Design and Fabrication of the 1.9 K Magnet Test Fa- Cility at BNL, and Test of the First 4 M-Long MQXF Coil

Home , Cryostat

Mon-Af-Po1.01-05 1 Design and Fabrication of the 1.9 K Magnet Test Fa- cility at BNL, and Test of the First 4 m-Long MQXF Coil

J. Muratore, M. Anerella, P. Joshi, P. Kovach, A. Marone, P. Wanderer

Abstract— The future high luminosity upgrade of the Large Had- vertical superconducting magnet test facility of the SMD at ron Collider (LHC) at CERN will include twenty 4.2 m-long BNL has been upgraded to perform testing in superfluid He at Nb3Sn high gradient quadrupole magnets which will be compo- 1.9 K and 1 bar, the operational condition at the LHC. This nents of the triplets for two LHC insertion regions. In order to test these and four pre-production models, the vertical supercon- has involved extensive modifications of the 40-year old, 4.5 K ducting magnet test facility of the Superconducting Magnet Divi- cryogenics plant and vertical test facility at the SMD. These sion (SMD) at Brookhaven National Laboratory (BNL) has been upgrades include new piping, compressors, and other critical upgraded to perform testing in superfluid He at 1.9 K and 1 bar, components; a new vertical test cryostat with a top plate and the operational condition at the LHC. This has involved extensive hanging fixture which can accept larger diameter and longer modification of the 4.5 K cryogenics plant, including piping, magnets, up to an actual length of 5 m; a 1.9 K heat exchang- compressors, and other upgraded components; a new vertical test cryostat which can accept larger diameter magnets; a modern- er; a newly designed warm bore tube; and a lambda plate de- ized power supply system upgraded with IGBT switches and fast signed with a novel sealing scheme to provide for both shutoff capability, and that can supply 24 kA to test high field strength and minimum heat loss. In addition, the former short Nb3Sn magnets; and completely new data acquisition, signal sample cable test facility 30 kA power supply has been up- analysis, and control software and hardware, allowing for fast, graded with an energy extraction system using IGBT switches high precision, large volume data collection. This paper reports and fast shutoff capability, and new high energy dump resis- on the design, assembly, and commissioning of this upgraded test tors, and has been reconfigured to supply 24 kA to test high facility, and presents results of the first magnet test performed. field Nb3Sn magnets. Index Terms—accelerator magnet, superconducting coils, We have also assembled completely new data acquisition, quench protection, test facilities signal analysis, and control software and hardware, allowing for faster, higher precision, and larger volume data collection. This paper describes the design, assembly, and commis- I. INTRODUCTION sioning of the upgraded test facility, and reports on the first he future high luminosity (HiLumi) upgrade of LHC at magnet test performed, on a mirror model, which consists of a single coil quadrant and an iron yoke which fills the space of T CERN will include twenty 4.2 m-long Nb3Sn high gradient quadrupole magnets which will be components of the Q1 the other three quadrants. The test facility and its upgrades can and Q3 triplets for two insertion regions of the LHC. These be discussed in four general areas: 1) cryogenics facility, magnets, denoted as MQXFA, will be supplied by the US Ac- 2) vertical test cryostat and magnet fixture, 3) power supply celerator Upgrade Project (AUP), a collaboration of BNL, system with quench protection, and 4) data acquisition, con- Fermi National Accelerator Laboratory, and Lawrence Berke- trol, and analysis hardware and software. ley National Laboratory. Each magnet will have four two- layer coils wound with 40-strand Nb3Sn Rutherford cable in TABLE I order to generate the high field gradients necessary for the REQUIRED OPERATIONAL PARAMETERS FOR MQXFA TESTS LHC HiLumi upgrade. Table I shows required operational pa- Coil inner aperture D = 150 mm rameters for these magnets, which include operating at cur- Coil magnetic length l = 4.2 m rents up to 17.89 kA, known as the ultimate current. More in- Coil actual length l = 4.523 m formation about the MQXFA magnets and the test require- Total magnet length l = 5 m (nom) ments can be found in [1], [2]. Operational temperature T = 1.9 K In order to test these and four pre-production models, the LHC nominal operating current (1.9 K) Inom = 16.470 kA

