1. David Cussans, University of BristolCERN, 7 th October 2008 1 Air Core Magnetic Components for CMS SLHC Tracker DC-DC converters David Cussans, Powering Working Group, CERN, 7 October 2008

2. David Cussans, University of BristolCERN, 7 th October 2008 2 Air Core Magnetics for SLHC  Started to look at different types of air-core magnetic components for CMS SLHC Tracker power supply. Toroidal inductor Planar transformer

3. David Cussans, University of BristolCERN, 7 th October 2008 3 Buck Configuration  Baseline configuration Proof of principle device being developed at CERN See talk by Michelis at TWEPP07, http://indico.cern.ch/contributionDisplay.py?contribId=48&sessionId=21&confId=11994 http://indico.cern.ch/contributionDisplay.py?contribId=48&sessionId=21&confId=11994 Simple and flexible configuration Step-down ratio determined by switching duty cycle Energy stored in inductor Large dV/dt across switches when they open/close Not the best configuration available in terms of EMI Difficult to retain high efficiency at high step-down.

4. David Cussans, University of BristolCERN, 7 th October 2008 4 Buck Configuration Vin=12-24 V Vout=1.5-3V Iout=1-2A Rad-hard technology Inductor Power dissipation Michaelis, Faccio, et. al., CERN

5. David Cussans, University of BristolCERN, 7 th October 2008 5 Transformer Based Converter  With a well designed transformer there is little external magnetic flux ( the magnetic field transfers the energy during the cycle, rather than storing it ) Pointed out by Brian Hawes, Oxford  Step-down ratio determined by turns ratio. Switches operate at ~ 50% mark/space ratio. Max. frequency determined by shortest pulse, so can be higher (all other factors being equal) than buck configuration

6. David Cussans, University of BristolCERN, 7 th October 2008 6 Transformer Based Converter  Commercial chips available aimed at low-EMI DC-DC converters See Linear Technologies LT1533 Claim < 100 μV pk-pk noise possible with 1A supply Would use sync. Rectifiers LT1533 only 200kHz

7. David Cussans, University of BristolCERN, 7 th October 2008 7 PCB based magnetics  Many commercial DC-DC converters use inductors and transformers with windings fabricated into the PCB Haven't done a market survey, but just pry the covers off a few 48V—>2.5V converters and have a look.... The converters I have seen use ferrite components.  Is is practical to manufacture air-core magnetics into the PCB carrying the active components?

8. David Cussans, University of BristolCERN, 7 th October 2008 8 Air Core Toroid  A toroid has only small external field Still get field generated from overall “single turn”  What values can be achieved in a PCB? Calculation straight-forward. See for example http://ieeexplore.ieee.org/iel5/4341939/4341940/04342265.pdfhttp://ieeexplore.ieee.org/iel5/4341939/4341940/04342265.pdf Prototype: h=1.6mm, d i = 12mm, d o = 28mm, N= 30 (Double-side FR4) L ≃ 244nH

9. David Cussans, University of BristolCERN, 7 th October 2008 9 Air-core Toroid  Built prototype in low-cost standard PCB process. 35 μm copper.  L = 240 +/- 10uH  R dc = 205 +/- 10 mΩ Resistance too high Want <90mΩ Need more layers or exotic PCB fabrication (e.g. filled vias ) Could make lower L, lower R dc

10. David Cussans, University of BristolCERN, 7 th October 2008 10 Magnetic Field Measurements  Difficult to measure absolute magnetic field, but can get a good idea of relative field  Use tracking generator in spectrum analyser to drive test inductor and field probe to measure B

11. David Cussans, University of BristolCERN, 7 th October 2008 11 Magnetic Field Measurements  Measured at 10cm in axial direction  Single loop  Single 500nH air-core solenoid  Two solenoids, field parallel  Two solenoids, field anti-parallel  PCB toroid

12. David Cussans, University of BristolCERN, 7 th October 2008 12 Magnetic Field Measurements  Can change distant field by large amount without significantly changing inductance.

13. David Cussans, University of BristolCERN, 7 th October 2008 13 Air Core Planar Transformer  Research into planar transformers to drive column-parallel CCDs done by Brian Hawes at Oxford for the LCFI project.  Need to drive 10A @ 50MHz  Spice models developed and matched against physical parts  FEMM and FastHenry used for FEM modeling  Brian has supplied us with some of the prototypes.

14. David Cussans, University of BristolCERN, 7 th October 2008 14 Air Core Prototypes  Prototypes in 10-layer PCB

15. David Cussans, University of BristolCERN, 7 th October 2008 15 Air Core Prototypes  16:1 turns ratio Four layers each with four turns in series for primary Five single turn secondary windings in parallel Primary centre-tapped

16. David Cussans, University of BristolCERN, 7 th October 2008 16 Lumped Element Model  Modeled in Spice Input Impedance

17. David Cussans, University of BristolCERN, 7 th October 2008 17 Measured Impedance  Iterate to match model to measurement

18. David Cussans, University of BristolCERN, 7 th October 2008 18 Finite Element Modeling of B-Field  Shows section through transformer  8:1 primary:secondary with centre-taps  35um copper on 50um dielectric  Field lines with only primary energized, 250mA  B at 10cm = 287 nT

19. David Cussans, University of BristolCERN, 7 th October 2008 19 Finite Element Modeling of B-Field  Field from secondary largely cancels the field from the primary:  Primary/Secondary 250mA/1A, B = 12nT at 10cm

20. David Cussans, University of BristolCERN, 7 th October 2008 20 Finite Element Modeling of B-Field  Stray field can be shielded without serious disturbance of transformer operation  35um copper shield, 200um from outer layers, B = 60pT at 10cm  Resistive loss = 2.8mW

21. David Cussans, University of BristolCERN, 7 th October 2008 21 Transformer Prototype  Manufactured in-house

22. David Cussans, University of BristolCERN, 7 th October 2008 22 Planar transformer  Reflection from primary with secondary terminated to 50Ω  Resonances above likely operating frequency (good)

23. David Cussans, University of BristolCERN, 7 th October 2008 23 Test-Stands  Setting up measurement stand for conducted noise. Copying CERN / Aachen test-stands All components (LISN, current probes, differential probes, spectrum analyzer) in-hand Will be able to measure noise produced by prototype converters. Will be able to measure susceptibility to noise.  Setting up a CMS strip module test-stand ARC stand (Aachen design) Many thanks to UCSB for providing components.

24. David Cussans, University of BristolCERN, 7 th October 2008 24 Test-stands  Combine to give module test-stand with injected noise, or test DC-DC converter.

25. David Cussans, University of BristolCERN, 7 th October 2008 25 Short-Term Plan of Action  Supply test inductors on interface board to Aachen ( 500nH, <100mΩ toroid, also solenoid on interface board for comparison.  Make a DC-DC converter prototype based on transformer. Produce another transformer with 10:1 ratio and aim for 15V --> 1.3V converter. Concentrate on low-noise rather than high efficiency Use air-core toroid as output filter?  Measure noise of prototype DC-DC converter ( at CERN, Aachen or Bristol)

26. David Cussans, University of BristolCERN, 7 th October 2008 26 Conclusions  Designing the magnetic components will be a vital part of designing any DC-DC switch mode converter. Can't be done in isolation. Needs contact with silicon designers. Needs careful evaluation  Fabricating magnetics into PCB offers increased integration and possibly improved reliability and performance.  Transformer based converter can give high step-down without efficiency degrading.