T single coil quadrant and an iron yoke which fills the space of - - PDF document

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- ron Collider (LHC) at CERN will include twenty 4.2 m-long Nb3Sn high gradient quadrupole magnets which will be compo- nents of the triplets for two LHC insertion regions. In order to test these and four pre-production models, the vertical supercon- ducting magnet test facility of the Superconducting Magnet Divi- sion (SMD) at Brookhaven National Laboratory (BNL) has been upgraded to perform testing in superfluid He at 1.9 K and 1 bar, the operational condition at the LHC. This has involved extensive modification of the 4.5 K cryogenics plant, including piping, compressors, and other upgraded components; a new vertical test cryostat which can accept larger diameter magnets; a modern- ized power supply system upgraded with IGBT switches and fast shutoff capability, and that can supply 24 kA to test high field Nb3Sn magnets; and completely new data acquisition, signal analysis, and control software and hardware, allowing for fast, high precision, large volume data collection. This paper reports

• n the design, assembly, and commissioning of this upgraded test

facility, and presents results of the first magnet test performed. Index Terms—accelerator magnet, superconducting coils, quench protection, test facilities

• I. INTRODUCTION

he future high luminosity (HiLumi) upgrade of LHC at CERN will include twenty 4.2 m-long Nb3Sn high gradi- ent quadrupole magnets which will be components of the Q1 and Q3 triplets for two insertion regions of the LHC. These magnets, denoted as MQXFA, will be supplied by the US Ac- celerator Upgrade Project (AUP), a collaboration of BNL, Fermi National Accelerator Laboratory, and Lawrence Berke- ley National Laboratory. Each magnet will have four two- layer coils wound with 40-strand Nb3Sn Rutherford cable in

• rder to generate the high field gradients necessary for the

LHC HiLumi upgrade. Table I shows required operational pa- rameters for these magnets, which include operating at cur- rents up to 17.89 kA, known as the ultimate current. More in- formation about the MQXFA magnets and the test require- ments can be found in [1], [2]. In order to test these and four pre-production models, the

This work was supported by the U.S. Department of Energy, Office of Sci- ence, Office of High Energy Physics, through the US LHC Accelerator Re- search Program and by the High Luminosity LHC project at CERN. The U.S. Government retains and the publisher, by accepting the article for publication, acknowledges that the U.S. Government retains, a non-exclusive, paid-up, ir- revocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for U.S. Government purposes. The authors are with the Superconducting Magnet Division, Brookhaven National Laboratory, Upton, NY 11973 USA (e-mail: Muratore@bnl.gov).

vertical superconducting magnet test facility of the SMD at BNL has been upgraded to perform testing in superfluid He at 1.9 K and 1 bar, the operational condition at the LHC. This has involved extensive modifications of the 40-year old, 4.5 K cryogenics plant and vertical test facility at the SMD. These upgrades include new piping, compressors, and other critical components; a new vertical test cryostat with a top plate and hanging fixture which can accept larger diameter and longer magnets, up to an actual length of 5 m; a 1.9 K heat exchang- er; a newly designed warm bore tube; and a lambda plate de- signed with a novel sealing scheme to provide for both strength and minimum heat loss. In addition, the former short sample cable test facility 30 kA power supply has been up- graded with an energy extraction system using IGBT switches and fast shutoff capability, and new high energy dump resis- tors, and has been reconfigured to supply 24 kA to test high field Nb3Sn magnets. We have also assembled completely new data acquisition, 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- sioning of the upgraded test facility, and reports on the first magnet test performed, on a mirror model, which consists of a single coil quadrant and an iron yoke which fills the space of the other three quadrants. The test facility and its upgrades can be discussed in four general areas: 1) cryogenics facility, 2) vertical test cryostat and magnet fixture, 3) power supply system with quench protection, and 4) data acquisition, con- trol, and analysis hardware and software.

