

The TileCal is a sampling calorimeter that forms the central region of the hadronic calorimeter of ATLAS and is located at $\left|\eta \right|<1.7$ and 2.28 m < r < 4.23 m. The ATLAS experiment makes use of a right-handed coordinate system with its origin located at the nominal Interaction Point (IP) in the centre of the detector and the z-axis along the beam pipe. The x-axis points from the IP to the centre of the LHC ring, and the y-axis points upwards. Cylindrical coordinates (r,$\varphi$) are used in the transverse plane, where $\varphi$ is defined as the azimuthal angle around the z-axis. The pseudorapidity is defined in terms of the polar angle $\theta$ as $\eta =-lntan(\theta /2)$. The detector performs several critical functions within the ATLAS experiment such as the measurement and reconstruction of hadrons, jets, hadronic decays of $\tau$-leptons, and missing transverse energy [6]. It also contributes to muon identification and provides inputs to the Level-1 calorimeter trigger system [7,8]. The readiness of the TileCal for Run1 of the LHC, which took place during 2011 and 2012, was presented in a paper by the ATLAS collaboration [9].

The TileCal is physically partitioned into three cylindrical barrel regions along the beam axis, as seen in Figure 1. These are known as the Long Barrel located at $\left|\eta \right|<1.0$ and Extended Barrels, of which there are two, located at $0.8<\left|\eta \right|<1.7$. Each barrel region consists of 64 wedge-shaped modules that cover the azimuthal angle $▵\varphi \sim 0.1$ rad. The modules are composed of plastic scintillator tiles, functioning as the active media, inter-spaced by steel absorber plates. The ultraviolet light produced by the particle–matter interaction is transported by wavelength shifting fibres to Photo-Multiplier Tubes (PMTs), thereby producing signals that are sent to the on-detector Front-End (FE) electronics. The PMTs and FE electronics of a module will be housed within a new configuration. This configuration makes use of three-meter-long drawers, known as Super-Drawers (SDs), which are located at the outer radius of the TileCal. The Long and Extended Barrel modules utilise two SDs and one SD, respectively, resulting in a total of 256 SDs.

A Low-Voltage Power Supply (LVPS) is located adjacent to the SD that it services and is housed within a shielded container known as a finger. Each finger LVPS (fLVPS) contains a fuse board, an Embedded Local Monitoring Board (ELMB) and its associated motherboard, a wiring harness, as well as eight transformer-coupled buck converters known as Bricks [11]. Each of the eight Bricks function to step-down the $200$ VDC power received from off-detector power supplies to the $10$ VDC power required by the Phase-II Upgrade FE electronics.

2. fLVPS Brick History, Design, and Function

All of the TileCal fLVPS Brick versions are of an iterative design. The V6.5.4 Brick was the first to be installed and operated within the TileCal in 2007 [12]. The iterative nature of the fLVPS Brick design is motivated not only due to the success of previous designs and the inherent benefits of refining a well-understood system, but also due to the stringent design limitations imposed by their unique operating environment and location. One such limitation is that the overall physical dimensions of the fLVPS are set at 175 mm × 170 mm × 156 mm by the size of the available internal volume of the “fingers”, where they must fit. This results in a strong limitation being placed on the form factor of the Bricks. This restriction is still imposed for the Phase-II Upgrade Bricks as the volume of the fingers will not be altered during the upgrade.

The V6.5.4 Bricks functioned well at first, but began to exhibit a sensitivity for trips, which became apparent as the luminosity of the LHC was sequentially increased during commissioning. A trip is defined as a “spontaneous” event in which a Brick switches off. This trip rate scaled with the luminosity of the beam, further indicating that the performance of theV6.5.4 Brick would only further deteriorate and was, therefore, not fit for the purpose [13]. To address the issue of trips, as well as implement additional design changes motivated by experience gained with the V6.5.4 Bricks, the V7.5.0 Bricks were designed and subsequently installed within the TileCal in 2013. The V7.5.0 Bricks are to remain within the TileCal up until the Phase-II Upgrade, which will take place during the Long-Shutdown 3 (LS3).

