1. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE ENERGY EXTRACTION Development of Semi-Conductor Switches Alexandre Erokhin, Gert Jan Coelingh, Bozhidar Panev, Knud Dahlerup-Petersen TE-MPE-EE

2. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE OUTLINE  S.C.E.E.: Why? Where? When?  Current development and topologies  IGxT x = B or C is the question  Design Criteria  600 A  SM18  Summarizing Questions  Where are we today?

3. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Reminder

4. WHY ?  THREE DIFFERENT REASONS  1 - Demand coming from MSC group (SM18) for Extraction Systems : Fast (< 1 ms), High Current (10 kA, 30 kA and 14 kA), Medium Energy (4 – 5 MJ).  Fast (< 1 ms) means; can not be done with Electro-Mechanical Breakers! »Alternative: Semi-Conductors  2 - LHC: Increasing worry 600 A spare Energy Extraction Switches/Systems Back-Up solution in case of major events or degrading… Maintenance (preventive and corrective) is time and resource consuming  3 – (HL-)LHC: Development of New Inner Triplet Magnets most likely will need Fast, High Current Energy Extraction Systems – In house knowledge and experience will be present!

5. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE WHERE & WHEN ? TWO MAIN LOCATIONS: 1 - SM18 for Extraction Systems : FRESCA HFM 10 kA circuit - 2016-2017 D Cluster 30 kA circuit - 2018 A Cluster 14 kA circuit - 2019 2 – LHC tunnel: Replacement of existing 600 A Energy Extraction Systems UA and UJ33 areas radiation free – LS3 RR areas rad- tolerant version – LS3 2a - LHC: New Inner Triplet Magnets – LS ?

6. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Current Development 1 – High Current for SM18: 10 kA – 30 kA Design phase 2 – Medium Current for LHC tunnel 600 A Test phase (Re-design)

7. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Different Topologies 2 – Medium Current for LHC tunnel Redundant (min. 2 elements in series) No elements in parallel (600 ADC max) Bi-polar (current in both directions) semi-conductor is mono-polar 1 – High Current for SM18: Not redundant (no elements in series) Many elements in parallel (~2kA per branch) Mono-polar (current in one direction)

8. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Semi-Conductors candidates 3.5 kA

9. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Insulated Gate Bipolar Transistor (IGBT)   Developed based on (power) MOSFET technology in the 1980s   Combines the power handling capability of the bipolar transistor and the advantages of the isolated gate drive of the power MOSFET.   It has the advantages of being a minority carrier device leading to good performance in the on-state, even for high voltage devices.   With the high input impedance of a MOSFET it can be driven On or Off with a very low amount of power (voltage-driven).   The voltage drop in the on-state for high current DC applications between 1.5 to 3.5 V   IT IS NOT A BI-POLAR DEVICE! Current only in one direction.

10. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Integrated Gate-Commutated Thyristor (IGCT)   Developed in the 1990’s based on Gate-Turn-Off Thyristor (GTO) and the Gate Commutated Thyristor (GCT).   The IGCT is basically an improved GTO with also a four layer npnp structure with anode, cathode and integrated gate terminals. The integration of the gate terminals leads to extreme low coupling inductances allowing a faster turn-off process with lower losses.   The result is a combination of the strengths of both the GTO and the Insulated Gate Bipolar Transistor (IGBT) without the complex control unit of the GTO and the snubber-needs of the IGBT.   Voltage drop in on-state for high current DC applications between 1.5 V and 2.5 V

11. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Design criteria 1 Reliability Major failure causes due to Thermal design (DC application) Device voltage rating - stray inductances Cosmic Rays (FIT see next slide) Forward losses (Overrating!) Commercial Availability

12. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Design criteria 2 The Failures In Time (FIT) rate of a semi- conductor is the number of failures that can be expected in 10 9 hours of operation or for 500 devices during 2 million hours (228 years)  strongly dependent on the applied voltage  small dependence on temperature Typical FIT numbers for IGCT: 100 & Gate driver: 200 => 300 IGBT: 250 & Gate driver: 150 => 400

13. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE 600 A

14. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE 600A Energy Extraction Systems   202 systems installed in the LHC tunnel Corrector circuits with stored energy between 2.2 and 150 kJ   In 15 different locations; 8 x UA parallel service tunnel and 6 x RR and 1 x UJ tunnel caverns   Systems developed in close collaboration between CERN and the Budker Institute of Nuclear Physics (BINP), Novosibirsk, Russia. Reminder

15. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Remember: Different Topologies  2 – Medium Current for LHC tunnel Redundant (min. 2 elements in series) No elements in parallel (600 ADC max) Bi-polar (current in both directions) semi-conductor is mono-polar  1 – High Current for SM18: Not redundant (no elements in series) Many elements in parallel (2kA per branch) Mono-polar (current in one direction)

16. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Why worry after LS1? This is not building 281

17. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE

18. Why worry after LS1?  Still 50 spare breakers available  Spare consumption rate (before LS1) approx. 10/year 5 years of “operation”  Spare consumption rate (after LS1) ?????? < 5/year but not less than 2 !  Manufacturer closed premises at the end of 2012 and the production suspended  Breakers are now obsolete  Spare Systems reduced from 23 to 7 due to water incident in UA67. So far, unknown if and how much we can recuperate the damaged systems

19. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Global planning 600 A EE   Change (partially ?) to semiconductor-based switches, used as static breakers   IGBT (Insulated-Gate Bipolar Transistor) or IGCT (Integrated Gate-Commutated Thyristor) R&D on-going / necessary: 2012 development and testing of mono-polar Lab version 2013 development and testing of Bi-polar Lab version 2014/2015 Design, development and testing of prototype 2016 Final Design: proto to series 2017/2018 Market Survey/Tender

20. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Achievements   Based on IGCT (Integrated Gate-Commutated Thyristor) R&D on-going: 2012 development and testing of mono-polar Lab version  2 IGCT and 1 Diode in series 2013 development and testing of Bi-polar Lab version  2 IGCT and 1 Diode in series + anti-parallel 2 IGCT and 1 Diode in series   IGBT study on going Candidates found

21. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE IGCT semi-conductor

22. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Monopolar IGCT 600 ADC Switch

23. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Monopolar Switch Test Circuit

24. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Energy Extraction 600A

25. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE On-State Characteristics two IGCT + one diode) @ 25°C Voltage drop at 85 ⁰C 2.7

26. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE single components @ 25°C On-State Characteristics Single Components @ 25°C

27. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Bi-Polar IGCT 600ADC switch Diode is needed to cover the reverse-voltage during extraction

28. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Bi-Polar IGCT 600ADC switch

29. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Electronics IGCT  Electronics comes from existing 600 A EE systems (3 cards modified by Bozhidar)  Optical transmission of signals (Semi-Conductor type independent)  Redundancy in electronics. One Interface PCB per IGCT in one branch

30. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Energy Extraction 250A

31. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Transition pos to neg – 40A/s Circuit Current Dump Resistor Current Voltage EE system Current negative IGCT Current positive IGCT Less problems at 10A/s and Power Converter settings can be optimised

32. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Next steps 1  Choice of semi-conductor: IGBT slightly higher forward losses compared to IGCT Workaround possible by overrating the device Available in Press-Pack or Modules Wide choice of manufacturers (Infineon, ABB, Westcode) Voltage driven («static» switch device «fail safe») Snubber capacitors and varistors needed Driver needed

33. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Next steps 2  Choice of semi-conductor ctd. : IGCT Lower forward losses Available in press pack only ABB and Mitsubishi are today the only manufacturers Current controlled («active» switch device.. less fail safe) Snubbers needed (Varistor) but simpler than IGBT Integrated driver

34. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Next steps 3  For LHC application; any Semi-Conductor EE system will need water-cooling Total forward losses: 3V x 600 A = 1800 W x 2 systems = 3.6 kW per rack Water cooling capacity in the tunnel areas available?  Power Converter Boost Voltage: enough to overcome extra voltage drop? Seems ok for 200 oo 202 systems if U F is lower or equal to 3.5V  Power Converter 0 ampere crossing: can it be regulated better?  Desperate need for a designer!

35. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Modular Design  Independent of choice of semi-conductor: Identical Electronics Chassis with optical in/output Predefined Power Chassis with either IGBT, either IGCT switch, including drivers eventually included dump-resistor Similar as the power part of the EPC power converter One water in- and outlet per rack Water-cooling topology is a lot(!) simpler (higher reliability) for modules compared to press-packs

36. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE General Rack- Layout

37. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Budgets Change (partially – only rad free areas) to solid state breakers Semi-conductor technology EE systems Budget for R&D? Budget for replacement of 202 systems IGBT: 5.8 MCHF Budget for replacement of 202 systems IGCT: 6.7 MCHF Only UA/UJ areas (rad free) would be 75% of total Plan B: New Electro-Mechanical Circuit Breaker Budget for replacement of 202 systems EM-CB: 5.1 MCHF Similar problems as today to be expected  but before getting there.. development starting from scratch – Very time consuming and expensive study!

38. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE SM18

39. Remember: Different Topologies  2 – Medium Current for LHC tunnel Redundant (min. 2 elements in series) No elements in parallel (600 ADC max) Bi-polar (current in both directions) semi-conductor is mono-polar  1 – High Current for SM18: Not redundant (no elements in series) Many elements in parallel (2kA per branch) Mono-polar (current in one direction)

40. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Parallelism - Design  Highest available device is rated 5500A (IGCT) or 4800 A (IGBT) But for DC applications this means 2500A in order to stay in the SOA (Safe Operating Area) A5 surface dimensions to give an idea  For 10 kA this means roughly 4 devices parallel Major advantage: U F -25%  For 30 kA this means 12 devices parallel

41. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Master-Slave drivers commercially available Up to 4 IGBTs in parallel: hardware commercially available For a larger number of IGBTs => Careful synchronising on the high level electronics side. Still to be investigated.

42. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Parallelism - Design  Transient processes commutating very fast very high currents creates very high voltage spikes Stray inductance to be kept as low as possible 1 uH of non compensated stray inductance will lead to kV spikes Important parameter to stay within design voltage Need for intelligent geometrical mechanical design Trade-off efficient cooling & stray inductance Need for multi-level snubber compensation IGCT is slower than an IGBT: smaller/less snubbers can be used

43. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Snubbers Needed to reduce DV/Dt during switching Several layers of snubber capacitors over: the complete switch each semi conductor-branch each semi conductor

44. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Simulations – electrical and thermal  Simulations should start as soon as possible Simulate parasitical capacitances and inductances  Narrow down semi-conductor brand and type and get pspice models If not available create our own…

45. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE SM 18 Budgets EE switches   Fresca – HFM – 2 nd circuit 10 kA Budget switch: 100 – 125 kCHF Operational 2016 !!!   Cluster D circuit 30 kA  Budget switch: 300 – 350 kCHF  Operational 2017? Tbc..  Horizontal Banc A circuit 14 kA  Budget switch: 150 – 175 kCHF  Operational 2019? Tbc..

46. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Summarizing Questions   In spite of the already huge job done so far and the gathered knowledge and experience there are still important questions: Paralleling large numbers of IGBTs or IGCTs. Geometric mechanical topology. Clear ideas up to 10kA (since it can fit to one rack). 14-30kA circuits layout not so clear. Still to be investigated. Design will be Trade-off between efficient cooling & stray inductance. Where to focus? Bus bars topology. For 30kA at 1.7 A/mm 2 => 17650 mm 2 =>42x42cm! Water cooled bus bars ? Indirect cooling? Topology should be chosen also according to SM18 infrastructure Switch-on-off criteria – No overcurrent during ramp-up? Will the last off- switching branch take all current (ns range)?

47. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Where we are today? Crucial point: continue only by ourselves (CERN re-inventing the wheel) or Study existing installations in collaborating institutes 2016.. not much time => no time for trial-and-error Main responsibility still LS1 and, after re-start, LHC operation Not to copy but learning from their experiences Desperate need for Electro-Mechanical CAD designer (preferably with plumbing experience)

48. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE SM18 and/or LHC Dump-Resistors

49. MPE-TM 17/10/2013 G.J. Coelingh TE-MPE-EE Compact Dump-Resistors Development of Energy Absorbers for medium- to high energy applications : Compact design for use in LHC underground locations Primary cooling by immersion into dielectric, heat-transfer liquid Built-in liquid-to-water heat exchanger Modular design to facilitate change of resistance value (typically 10 to 300 mOhm). For energy absorption up to 10 MJ. Development cost + prototype: 150 kCHF Agreement of principle for collaboration with BARC, India