1. Low Noise Dc to DC Converters for the sLHC Experiments Low Noise Dc to DC Converters for the sLHC Experiments TWEPP 2010 Aachen, Germany 21/9/2010 TWEPP 2010G. Blanchot, PH/ESE B.Allongue, G.Blanchot, F.Faccio, C.Fuentes, S.Michelis, S.Orlandi CERN – PH-ESE 1

2. Outline  DCDC based powering scheme.  DCDC development status.  Compatibility of DCDC with front-end systems.  Noise optimized DCDC Plug-in-Boards.  Shielding and radiated magnetic field.  Performance of AMIS2-PIB using radtol ASIC.  Performance of SM01B-PIB using commercial chip.  Conclusions. TWEPP 2010G. Blanchot, PH/ESE 2

3. TWEPP 2010G. Blanchot, PH/ESE Distribution scheme example (ATLAS Short Strip concept) Identified integration issues:  Radiated magnetic field from stage 1 DC/DC.  Board layout, coil topologies, shields  High noise susceptibility of modules.  System tests with hybrids  Board area, material budget on stage 1 DC/DC.  ASIC development, compact layout. 10-12V 2 Converter stage2 on-chip Detector Intermediate voltage bus(ses) Converter stage 1 block Hybrid controller SC and optoelectronics 10-12V Module/Stave Scheme based on 2 conversion stages: Stage 1: On Module Buck DC/DC Stage 2: On Chip Switched Capacitor Module 3

4. DCDC Development Status TWEPP 2010G. Blanchot, PH/ESE  Critical achievements:  Radiation tolerant technologies have been selected.  Buck converter ASIC prototypes have been produced and tested  Air core inductors topology has been selected.  Standard buck converter prototypes have been produced, tested and used with systems.  Radiation tolerant buck converter prototypes using ASICs have been produced.  Critical achievements:  Radiation tolerant technologies have been selected.  Buck converter ASIC prototypes have been produced and tested  Air core inductors topology has been selected.  Standard buck converter prototypes have been produced, tested and used with systems.  Radiation tolerant buck converter prototypes using ASICs have been produced. Coils optimization 7 mm Package QFN48 5 mm Package QFN32 7 mm Package QFN48 Radiation tolerant ASICs AMIS2 IHP1 IHP2: DCDC Prototypes IHP Technology still under development 4

5. Buck Converter TWEPP 2010G. Blanchot, PH/ESE  Identification of main noise sources:  Switched voltage at node area N3 = Vin at switch frequency + harmonics.  On-time switching loop “A” = uprising current.  Off-time switching loop “B” = down-rising current.  Transition-time loop “C” = fast current transition inside switches.  Magnetic field emitted by the main coil L= triangular current.  These noise sources originate noise currents within the DCDC board that in turn radiate fields along cables and interconnections.  Identification of main noise sources:  Switched voltage at node area N3 = Vin at switch frequency + harmonics.  On-time switching loop “A” = uprising current.  Off-time switching loop “B” = down-rising current.  Transition-time loop “C” = fast current transition inside switches.  Magnetic field emitted by the main coil L= triangular current.  These noise sources originate noise currents within the DCDC board that in turn radiate fields along cables and interconnections. B A C 5

6. ISL6540 Proto2 Noise reduction in DC/DCs TWEPP 2010G. Blanchot, PH/ESE  DCDC optimization:  Voltage nodes and current loop areas have been reduced significantly.  The subsequent reduction in radiated noise results in reduction of conducted noise along cables.  DCDC optimization:  Voltage nodes and current loop areas have been reduced significantly.  The subsequent reduction in radiated noise results in reduction of conducted noise along cables. ISL6540 Proto3 ISL6540 Proto5 The noise level is characterized on a reference test stand: The noise level is characterized on a reference test stand: The noise level has been considerably reduced on DCDC prototypes that used an Intersil ISL6540 controller and air core coils. The noise level has been considerably reduced on DCDC prototypes that used an Intersil ISL6540 controller and air core coils. Very good performance was achieved in Proto5. Very good performance was achieved in Proto5. Equally good performance was achieved using the AMIS2 ASIC. Equally good performance was achieved using the AMIS2 ASIC. This level of performance was however not enough for the sensitive trackers front- end electronics and detectors. This level of performance was however not enough for the sensitive trackers front- end electronics and detectors. AMIS2 QFN48 Test Board 6