During LS3, the latest V8.6.0 Bricks will be installed (pending final review). The replacement of the V7.5.0 Bricks is motivated by the high-luminosity environment that will be created by the HL-LHC, which necessitates more-stringent radiation hardness requirements for the active components. The TileCal detector will also take the opportunity to implement further design improvements of the low-voltage power distribution system. These improvements are the introduction of a third stage to the system and the introduction of tristate functionality.

The third stage of the power distribution system takes the form of point-of-load voltage regulators, which undertake the final step-down of the 10 VDC power received from the Bricks to the voltages required by the local front-end electronic circuitry. The introduction of this stage allows for a single Brick variant with a homogeneous output voltage as opposed to the V6.5.4 Brick, which had eight sub-variants, each of which had an output voltage tailored to the specific on-detector electronics that it powered. This resulted in not only increased production costs, but also made the storage of spares more cumbersome.

The newly introduced tristate functionality refers to the capability of the LV power distribution system control the on/off state of an individual Brick utilising a tristate voltage signal, which is routed to the Bricks’ startup circuitry. This is a significant improvement as the on/off control of the legacy LV system was at the granularity of half of an fLVPS. One such implication of the legacy control was that the front-end electronics of a half a super-drawer would need to be power cycled even if only a portion of the drawer was exhibiting errors associated with single-event upsets. Although no negative behaviour was “observed” due to the needless power cycling of the font-end electronics, it is nonetheless undesirable.

As shown in Figure 2, the Phase-II Upgrade Brick, of which there will be a total of 2048 installed within the TileCal, provides a nominal output current of 2.3 A at 10 VDC.

The Bricks’ design is centred on the Analog Devices LT1681 controller chip, which is a pulse-width modulator that operates at a fundamental frequency of 300 kHz. The pulse-width modulation is controlled by two inputs. The first input is received from a slow feedback path that monitors the output voltage with a bandwidth of approximately 1 kHz. This feedback path makes use of an isolation amplifier, ensuring galvanic isolation of the primary and secondary sides. The second input is a fast feedback path that monitors the current through the low-side transistor on the primary side of the Brick. The LT1681 provides an output clock to a Field Effect Transistor (FET) driver, which performs the switching of the two MOSFETs located on the primary side.

The design employs synchronous switching of both MOSFETs, that is both the high-side and low-side MOSFETs turn on and conduct for the duration that the output clock is in the high state, and both are off when the clock is in the low state. When the FETs conduct, the input current at 200 VDC flows through the primary windings of the transformer, which, in turn, transfer energy to the secondary windings. The primary to secondary winding ratio is 14:1, and the transformer has a nominal power loss of 3.95 W. The transformer serves additional purposes in that it provides galvanic isolation of the primary and secondary sides of the Brick and also has additional windings for other auxiliary power supplies on the Brick, required for control and monitoring purposes.

A buck converter is implemented on the secondary side of the Brick. The secondary (output) side also contains an inductor–capacitor stage for the filtering of high-frequency noise, which would otherwise be propagated to the front-end electronics. The V8.6.0 Brick utilises the same inbuilt protection circuitry implemented on previous iterations of the Brick, albeit with different trip points. That is, the design utilises three types of inbuilt protection circuitry, Over-Voltage Protection (OVP), Over-Current Protection (OCP), and Over-Temperature Protection (OTP). These circuits, if activated, initiate a shutdown of the Brick. Their activation depends on preset thresholds, which are discussed in Section 5.

The temperature measurements are taken by two thermistors located at the different positions that are illustrated in Figure 3. Thermistor T2 is adjacent to the primary-side switching MOSFETs of the Brick, whereas Thermistor T3 is located adjacent to the LC buck on the secondary side. The thermistors are located on the underside of the Brick.

3. Quality Assurance Testing of the Phase-II Upgrade fLVPS Brick

Access to the LVPS Bricks is of the order of once per year as the ATLAS detector and is required to be in the open position during the Year-End Technical Stop (YETS). Therefore, any Bricks that experience a permanent failure will remain in the same state until the upcoming YETS. This results in one-eighth (sixth) of a module’s Front-End (FE) electronics, in the case of a single Brick failure, being offline for a commensurate period of time for a long (extended) barrel module, respectively. The off-line FE electronics are entirely unable to collect collision data. Due to this, the reliability of the Bricks is of the utmost importance. A failure is defined as the permanent inability of the fLVPS to provide power to its associated front-end electronics, necessitating replacement.