LHC ultimate operating current (1.9 K) I = 17.890 kA This work was supported by the U.S. Department of Energy, Office of Sci- ult ence, Office of High Energy Physics, through the US LHC Accelerator Re- Conductor limit at 1.9 K Iss = 21.600 kA search Program and by the High Luminosity LHC project at CERN. The U.S. Conductor limit at 4.5 K Iss = 19.550 kA Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains, a non-exclusive, paid-up, ir- Magnet inductance (at 1.9 and 1 kA) L1 = 40.9 mH revocable, world-wide license to publish or reproduce the published form of Magnet inductance (at 1.9 and Inom=16.5 kA) L16.5 = 32.8 mH this manuscript, or allow others to do so, for U.S. Government purposes. Operating stored energy (at B , I ) E = 4.5 MJ The authors are with the Superconducting Magnet Division, Brookhaven nom nom max National Laboratory, Upton, NY 11973 USA (e-mail: [email protected]). Dump resistor (energy extraction) options RD = 30, 37.5, 50, 75, 150 mΩ

II. CRYOGENICS FACILITY 0.0837 bar and it delivers 2.7 g/s at 1.9 K. Vapor pressure in The cryogenics facility with its large infrastructure and the heat exchanger is 16 mbar. Cooling capacity is 40 W. Out- multiple vertical test cryostats at the SMD has been used since dated control and diagnostic programming has been replaced the 1970’s to test superconducting magnets at 4.5 K (nominal) with more efficient LabVIEW software. As with the other and up to 10 kA current for magnets mostly wound with NbTi components of the cryogenics system, the vacuum pump has conductor, and it has supported such projects as SSC, RHIC, been refurbished with many new parts and critical spares have and the present LHC insertion region dipole magnets. For use been purchased. in the present HiLumi project, the facility had to be modern- Warm Helium Return ized and upgraded to test at 1.9 K and 1 bar, and up to 24 kA 4.5 K Helium Input current for the state-of-the-art Nb3Sn MQXFA quadrupoles.

BNL management funded the purchase of new parts and criti- Liquid Helium 4.5 K Storage Dewar 16 mbar 10000 L cal spares, preventive maintenance of critical components, and 1 bar installation of backup systems to improve reliability and miti- Nash 1.9 K Vacuum 1.9 K Buffer Tank Station to Helium Gas gation of risks inherent in a 40-year old facility. Fig. 1 shows a Storage flow chart of the cryogenics facility and its components. 2-Stage CVI/Magcool Purifier A & B

1.9 K Test Cryostat High Pressure Helium Gas High Pressure Mycom 800 hp Sullair 500 hp Sullair 350 hp Sullair 100 (75 kW) Helium Gas Compressor Compressor Compressor @ 8K (2 Stages) (Single Stage) (Single Stage) 160 g/s 80 g/s 51 g/s Buffer Tank Reciprocating to Suction CS4 Helium Expander CTI Model 4000 Buffer High Pressure Helium 1500 W Tank CS4 Refrigerator/ CVI – Magcool Gas Storage Liquifier Precooler/Subcooler 8.5 x 106 L @ 15.5 bar (Cold Box) Fig. 2. 1.9 K Process Flow Chart for Cryogenics Facility at SMD Liquid Helium

Helium Wet Expander Koch Model 1600 Warm Return The SMD has five vertical test cryostats, of which the Warm Return Liquid 6.1 m-deep Test Cryostat 2 was modified by inserting a new, Helium