7. Proto5 with UniGe Module TWEPP 2010G. Blanchot, PH/ESE Position of hybrids Conducted noise test Radiated noise at corner Radiated noise on top of hybrid Reference noise: ENC Average:560 ENC Sigma:32 Measurements performed with the help of Sergio Gonzalez from the University of Geneva. The susceptibility against radiated fields of the UniGe module was measured using the Proto5 DCDC. 7

8. Proto5 on UniGe Module TWEPP 2010G. Blanchot, PH/ESE Conducted noise test Radiated noise at corner Radiated noise on top of hybrid Proto5, shielded coil: ENCSigma Proto5, shielded coil: ENCSigma Reference: 560 32 Reference: 560 32 Conducted: 57133 Conducted: 57133 Radiated Corner: 58335 Radiated Corner: 58335 Radiated Top: 930176 Radiated Top: 930176  VCC and VDD powered from the same DCDC converter, without regulator on VCC.  The system is insensitive to the conducted noise of the converters.  High noise is observed when the DCDC is very close of the hybrids: susceptibility to radiated couplings.  VCC and VDD powered from the same DCDC converter, without regulator on VCC.  The system is insensitive to the conducted noise of the converters.  High noise is observed when the DCDC is very close of the hybrids: susceptibility to radiated couplings. 8

9. Radiated fields susceptibility measured at Liverpool Shielded 3cm probe, 12 MHz, 6mA DCDC Edge Position Shielded DCDC Edge Position Noise increases on all channels. Noise increases on all channels. B field coupling only. B field coupling only. Noise increases on all channels (all above 1000 electrons) due to B field. Noise increases on all channels (all above 1000 electrons) due to B field. Alternating pattern due to E field. Alternating pattern due to E field. No global increase: B field is shielded. No global increase: B field is shielded. Alternating pattern on two first chips: some E field remains, probably due to leaking E field. Alternating pattern on two first chips: some E field remains, probably due to leaking E field. Noise Reference = 650 electrons TWEPP 2010 G. Blanchot, PH/ESE Radiated fields need therefore to be mitigated further on. 9

10. Noise Optimized Plug-in-Boards TWEPP 2010G. Blanchot, PH/ESE  New generation of DCDC plug-in board to be used with systems:  A form factor compatible with front-end systems under development now.  More compact design.  Power interface: connector or bonds.  In some cases: control logic.  Better control of the noise sources for lower conducted and radiated couplings:  Understanding of how electromagnetic fields are emitted from power loops and switching nodes.  The reduction of radiated fields will result in reduced conducted noise.  Introduction of an electromagnetic shield:  to cancel E field couplings with front-end systems.  to mitigate the radiated B field down to compatible levels.  A thermal interface must be provided for cooling.  New generation of DCDC plug-in board to be used with systems:  A form factor compatible with front-end systems under development now.  More compact design.  Power interface: connector or bonds.  In some cases: control logic.  Better control of the noise sources for lower conducted and radiated couplings:  Understanding of how electromagnetic fields are emitted from power loops and switching nodes.  The reduction of radiated fields will result in reduced conducted noise.  Introduction of an electromagnetic shield:  to cancel E field couplings with front-end systems.  to mitigate the radiated B field down to compatible levels.  A thermal interface must be provided for cooling.  3 different DCDC-PIB have been designed and produced:  AMIS2_DCDC: 2 versions with AMIS2 radiation tolerant ASIC, implementing noise cancellation techniques.  10V down to 2.5V, rated for 2A.  The noise optimization method is explained by Cristian Fuentes at the Power WG.  SM01B: 1 version using a commercially available buck converter chip similar to AMIS2  10V down to 2.5V rated 5A for the 0.25um ABCN modules in use today.  Another DCDC-PIB is under design for bonding onto ATLAS Stavelets  3 different DCDC-PIB have been designed and produced:  AMIS2_DCDC: 2 versions with AMIS2 radiation tolerant ASIC, implementing noise cancellation techniques.  10V down to 2.5V, rated for 2A.  The noise optimization method is explained by Cristian Fuentes at the Power WG.  SM01B: 1 version using a commercially available buck converter chip similar to AMIS2  10V down to 2.5V rated 5A for the 0.25um ABCN modules in use today.  Another DCDC-PIB is under design for bonding onto ATLAS Stavelets 10