The reliability of an electronic device, such as a Brick, can differ from its predicted design reliability due to latent and patent defects associated with its production and that of its components. Quality assurance testing is to be undertaken on all fLVPS Bricks post-manufacturing to address this phenomenon [14], the purpose of which is to increase the reliability of the surviving population by screening out the sub-population of Bricks that exhibit such defects. As seen in Figure 4, the quality assurance procedure is composed of five distinct tests, namely an automated visual inspection, X-ray scan, initial testing, Burn-in, and final testing. Each test is tailored to ensure the high reliability of the Brick population, once installed on the detector, by detecting and/or stimulating any patent or latent defects that are present.

Upon completion, every Brick receives a unique identification barcode in order to allow for the ease of tracking during its lifetime in the experiment. The Bricks then undergo an automated visual inspection, whereby visible manufacturing defects are identified and corrected. The Bricks proceed to the next step, whereby they receive an X-ray scan, which focuses on the identification of possible solder outgas voiding that may have occurred between certain components during the reflow process, such as the primary-side switching FETs and the ceramic post to which they are affixed. The motivation for the detection of these voids is due to their propensity to reduced thermal conductivity between the electronic component tab and the ceramic post at the thermal interface.

The initial and final testing are identical. Each utilise the same test apparatus and procedure and are so named due to their occurrence before or after Burn-in, respectively [15]. Both tests measure and evaluate various Brick performance metrics and function to ensure that the Brick under test is operating within its design specifications. Undertaking performance testing both before and after Burn-in provides valuable insight into any performance degradation resulting from the Burn-in procedure while also serving to screen out Bricks that have been permanently damaged. However, Bricks that have latent defects that function marginally may not be detected by performance testing alone and would be installed within the TileCal. These Bricks would fail early in the life of the Phase-II Upgrade TileCal and give rise to a high initial failure rate, commonly observed in the electronic infant mortality region.

System-level Burn-in of all manufactured Bricks is to be undertaken. That is, the Bricks will undergo Burn-in once fully assembled and operational. Burn-in is required as some patent and/or latent defects may not have been detected by the initial testing. Burn-in is primarily focused on detecting patent defects that appear during the early life of the Bricks, but it should be noted that latent defects, which usually appear during normal operation, can be converted into patent defects via the application of external overstress [16]. Burn-in testing entails subjecting the Bricks to a Burn-in procedure in which they are exposed to overstress conditions, such as an increased operating temperature and applied load, in relation to their nominal operating conditions. This functions to stimulate failure mechanisms within the non-performant Brick population. These Bricks need not experience a catastrophic failure during Burn-in, but should be identifiable due to the identification of possible deterioration in their performance during final testing. It is worth emphasising that Burn-in does not improve the reliability of an individual Brick, but rather, the reliability of the surviving brick population as any identifiable non-performant Bricks are removed from the population.

Efforts will be made to repair the Bricks removed from the population. The repair of the Bricks is motivated by the fact that they have a high per-unit cost and, more importantly, is due to the fact that the active components used in their production are essentially one of a kind, as they are independently certified for radiation hardness in production lots. These lots, containing a fixed unit number, are approved for use within the production. The Bricks will be subjected to the quality assurance procedure again if repaired. This process will repeat until the Brick passes the entire quality assurance procedure or is deemed irreparable. The Burn-in apparatus employed is a custom design due to the custom nature of the Bricks that it is tasked with testing. This testing procedure is referred to as the Burn-in procedure in the subsequent paragraphs. The development of the Burn-in station is considered below.

4. Custom Phase-II Upgrade fLVPS Brick Burn-In Station

The Burn-in station, pictured in Figure 5, is an iterative design similar to that of the Bricks and is tasked with applying a Burn-in procedure. The key design requirement for the latest Burn-in station is that it is able to consistently apply the Burn-in procedure to each of the Phase-II Upgrade Bricks throughout the testing of the entire Brick population. Due to the scale of the Brick quality assurance testing, further requirements are imposed as many junior researchers will be involved in the undertaking. That is, it is a necessity that the operation of a Burn-in station be both simple and safe, allowing its use by non-experts.