Turbo-Expander 4.5 K Lambda redesigned inner He vessel, with 4.5 K heat shield, and which High Pressure Helium Dual Plate Helium Gas Liquifier Turbines 1 bar CVI/Air Liquide Cryenco is rated at 5.86 bar, into the existing outer dewar and which 1000 W Cryenco Gardner 10000 L 1.9 K 3785 L 3785 L Storage Dewar extended the useful length by 200 mm, in order to accommo- Storage Storage Dewar Dewar Cryenco date wider and longer magnets, up to 5 m long, and can there- Inline Purifier fore accept the MQXFA quadrupole magnets, which are close 1.9 K Test Cryostat 2 Fig. 1. SMD Cryogenics Facility Main Process Flow Chart to 5 m. In addition, a new top plate assembly (76.2 mm-thick 304 The main source of liquid helium is a 1500 W CTI Model stainless steel) has also been built with enough connector trees 4000 Refrigerator/Liquefier with two reciprocating expansion to accommodate the large amount of MQXFA instrumenta- engines rated at 250 rpm. With both expansion engines and a tion, including voltage taps, quench protection heaters, strain Koch Model 1600 Wet Expander, rated at 60 rpm, running, the gauges, temperature sensors, and level probes. The top plate liquefaction capability is 320 L/hr, with up to 17570 L storage assembly also includes a G10-stainless steel bilaminar lambda capacity. Refrigerator upgrades included a new liquid nitrogen plate, 24 kA vapor-cooled copper leads, two 1 kA NbTi leads heat exchanger, doubling the number of inline purifiers to for the CLIQ (coupling loss induced quench) protection sys- four, rebuilding of expansion engines and wet expander, and tem, a 1.9 K heat exchanger, and multiple liquid helium fill new diagnostic and control software written in LabVIEW. lines. Fig. 3 shows a rendering of the top plate. The lambda The main He compressor is a 597 kW 2-Stage My- plate is below and not shown but is discussed in detail in [3]. comm 800 which, along with a 260 kW Sullair 350 compres- The lambda plate is composed of a 38.1 mm-thick G10 sor, can supply up to 210 g/s flow. These compressors have plate bonded with Stycast 2850FT on top of a 19.05 mm-thick been refurbished with many new parts and the Mycomm has 304 stainless steel plate; the G10/stainless double layer is a been equipped with a new “soft start” system to replace the compromise for lowering heat load and increasing strength, old system of six mechanical contacts and which eliminates and getting a more reliable seal to the inner He vessel flange. the unreliability of that old contactor start system. The start is Total heat load of the lambda plate has been calculated to be now done with reliable electronics and LabVIEW program- less than 1.4 W. ming to slow the ramp up to 480 V. In addition, a 373 kW Sul- The seal of the lambda plate with the He vessel flange is a lair 500 compressor has been provided as a backup in case of spring-energized face seal consisting of a cantilever spring loss of the Mycomm during a test. Not shown in the picture is made of Elgiloy (Co-Cr-Ni alloy) in a Teflon jacket and which a “dirty gas” recovery system which has been refurbished with 6 compresses from the load of the lambda plate and hanging new hardware. Total gas storage is 8.5 x 10 L. magnet by gravity. The load required to completely compress Fig. 2 shows a flow chart of the new 1.9 K cooldown sys- the seal is approximately 105 N/cm. The seal for this lambda tem. The vacuum pump is a 150 kW 2-stage Nash-Kinema, plate is designed to reach full compression and optimal sealing which pumps on the liquid He in a heat exchanger under the with 23 kN, which is more than provided by the top plate and lambda plate in the test cryostat. Suction pressure is hanging magnet. The seal was designed to work such that all