TABLE I REQUIRED OPERATIONAL PARAMETERS FOR MQXFA TESTS Coil inner aperture D = 150 mm Coil magnetic length l = 4.2 m Coil actual length l = 4.523 m Total magnet length l = 5 m (nom) Operational temperature T = 1.9 K LHC nominal operating current (1.9 K) Inom = 16.470 kA LHC ultimate operating current (1.9 K) Iult = 17.890 kA Conductor limit at 1.9 K Iss = 21.600 kA Conductor limit at 4.5 K Iss = 19.550 kA Magnet inductance (at 1.9 and 1 kA) L1 = 40.9 mH Magnet inductance (at 1.9 and Inom=16.5 kA) L16.5 = 32.8 mH Operating stored energy (at Bnom, Inom) Emax = 4.5 MJ Dump resistor (energy extraction) options RD = 30, 37.5, 50, 75, 150 mΩ

T

2

• II. CRYOGENICS FACILITY

The cryogenics facility with its large infrastructure and multiple vertical test cryostats at the SMD has been used since the 1970’s to test superconducting magnets at 4.5 K (nominal) and up to 10 kA current for magnets mostly wound with NbTi conductor, and it has supported such projects as SSC, RHIC, and the present LHC insertion region dipole magnets. For use in the present HiLumi project, the facility had to be modern- ized and upgraded to test at 1.9 K and 1 bar, and up to 24 kA current for the state-of-the-art Nb3Sn MQXFA quadrupoles. BNL management funded the purchase of new parts and criti- cal spares, preventive maintenance of critical components, and installation of backup systems to improve reliability and miti- gation of risks inherent in a 40-year old facility. Fig. 1 shows a flow chart of the cryogenics facility and its components.

• Fig. 1. SMD Cryogenics Facility Main Process Flow Chart

The main source of liquid helium is a 1500 W CTI Model 4000 Refrigerator/Liquefier with two reciprocating expansion engines rated at 250 rpm. With both expansion engines and a Koch Model 1600 Wet Expander, rated at 60 rpm, running, the liquefaction capability is 320 L/hr, with up to 17570 L storage

• capacity. Refrigerator upgrades included a new liquid nitrogen

heat exchanger, doubling the number of inline purifiers to four, rebuilding of expansion engines and wet expander, and new diagnostic and control software written in LabVIEW. The main He compressor is a 597 kW 2-Stage My- comm 800 which, along with a 260 kW Sullair 350 compres- sor, can supply up to 210 g/s flow. These compressors have been refurbished with many new parts and the Mycomm has been equipped with a new “soft start” system to replace the

• ld system of six mechanical contacts and which eliminates

the unreliability of that old contactor start system. The start is now done with reliable electronics and LabVIEW program- ming to slow the ramp up to 480 V. In addition, a 373 kW Sul- lair 500 compressor has been provided as a backup in case of loss of the Mycomm during a test. Not shown in the picture is a “dirty gas” recovery system which has been refurbished with new hardware. Total gas storage is 8.5 x 106 L.

• Fig. 2 shows a flow chart of the new 1.9 K cooldown sys-
• tem. The vacuum pump is a 150 kW 2-stage Nash-Kinema,

which pumps on the liquid He in a heat exchanger under the lambda plate in the test cryostat. Suction pressure is 0.0837 bar and it delivers 2.7 g/s at 1.9 K. Vapor pressure in the heat exchanger is 16 mbar. Cooling capacity is 40 W. Out- dated control and diagnostic programming has been replaced with more efficient LabVIEW software. As with the other components of the cryogenics system, the vacuum pump has been refurbished with many new parts and critical spares have been purchased.

• Fig. 2. 1.9 K Process Flow Chart for Cryogenics Facility at SMD

The SMD has five vertical test cryostats, of which the 6.1 m-deep Test Cryostat 2 was modified by inserting a new, redesigned inner He vessel, with 4.5 K heat shield, and which is rated at 5.86 bar, into the existing outer dewar and which extended the useful length by 200 mm, in order to accommo- date wider and longer magnets, up to 5 m long, and can there- fore accept the MQXFA quadrupole magnets, which are close to 5 m. In addition, a new top plate assembly (76.2 mm-thick 304 stainless steel) has also been built with enough connector trees to accommodate the large amount of MQXFA instrumenta- tion, including voltage taps, quench protection heaters, strain gauges, temperature sensors, and level probes. The top plate assembly also includes a G10-stainless steel bilaminar lambda plate, 24 kA vapor-cooled copper leads, two 1 kA NbTi leads for the CLIQ (coupling loss induced quench) protection sys- tem, a 1.9 K heat exchanger, and multiple liquid helium fill