The Burn-in station is composed of four distinct elements, namely the test-bed, cooling system, hardware, and software. The test-bed contains the majority of the Burn-in station hardware and the Bricks undergoing the Burn-in procedure, as seen in Figure A1. The test-bed is fully enclosed in a metal (main chassis) and perspex (top and front lid) box, reducing the impact of the outside environment on the operating temperature of the Bricks. The test-bed also has a safety role in that it prevents an unwitting user from coming into contact with any electrified elements. It is still foreseen that a safety interlock will be included on the front lid, which will shut-off the High-Voltage (HV) input to the Bricks should the lid be opened during the Burn-in procedure. The test-bed makes use of thermal and electrical insulation to prevent internal water condensation, as well as potential electrical shorts, while also being physically grounded.

An active cooling system is employed to control the operating temperature of the Bricks, a key Burn-in parameter, during Burn-in, as well as to remove heat produced by the dummy loads to which they provide power. The cooling system makes use of a commercial external water chiller, which brings distilled water to the desired setpoint before pumping it to the four Brick Cooling Plates (CPs) and, then, subsequently, to the four dummy load CPs, which are in a single cooling-loop. The implementation of a single cooling-loop allows for a single external water chiller to be used to control the Burn-in temperature, as well as cool the dummy load boards. The water chiller also has a total cooling capacity that allows it to control the temperature of two Burn-in test-beds. The benefit of this design is a reduction in the cost of the Burn-in stations, but this also results in additional complications with the cooling of the dummy load boards. The coolant temperature can be increased or decreased, thereby inducing a commensurate change in the temperature of the Bricks, allowing for the Brick Burn-in operating temperature to be set.

The Burn-in station hardware and software are discussed in detail in Appendix A and Appendix B, with the most important Burn-in performance measurements depicted in

Table 1. Burn-in station performance measurements and their origin. The measurements labels as used in the Burn-in LabVIEW Application (BLA) and shown in the Graphical User Interface (GUI) in Figure A7. The GUI or LabVIEW Front Panel shows the performance measurements of each Brick during a Burn-in procedure.

Origin	Measurement	Label on BLA
Brick	Input voltage	Vin [V]
	Output voltage	Vout [V]
	Input current	Iin [A]
	Output current	Iout [A]
	Temperature 2	T2 [°C]
	Temperature 3	T3 [°C]
Dummy Load Board	Brick voltage at load	Vload [V]
	Brick current at load	Iload [A]
High-Voltage Supply	Input voltage	Vin [V]
	Input current	Iin [A]

. The lack of a Temperature 1 measurement is an artifact of the removal of this measurement, which was used in previous iterations of the Brick. The measurements are taken by thermistors located at the 2nd and 3rd positions. Position 2 is adjacent to the primary-side switching MOSFETs of the Brick, whereas Position 3 is located adjacent to the LC buck on the secondary side.

4.1. Hardware

The Burn-in station hardware is composed of a Personal Computer (PC), a 200 VDC High-Voltage (HV) power supply, custom-designed Printed Circuit Boards (PCBs), electrical wiring, cooling plates, and a test-bed. The PCBs are subdivided into four types, namely the Main Board (MB), the Brick Interface Board (BIB), the Load Interface Board (LIB), and the Dummy Load Board (DLB). The location of the various PCB types within the test-bed is illustrated in Figure 6.

The PCBs are supplied by 220 VAC mains power, which is rectified using DC–DC converters to supply regulated power to the active components. A simplified block diagram of the Burn-in station hardware is provided Figure A1. Based on the legacy designs developed by Argonne National Laboratories (ANL), the Burn-in station PCBs were redesigned by the University of Texas Arlington (UTA), with refinement and testing being undertaken by the University of the Witwatersrand (Wits).

4.2. Software

The Burn-in LabVIEW Application (BLA) and Programmable Integrated Circuit (PIC) firmware were originally developed by Argonne National Laboratory (ANL) in 2006 for the V6 Bricks [12,17]. Based on the legacy designs developed by ANL, the Burn-in station PCBs have been redesigned by the University of Texas Arlington (UTA) and developed by the University of the Witwatersrand (Wits). During the testing and development of the Burn-in station, the BLA underwent major revisions to improve code readability, while only minor changes were made to the HV power supply instrumentation drivers and firmware (ADC channel configurations) of the interface boards. The software required for the operation of a Burn-in station can be divided into three categories, namely the BLA, the PIC firmware, and the high-voltage programmable power supply instrumentation drivers.