3 gaps can be handled if the seal is not fully compressed or the 15 kA water-cooled cable 15 kA / 14 V 100 µH mating parts not completely flat. The seals for the warm bore 85 mΩ Power Supply 15 kA DCCT 6 Pulse SCR 15.5 mF 62.0 mF tube and the 1.9 K heat exchanger are similarly spring- 3 Phase Bridge 480 V 15 kA water-cooled cable energized but in a radial configuration where the forces are X 6 applied radially by the insertion of the part rather than by IGBT (6 parallel units) Switch 24 kA water cooled cable gravity as is the case with the face seal. More details of the with snubber circuits upgraded cryogenics facility and vertical test cryostat, includ- 24 kA water cooled cable 800 mF ing lambda plate details, are shown in [3]. 33 mΩ 33 mΩ 1 Ω 15 kA water-cooled cable 15 kA / 14 V 100 µH 85 mΩ Power Supply 15 kA DCCT 24 kA Vapor-Cooled Leads Quench Heater Connector Tree 6 Pulse SCR 16.0 mF Heat 81.9 mF Bridge exchanger 15 kA water-cooled cable 4.5 K

X 6 1.9 K

IGBT (6 parallel units) Switch with snubber circuits

800 mF 33 mΩ 33 mΩ 1 Ω Fig. 4. 30 kA Power Supply System with Energy Extraction

Another source of quench protection is the CERN-supplied 24 kA vapor-cooled leads CLIQ system, details of which can be found in [5], [6].This system, which will be used in the LHC instead of EE, supplies Top plate 76.2 mm thick 304 SS a damped RC oscillation with maximum amplitude 500 V, 3rd CLIQ Lead which is triggered by a quench detector trip, along with the (if needed) QPH system, the power supply shutoff, and the fast data log- He fill lines ger. The AC loss heating generated by the CLIQ signal creates Connector Trees CLIQ Leads added quench volume in the coils and therefore lowers the maximum quench temperature by spreading the energy. In or-

Fig. 3. Top plate and header system designed for MQXFA magnets. der to verify CLIQ performance in the MQXFA testing, its connection to the magnet and the power supply system must be configured as shown in Fig. 5. The string of diodes shown III. POWER SUPPLY AND QUENCH PROTECTION is necessary to keep current from back-flowing during ramp- For the main power supply to power MQXFA magnets, the ing of the magnet during the first few thousand amperes. Also previous 30 kA short sample cable facility power supply, the configuration shown allows for the triggering of both which consists of two 15 kA power supplies in parallel, has CLIQ and EE independently when the QD trips. Therefore de- been updated and modified to supply 24 kA to an inductive lays can be utilized in both systems to add versatility during load. New electronics and feedback loop and PID tuning have quench protection studies when testing the pre-production been added to handle magnet-sized inductances, and filtering MQXFA models. was re-configured to minimize the ripple. A schematic detail- Filter DCCT ing the re-designed power supply can be seen in Fig. 4. It is al- 24 kA Cable (60 ft) 30 VDC S1, 6 IGBT 1000 A Shunt 15 kA Switches In so showing that a fast energy extraction (EE) system (for D2 Parallel GAS COOLED LEAD Power

Supply D3 quench protection) has been added by installing six IGBT CLIQ Circuit CERN Built L1 Vo switches and a 150 mΩ ceramic, non-inductive dump resistor REE1, Energy for each power supply. Each dump resistor is rated to dissipate Extraction Tc L2 Resistor up to 5 MJ of energy during a quench. Values of dump re- 20 DIODES sistance can be varied over a range (0, 30, 37.5, 50, 75, and 15 kA Cable (85 ft) C L3