• lines. Fig. 3 shows a rendering of the top plate. The lambda

plate is below and not shown but is discussed in detail in [3]. The lambda plate is composed of a 38.1 mm-thick G10 plate bonded with Stycast 2850FT on top of a 19.05 mm-thick 304 stainless steel plate; the G10/stainless double layer is a compromise for lowering heat load and increasing strength, and getting a more reliable seal to the inner He vessel flange. Total heat load of the lambda plate has been calculated to be less than 1.4 W. The seal of the lambda plate with the He vessel flange is a spring-energized face seal consisting of a cantilever spring made of Elgiloy (Co-Cr-Ni alloy) in a Teflon jacket and which compresses from the load of the lambda plate and hanging magnet by gravity. The load required to completely compress the seal is approximately 105 N/cm. The seal for this lambda plate is designed to reach full compression and optimal sealing with 23 kN, which is more than provided by the top plate and hanging magnet. The seal was designed to work such that all

Mycom 800 hp Compressor (2 Stages) 160 g/s Sullair 500 hp Compressor (Single Stage) 80 g/s Sullair 350 hp Compressor (Single Stage) 51 g/s

Buffer Tank

Reciprocating Helium Expander CTI Model 4000 1500 W Refrigerator/ Liquifier CVI – Magcool Precooler/Subcooler (Cold Box) Turbo-Expander Helium Liquifier CVI/Air Liquide 1000 W Helium Wet Expander Koch Model 1600

Cryenco Inline Purifier Cryenco 10000 L Storage Dewar

High Pressure Helium Gas High Pressure Helium Gas @ 8K Liquid Helium Liquid Helium Warm Return

1.9 K Test Cryostat 2 Cryenco 3785 L Storage Dewar Gardner 3785 L Storage Dewar

Dual Turbines High Pressure Helium Gas Lambda Plate Warm Return 4.5 K 1 bar 1.9 K

Buffer Tank to Suction CS4 Sullair 100 (75 kW)

CS4

1.9 K Test Cryostat 2-Stage CVI/Magcool Purifier A & B Liquid Helium Storage Dewar 10000 L High Pressure Helium Gas Storage 8.5 x 106 L @ 15.5 bar Nash 1.9 K Vacuum Station Buffer Tank to Helium Gas Storage Warm Helium Return 4.5 K Helium Input

1.9 K 4.5 K 16 mbar 1 bar

3

gaps can be handled if the seal is not fully compressed or the mating parts not completely flat. The seals for the warm bore tube and the 1.9 K heat exchanger are similarly spring- energized but in a radial configuration where the forces are applied radially by the insertion of the part rather than by gravity as is the case with the face seal. More details of the upgraded cryogenics facility and vertical test cryostat, includ- ing lambda plate details, are shown in [3].

• Fig. 3. Top plate and header system designed for MQXFA magnets.
• III. POWER SUPPLY AND QUENCH PROTECTION

For the main power supply to power MQXFA magnets, the previous 30 kA short sample cable facility power supply, which consists of two 15 kA power supplies in parallel, has been updated and modified to supply 24 kA to an inductive

• load. New electronics and feedback loop and PID tuning have

been added to handle magnet-sized inductances, and filtering was re-configured to minimize the ripple. A schematic detail- ing the re-designed power supply can be seen in Fig. 4. It is al- so showing that a fast energy extraction (EE) system (for quench protection) has been added by installing six IGBT switches and a 150 mΩ ceramic, non-inductive dump resistor for each power supply. Each dump resistor is rated to dissipate up to 5 MJ of energy during a quench. Values of dump re- sistance can be varied over a range (0, 30, 37.5, 50, 75, and 150 mΩ). The switches are fast as the IGBT turnoff time is about1-2 μs. In addition, each IGBT is equipped with a spe- cially designed snubber circuit to limit the maximum switch- ing transient collector-emitter voltages to 800 V to avoid dam- age, and though each IGBT is rated at 3.6 kA, they are being limited to 2.0 kA for an added margin of protection. All IGBT devices are continuously being monitored during testing for such critical parameters as temperature, current sharing, and collector-emitter voltages. At a trip of the quench detector (QD), a fast data log of these parameters is generated to be an- alyzed for proper operation. Further details of this system can be found in [4].