5. Phase-II Upgrade fLVPS Brick Preliminary Burn-In Procedure

The Burn-in procedure subjects the Bricks to sub-optimal operating conditions, which function to stimulate failure mechanisms within the Bricks. This procedure is the subject of ongoing research. These conditions needs to fall within the extrema allowed for by the Brick design and operation. There are two reasons for this. The first is related to effective Burn-in, and the second is related to the implementation of the Burn-in procedure. Firstly, the failure mechanisms stimulated must be of the kind that can be experienced during operation within the TileCal such as an increased operating temperature. Secondly, the Burn-in parameters must be restricted to the parameters provided in

Table 2. Preliminary Phase-II Upgrade Brick Burn-in parameters, nominal V8.6.0 Brick operating parameters (over-voltage, over-current, over-temperature), and protection circuitry trip points.

Parameter	Burn-In	Nominal	Protection
Operating temperature	60 °C	35 °C	70 °C (OTP)
Applied load	5.0 A	2.3 A	6.9 A (OCP)
Output voltage	10 V	10 V	11.85 V (OVP)
Run-time	8 h	-	-

due to the Bricks’ inbuilt protection circuitry. A Brick will initiate a trip (immediate shutdown) if the OCP or OTP trip points are reached during Burn-in. If the operating parameters are set to within the variance of the trip points, intermittent Brick trips would occur during the Burn-in procedure, which would not only be operationally taxing, but also bring the validity of the Burn-in results into question, whereas if the parameters are set higher than this variance, the Bricks will start and immediately trip so that Burn-in could not take place.

The run-time of the Burn-in procedure is an important parameter. One needs to weigh the efficacy of the Burn-in process against the practicality of undertaking the Burn-in of a large population of Bricks within a finite time period. The above points combined with previous experience obtained through the Burn-in of the V7.5.0 Bricks were considered, resulting in the Burn-in parameters provided in Table 2 being selected. The 8 h run-time originated from the development of the V6.5.4 Brick and its associated quality assurance procedure by Argonne National Laboratory. The explicit calculations in which this run-time was determined were unfortunately not recorded, with information provided only in the form of internal presentations of the Burn-in procedure. As this Burn-in duration has been successful when applied to two previous Brick versions, it was considered as a good starting point for the preliminary Burn-in procedure. That being said, numerous components have been changed in the subsequent Brick versions, and a naive assumption of the equivalency of their behaviour during Burn-in presents a major risk. Therefore, a quantitative analysis of the Burn-in duration and its efficacy are the subject of ongoing research.

A generalised Phase-II Brick Burn-in procedure thermal profile is illustrated in Figure 7. The $T_2$ temperature value originates from a thermistor located in between the primary-side MOSFETs. This thermistor, which records one of two on-Brick temperatures, was selected as it is utilised in the Brick OTP circuitry in which it measures the highest on-Brick temperatures originating from the MOSFET switching losses. The thermal profile can be sub-divided into three distinct regions. These are known as the rise, soak, and fall regions. At the onset of the Burn-in procedure, the Bricks are switched on and put under a 5 A load. The Bricks and the cooling system are in thermal equilibrium at 20 °C at this point. This lower bound was selected due to it being the average on-detector temperature of the Bricks when they are not in operation. The temperature steadily climbs via the action of the cooling system, increasing the temperature of the heat-sinks. This process is assisted by the Bricks’ self-heating due to inherent power conversion inefficiency. The characteristic curve is a result of the heat capacity of the system and the conductive, as well as radiative losses. The soak region begins after time T_rise and continues for a total time period known as the soak time at the Burn-in temperature of 60 °C. The Burn-in procedure run-time is taken as the sum of the rise and soak time. This is because the Brick must remain under load during the entirety of the Burn-in procedure. At the eight-hour mark, the Burn-in procedure is terminated with the Bricks being switched off and brought down to their starting temperature. The time that it takes for the Bricks to be brought down to their starting temperature is known as the fall time.

6. Commissioning of the fLVPS Brick Burn-In Station

The commissioning of the Burn-in station is partitioned into two distinct commissioning procedures. Preliminary commissioning is undertaken to ensure that the test station is safe for operation and that it is able to apply a Burn-in procedure. Final commissioning focuses on the calibration of the test station.