150 mΩ). The switches are fast as the IGBT turnoff time is Filter DCCT Current L4 Sensor about1-2 μs. In addition, each IGBT is equipped with a spe- 30 VDC S2, 6 IGBT 15 kA Switches In D2 Parallel GAS COOLED LEAD cially designed snubber circuit to limit the maximum switch- Power Supply Magnet & D3 Cryostat ing transient collector-emitter voltages to 800 V to avoid dam- 24 kA Cable (60 ft) REE2, High Voltage age, and though each IGBT is rated at 3.6 kA, they are being CLIQ Switch Power Supply IGBT Control Power Supply Control Energy Control Control limited to 2.0 kA for an added margin of protection. All IGBT Extraction Resistor BNL Built Quench Detector BNL Built Control For CLIQ devices are continuously being monitored during testing for Circuit such critical parameters as temperature, current sharing, and P. Joshi. BNL July 18 2017 Fig. 5. 30 kA Power Supply System with CLIQ collector-emitter voltages. At a trip of the quench detector (QD), a fast data log of these parameters is generated to be an- The MQXFA quench detection system is based on a redun- alyzed for proper operation. Further details of this system can dant scheme which relies on a number of voltage signals. be found in [4]. When any one of the signals reaches its specified threshold

4 voltage after a specified validation time, a fast discharge is in- will have only those taps needed for quench detection, as dis- stigated (current decays through dump resistor). These signals cussed earlier. include: the half coil difference, whole coil, quarter coil dif- As can be seen by the quench training results in Fig. 6 ferences, whole coil minus the calculated ramp induced volt- for the MQXFPM1 magnet, there were 19 spontaneous age and SC lead voltages. See [3] for a detailed schematic and quenches. The magnet reached the nominal current in four a table of nominal threshold voltages and validation times. quenches and the ultimate current by the tenth, showing that There are also designed into the system a number of inter- the long coil at least equaled the results of the short mirror test locks which also rely on various signals which, if they do not [7]. Out of the 19 quenches, 14 were in inner layer pole turn meet specified conditions such as voltage threshold or on-off straight sections, which have the highest local field values. state, will instigate a slow discharge, with the dump resistor Only two quenches were in the outer layer, and these were the out of the circuit. These include: vapor-cooled lead voltages, first quench and the third (first after thermal cycle). The last QPH capacitor bank not charged, IGBT switch temperature or quench which was the highest at 19.241 kA, 7.6 % higher than voltage too high, and many other monitored critical parameters. In addition, in Nb Sn magnets experience voltage spikes the ultimate, and was at 4.4 K, not superfluid; the previous 3 quench was at 300 A/s and was similar to the 20 A/s quenches. due to flux jumps, so current-dependent voltage thresholds as The straight section voltage signals gave evidence that the lo- high as several volts at low currents are necessary in order to cation of the pole turn quenches were not the same. These avoid false QD trips. The variation of thresholds with current facts imply that the magnet was still training at the end of test- is set in the programming and so is automatic during ramps. ing.

23000 IV. DATA ACQUISITION 22000 SS 22095A 300 A/s 3.265 - 4.424K 21000 NO QUENCH NO QUENCH The older data acquisition (DAQ) system has been replaced 20000 by new National Instruments (NI)/LabVIEW hardware and 19000 software in order to provide the required faster and higher pre- 18000 ULT 17890A