• Fig. 4. 30 kA Power Supply System with Energy Extraction

Another source of quench protection is the CERN-supplied CLIQ system, details of which can be found in [5], [6].This system, which will be used in the LHC instead of EE, supplies a damped RC oscillation with maximum amplitude 500 V, which is triggered by a quench detector trip, along with the QPH system, the power supply shutoff, and the fast data log-

• ger. The AC loss heating generated by the CLIQ signal creates

added quench volume in the coils and therefore lowers the maximum quench temperature by spreading the energy. In or- 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 is necessary to keep current from back-flowing during ramp- ing of the magnet during the first few thousand amperes. Also the configuration shown allows for the triggering of both CLIQ and EE independently when the QD trips. Therefore de- lays can be utilized in both systems to add versatility during quench protection studies when testing the pre-production MQXFA models.

• Fig. 5. 30 kA Power Supply System with CLIQ

The MQXFA quench detection system is based on a redun- dant scheme which relies on a number of voltage signals. When any one of the signals reaches its specified threshold

Heat exchanger CLIQ Leads Connector Trees Quench Heater Connector Tree 24 kA vapor-cooled leads Top plate 76.2 mm thick 304 SS He fill lines 3rd CLIQ Lead (if needed)

3 Phase 480 V

X 6

15 kA DCCT 100 µH 15.5 mF 85 mΩ 62.0 mF 24 kA water cooled cable 15 kA water-cooled cable

15 kA / 14 V Power Supply 6 Pulse SCR Bridge

X 6

15 kA DCCT 100 µH 16.0 mF 85 mΩ 81.9 mF 33 mΩ 33 mΩ 15 kA water-cooled cable

IGBT (6 parallel units) Switch with snubber circuits

24 kA water cooled cable 24 kA Vapor-Cooled Leads 1.9 K 4.5 K 800 mF 800 mF 1 Ω 33 mΩ 33 mΩ 1 Ω

IGBT (6 parallel units) Switch with snubber circuits

15 kA / 14 V Power Supply 6 Pulse SCR Bridge

15 kA water-cooled cable 15 kA water-cooled cable

Magnet & Cryostat 24 kA Cable (60 ft)

CLIQ Circuit CERN Built

REE1, Energy Extraction Resistor REE2, Energy Extraction Resistor

S2, 6 IGBT Switches In Parallel S1, 6 IGBT Switches In Parallel

DCCT DCCT 30 VDC 15 kA Power Supply BNL Built Control For CLIQ Circuit

High Voltage Power Supply Control

BNL Built Quench Detector

IGBT Control Power Supply Control

Filter Filter

Tc Vo

GAS COOLED LEAD GAS COOLED LEAD

CLIQ Switch Control C

• P. Joshi. BNL July 18 2017

Current Sensor D2

24 kA Cable (60 ft)

D2

D3 D3 1000 A Shunt

15 kA Cable (85 ft)

L1 L2 L3 L4

20 DIODES

30 VDC 15 kA Power Supply

4

voltage after a specified validation time, a fast discharge is in- stigated (current decays through dump resistor). These signals include: the half coil difference, whole coil, quarter coil dif- ferences, whole coil minus the calculated ramp induced volt- age and SC lead voltages. See [3] for a detailed schematic and a table of nominal threshold voltages and validation times. There are also designed into the system a number of inter- locks which also rely on various signals which, if they do not meet specified conditions such as voltage threshold or on-off state, will instigate a slow discharge, with the dump resistor

• ut of the circuit. These include: vapor-cooled lead voltages,

QPH capacitor bank not charged, IGBT switch temperature or voltage too high, and many other monitored critical parame-

• ters. In addition, in Nb3Sn magnets experience voltage spikes

due to flux jumps, so current-dependent voltage thresholds as high as several volts at low currents are necessary in order to avoid false QD trips. The variation of thresholds with current is set in the programming and so is automatic during ramps.