6.1. Preliminary Commissioning

Preliminary commissioning begins with a visual inspection of all safety-critical hardware. Particular emphasis is placed on the temperature control system and all electrical connections due to the inherent danger posed by their close proximity. This process consists of a review of all coolant pipe joints, coolant levels within the water chiller, electrical and thermal insulation of the cooling plates, grounding of the test-bed, inspection of all cable connections and the PCBs to which they are connected, and ensuring that the mica films located between the dummy load power MOSFET tabs are correctly affixed. It was found that the thermostats located above the dummy load MOSFETs, which act as inbuilt thermal protection circuitry, have a propensity to trip during normal operation. As a result, the thermostats were replaced with an identical package with a trip point of 130 °C, as opposed to the previous thermostats, which had a trip point of 110 °C. The maximum allowable MOSFET junction temperature was taken into account when making this design change.

Thermography was used as a part of the commissioning process due to its ability to provide a macroscopic view of the thermal behaviour of the Burn-in station. The findings of the thermal review were corroborated by the use of an external thermocouple. Key areas such as the dummy load power MOSFET and the primary-side switching MOSFETs of the LVPS Bricks were reviewed with their operating temperatures being measured at 30.2 °C and 62.12 °C, respectively. The cooling plates were found to be at 38 °C (Figure 8, bottom-right) which was 2.0 °C lower than the temperature setpoint of the water chiller. This deviation will be taken into account when applying the Burn-in procedure.

6.2. Final Commissioning

The Burn-in station records and displays numerous behavioural parameters. The purpose of these measurements is to ensure that the Burn-in procedure is being correctly applied while also providing insight into the performance of the Bricks during Burn-in. The raw data originate from two distinct sources, as can be seen in Table 1. That is, the performance data are measured by the Bricks themselves and the dummy load boards to which they provide power. The data produced by these two sources take two types of paths to the PC, which are illustrated in Figure A3. The first path type considered consists of raw analog data measured by a Brick, or dummy load board, which is then converted by the ADC of the associated interface board into a representative serial bit stream. The bit stream is then transmitted to the BLA, where it is scaled, displayed, and logged. The scaling of a Brick’s raw digital data is illustrated in Figure 9. It was assumed that the relationship between the raw analog data and the corresponding digital data is linear within a small deviation from nominal. Therefore, a linear function $f\left(x_j\right)$ was used to perform the scaling of the j-th raw data $x_j$. The scaling of $x_j$ was achieved by the selection of appropriate calibration constants $a_j$ and $b_j$, which correspond to a gain and offset, respectively. The selection of appropriate gain and offset values to ensure equality between the physical values and the recorded and displayed values constitutes the calibration of the Burn-in station.

The calibration constants are unique for each j-th value of a Brick and for each of the eight Bricks. This procedure is followed for $j<4$. For the remaining two measurements, the data need to be converted from a voltage $\left(V_0\right)$ that is representative of the temperature of the $T_2$ or $T_3$ thermistors (Part No. NCP18XH103J03RB, Murata, Kyoto, Japan) into their actual temperature. This takes the form of two additional steps. The first step uses the scaled voltage to determine the thermistors’ resistance $R_j$. This resistance is then used to calculate the thermistors’ temperature. The relationship between the voltage received from the voltage divider and the thermistors’ operating temperature is illustrated in Figure 10 left. The thermistor B value of 3428 k was chosen with the thermistor calibration curve for this value illustrated in Figure 10, right. The dummy load raw data undergo the same process of scaling and associated calibration as the current and voltage measurements of the Bricks for each of the eight dummy load channels (1 channel per Brick).

7. Summary and Outlook

A Burn-in apparatus, referred to as the Burn-in station in this paper, for the ATLAS Tile Calorimeter Phase-II Upgrade transformer-coupled buck converters has been developed and manufactured. The motivation for and application of Burn-in testing were provided within the context of the improved reliability of the fLVPS Bricks. The history and function of the LVPS Bricks were furnished with the emphasis being placed on their inbuilt protection circuitry, which sets limits on the possible Burn-in parameters. The four distinct elements of a Burn-in station, namely the test-bed, cooling system, hardware, and software, were introduced with detailed information provided in the Appendix. The Phase-II Brick Burn-in procedure was provided in which the Burn-in parameters were stated and motivated. The generalised Phase-II Brick Burn-in procedure thermal profile was also presented and described with the paper, culminating in the explanation and presentation of the initial and final commissioning of the Burn-in station.