CURRENT (A) CURRENT 17000 cision data collection. The main DAQ systems include a fast NOM 16480A 16000 A3-A4 Pole Multiturn (at 3.25-4.3K) data logger, slow data logger/monitor, strain gauge DAQ, and 15000 A5-A6 Pole Turn Long Straight Section QUENCH 144K (TOP) - 110K (BOT) A6-A7 Pole Turn NonLead End Section 14000 A7-A8 Pole Turn Short Straight Section QD DAQ. At this time, the fast logger has 128 differential 4.424K A8-B8 Ramp 13000 B3-B2 MIDPLANE MULTITURN channels with 16-bit ADC resolution and simultaneous sam- 12000 THERMAL CYCLES 117K (TOP) - 77K (BOT) Quenches 1-12 and 14-17 were at 20A/s. NO QUENCH-50A/s to 10kA then 20A/s 11000 Quench 3 was in A6-A7 but close to or at tap A7. NO QUENCH-50A/s pling of up to 250 kHz and is used mainly for voltage taps. Quench 13 was at 20A/s with 2hr hold at 17890A. 10000 NO QUENCH-20 A/s Slow logger has 64 18-bit channels for up to 500 kHz for 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 monitoring of vapor-cooled lead voltages, splice voltages, SC QUENCH NUMBER Fig. 6. MQXFPM1 Quench Training History lead voltages, temperatures, and other instrumentation. QD DAQ comprises 8 16-bit channels with up to 1.25 MHz sam- pling of signals used for quench detection, such as those listed VI. CONCLUSION above. Strain gauge DAQ has 60 22-bit channels with scan- ning of 30 channels each on two Agilent 34970A Data Acqui- As can be seen by the results for the MQXFPM1 magnet sition/Switching Units with 250 channels/s scan rate. The commissioning test, the upgraded facility as described in this number of DAQ channels can be expanded if necessary. report successfully fulfilled the acceptance test requirements for the MQXFA quadrupole magnets which are being built and supplied to the LHC for the insertion region Q1/Q3 for the V. TEST FACILITY COMMISSIONING AND TEST RESULTS high luminosity upgrade. These requirements include opera- The upgraded test facility was commissioned in late 2016 and tion at 1.9 K and 1 bar, powering up to 19.241 kA, and the early 2017 at 1.9 K. The first magnet tested was MQXFPM1, proper protection of the magnet during quench tests. For the a mirror magnet with the first long coil of MQXFA cross- test of the MQXFPM1, quench protection included a newly section. Up to that time, only short models had been tested, in- designed and faster quench detection system, energy extrac- cluding a short mirror MQXFSM1 [7]. In addition to a cam- tion with faster and state-of the-art electronic switching with paign of spontaneous training quenches, mostly at 20 A/s, to specially designed IGBT circuitry and infrastructure, and fu- verify the electric and mechanical stability of the long coil, ture magnets will be using the already upgraded energy extrac- there was also quench protection heater studies to determine tion system with a higher power and more versatile dump re- minimum quench energy and quench delay for the three sistors, and twelve new quench protection heater firing units, quench heater circuits, each having two strips on the inner lay- which meet the requirements for quench protection heater sys- er, the outer layer high field region near pole, and the outer tems in the LHC. layer low field region near mid-plane [8]. Future MQXFA quadrupole magnet tests will also include magnetic field ACKNOWLEDGMENT measurements, flux jump spike detection, ramp rate studies, and a quench antenna system to help locate quenches in re- The authors would like to thank William McKeon and Sebas- gions where voltage taps do not exist. The production models tian Dimiauta for their hard work and expertise.

REFERENCES [1] R. Carcagno, G. Ambrosio, G. Apollinari, “MQXFA Functional Requirements Specification”, US-HiLumi-doc-36, to be published. [2] R. Carcagno, G. Ambrosio, G. Apollinari, “MQXFA Magnets Functional Test Requirements”, US-HiLumi-doc-137, to be published. [3] J. Muratore, “BNL Magnet Test Facility”, presented at the Superconducting Magnets Test Stands Workshop, CERN, 13-14 June, 2016. [4] P. Joshi, “24 kA DC Energy Extraction Switch for LARP Magnet Testing at BNL’, IEEE Trans. Appl. Supercond., Po4.09-12, this conference, submitted for publication. [5] E. Ravaioli, “CLIQ,” Ph.D. dissertation, Enschede, 2015, presented on 19 June 2015. [6] E. Ravaioli, et al., “New, Coupling Loss Induced, Quench protection system for superconducting accelerator magnets,” IEEE Trans. Appl. Supercond., vol. 24, no. 3, pp. 1–5, June 2014. [7] G. Chlachidze, S. Stoynev, et al “LARP MQXFSM1 (Mirror) Magnet Test Summary”, FNAL Technical Division Note TD-15-018, 28 Aug 2015. [8] E. Ravaioli, et al., “Quench protection performance measurements in the first MQXF magnet models,” IEEE Trans. Appl. Supercond., Or24-01, this conference, submitted for publication.