• IV. DATA ACQUISITION

The older data acquisition (DAQ) system has been replaced by new National Instruments (NI)/LabVIEW hardware and software in order to provide the required faster and higher pre- cision data collection. The main DAQ systems include a fast data logger, slow data logger/monitor, strain gauge DAQ, and QD DAQ. At this time, the fast logger has 128 differential channels with 16-bit ADC resolution and simultaneous sam- pling of up to 250 kHz and is used mainly for voltage taps. Slow logger has 64 18-bit channels for up to 500 kHz for monitoring of vapor-cooled lead voltages, splice voltages, SC 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

• above. Strain gauge DAQ has 60 22-bit channels with scan-

ning of 30 channels each on two Agilent 34970A Data Acqui- sition/Switching Units with 250 channels/s scan rate. The number of DAQ channels can be expanded if necessary.

• V. TEST FACILITY COMMISSIONING AND TEST RESULTS

The upgraded test facility was commissioned in late 2016 and early 2017 at 1.9 K. The first magnet tested was MQXFPM1, a mirror magnet with the first long coil of MQXFA cross-

• section. Up to that time, only short models had been tested, in-

cluding a short mirror MQXFSM1 [7]. In addition to a cam- paign of spontaneous training quenches, mostly at 20 A/s, to verify the electric and mechanical stability of the long coil, there was also quench protection heater studies to determine minimum quench energy and quench delay for the three quench heater circuits, each having two strips on the inner lay- er, the outer layer high field region near pole, and the outer layer low field region near mid-plane [8]. Future MQXFA quadrupole magnet tests will also include magnetic field measurements, flux jump spike detection, ramp rate studies, and a quench antenna system to help locate quenches in re- gions where voltage taps do not exist. The production models will have only those taps needed for quench detection, as dis- cussed earlier. As can be seen by the quench training results in Fig. 6 for the MQXFPM1 magnet, there were 19 spontaneous

• quenches. The magnet reached the nominal current in four

quenches and the ultimate current by the tenth, showing that the long coil at least equaled the results of the short mirror test [7]. Out of the 19 quenches, 14 were in inner layer pole turn straight sections, which have the highest local field values. Only two quenches were in the outer layer, and these were the first quench and the third (first after thermal cycle). The last quench which was the highest at 19.241 kA, 7.6 % higher than the ultimate, and was at 4.4 K, not superfluid; the previous quench was at 300 A/s and was similar to the 20 A/s quenches. The straight section voltage signals gave evidence that the lo- cation of the pole turn quenches were not the same. These facts imply that the magnet was still training at the end of test- ing.

• Fig. 6. MQXFPM1 Quench Training History
• VI. CONCLUSION

As can be seen by the results for the MQXFPM1 magnet commissioning test, the upgraded facility as described in this 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 high luminosity upgrade. These requirements include opera- tion at 1.9 K and 1 bar, powering up to 19.241 kA, and the proper protection of the magnet during quench tests. For the test of the MQXFPM1, quench protection included a newly designed and faster quench detection system, energy extrac- tion with faster and state-of the-art electronic switching with specially designed IGBT circuitry and infrastructure, and fu- ture magnets will be using the already upgraded energy extrac- tion system with a higher power and more versatile dump re- sistors, and twelve new quench protection heater firing units, which meet the requirements for quench protection heater sys- tems in the LHC. ACKNOWLEDGMENT The authors would like to thank William McKeon and Sebas- tian Dimiauta for their hard work and expertise.

10000 11000 12000 13000 14000 15000 16000 17000 18000 19000 20000 21000 22000 23000 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 QUENCH CURRENT (A) QUENCH NUMBER A3-A4 Pole Multiturn (at 3.25-4.3K) A5-A6 Pole Turn Long Straight Section A6-A7 Pole Turn NonLead End Section A7-A8 Pole Turn Short Straight Section A8-B8 Ramp B3-B2 MIDPLANE MULTITURN THERMAL CYCLES NO QUENCH-50A/s to 10kA then 20A/s NO QUENCH-50A/s NO QUENCH-20 A/s Quenches 1-12 and 14-17 were at 20A/s. Quench 3 was in A6-A7 but close to or at tap A7. Quench 13 was at 20A/s with 2hr hold at 17890A. NO QUENCH 4.424K NO QUENCH

SS 22095A ULT 17890A NOM 16480A 300 A/s 3.265 - 4.424K 144K (TOP) - 110K (BOT) 117K (TOP) - 77K (BOT)

5

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.