To further assess if the final commissioning of the Wits Burn-in station was successful, an entire Burn-in procedure was applied to a V8.4.2 Brick. A special V8.4.2 Brick, with the “new” higher-efficiency MOSFETs and L4 inductor as updated in the V8.5.0 Brick, was used. This was performed so that the V8.4.2 Brick provided an approximate analog to the V8.5.0 Brick in terms of thermal characteristics and efficiency. This was necessitated as the V8.5.0 Brick shown in Figure 3 (and five others like it) had not been produced at the time of testing. The low sample size (one Brick) of the V8.4.2 Brick was due to the majority of the South African Bricks being sent directly to CERN for use in the TileCal Phase-II Upgrade test beam campaigns. In order to increase the sample size and corroborate the current findings, the measurements presented below will be repeated with the six V8.5.0 Bricks.

The measurements of the Bricks’ T2 and T3 temperature, as well as their efficiency during the Burn-in procedure, provide a probe into not only the performance of the Bricks, but also that of the Burn-in station. The T2 and T3 thermal profiles depicting the Burn-in of a single V8.4.2 Brick are illustrated in Figure 11. Three key observations can be made when comparing these thermal profiles to the idealised thermal profile illustrated in Figure 7. The first observation is that there is a difference of 10 °C between the T2 and T3 thermistor values, which was expected. This difference is due to their location, illustrated in Figure 3, being adjacent to the switching MOSFETs and LC-buck, respectively. These components output a different amount of heat during operation. The second observation that can be made is that the starting temperature is not 20 °C, as motivated previously in Section 5. This is a consequence of the way in which the Burn-in temperature control is currently being implemented. Due to the Bricks not being 100% efficient, “self-heating” produces a $\Delta T$ between the thermostat values and cooling plate of approximately 20 °C. Therefore, to achieve the Burn-in temperature of 60 °C, the cooling plates need to be maintained at 40 °C. The thermal profile can be brought back into the specifications by either starting the cooling station at the commencement of the Burn-in procedure or programming the cooling station to ramp its temperature. The former ad hoc option is to be implemented, but development is anticipated for the latter. Finally, the trailing edge of the thermal profile is not visible. This is due to the data acquisition being terminated at the end of the Burn-in procedure. Although the thermal behaviour is well understood and should subsequently decrease to 40 °C, this will be addressed.

The efficiency of the V8.4.2 Brick during Burn-in is illustrated in Figure 12 (left). The high efficiency observed at the start of the Burn-in procedure, which sharply decreases, can be attributed to the RC time constant associated with the input current measurement circuitry and is, therefore, not indicative of the Brick’s actual efficiency at startup. The input current has an inverse relationship with the efficiency with its behaviour at Brick startup seen in Figure 12 (right). A linear increased Brick efficiency, which also appears to be attributed to the input current measurement, can be observed. This behaviour is currently under investigation as a “steady-state” efficiency of approximately 70–75% is expected. This investigation will take the form of an external measurement of the Bricks’ input current, as well as the application of a longer-duration Burn-in procedure to investigate whether or not the current stabilises.

A University of the Witwatersrand Burn-in station has been developed, manufactured, and commissioned. The station is poised for use within the pre-production of approximately 100 units, in the fourth quarter of 2023, at the University of the Witwatersrand (Wits). This will serve as a preparatory step for the full-production and quality assurance testing of 1024 units in 2024 at this institution, which is well within the current Phase-II Upgrade schedule. The preproduction will also allow for the refinement of the Burn-in procedure as a substantially larger dataset will be produced. A second identical Burn-in station is currently being produced to reduce the anticipated bottleneck at the Burn-in stage of the quality assurance procedure at Wits. The same procedure is being mirrored in parallel at the University of Texas at Arlington (UTA), which will be producing the same amount of Bricks and undertaking identical quality assurance testing, with identical test stations, locally. Studies into the Burn-in duration, the observed Brick efficiency increase over time, and the refinement of the Burn-in station data acquisition are the subject of current work.