11. Noise Optimized Plug-in-Boards TWEPP 2010G. Blanchot, PH/ESE Enable Pgood Vin GND Vout PROTO5 AMIS2 SM01B  Board size reduction down to 26mmx13.5mmx9mm.  Increased switching frequency: 2 MHz on AMIS2, 3 MHz on SM01B.  A custom coil has been developped with an industrial partner : 250nH that will stand straight onto the AMIS2 ASIC.  A custom shield is under development now: it aims to replace the 200um copper foil boxes with Cu coated plastic cases to be soldered directly onto the PCBs.  Efficiencies above 80% achieved. SM01B reaches 87% at 2A, and is still at 80% for 4A load current at nominal input voltage.  Board size reduction down to 26mmx13.5mmx9mm.  Increased switching frequency: 2 MHz on AMIS2, 3 MHz on SM01B.  A custom coil has been developped with an industrial partner : 250nH that will stand straight onto the AMIS2 ASIC.  A custom shield is under development now: it aims to replace the 200um copper foil boxes with Cu coated plastic cases to be soldered directly onto the PCBs.  Efficiencies above 80% achieved. SM01B reaches 87% at 2A, and is still at 80% for 4A load current at nominal input voltage. SM01B 11

12. Shielding: Electric field TWEPP 2010G. Blanchot, PH/ESE B A C CB A N3 FE ASIC Bondings Filters Coil Electric field coupling Electric field shield  Electric field is mainly radiated by node N3: square wave of 10 V at switching frequency.  The field couples on the DC DC board filtered areas, on the output cables or traces and on the FE bondings.  The coupling is blocked very easily with the addition of a shielding case that segregates the filtered areas from the noisy areas on the DCDC board.  A plastic box with a very thin conductive layer is sufficient to provide E fied shielding.  The shield will reduce the conducted noise and also the couplings in the bondings.  Electric field is mainly radiated by node N3: square wave of 10 V at switching frequency.  The field couples on the DC DC board filtered areas, on the output cables or traces and on the FE bondings.  The coupling is blocked very easily with the addition of a shielding case that segregates the filtered areas from the noisy areas on the DCDC board.  A plastic box with a very thin conductive layer is sufficient to provide E fied shielding.  The shield will reduce the conducted noise and also the couplings in the bondings. 12

13. CB A N3 FE ASIC Bondings Filters Coil Shielding: Magnetic field TWEPP 2010G. Blanchot, PH/ESE B A C Radiated Magnetic Field Coupled Magnetic Field Radiated Magnetic Field  Magnetic field is mainly radiated by loops A and C and by the main coil L: triangular waves of up to 8A peak to peak at switching frequency with different emission spectrums for each loop.  The field couples on the DC DC board filtered areas, on the output cables or traces and on the FE board.  The coupling is mitigated with the addition of a shielding case that segregates the filtered areas from the noisy areas on the DCDC board.  To be effective, eddy currents must develop in the shield conductor material. At 2 MHz, δ = 50 µm of copper. The shielding effectiveness for Cu thickness from 10 µm to 100 µm will be studied.  The shield will reduce the conducted noise and also the couplings in the FE board.  Magnetic field is mainly radiated by loops A and C and by the main coil L: triangular waves of up to 8A peak to peak at switching frequency with different emission spectrums for each loop.  The field couples on the DC DC board filtered areas, on the output cables or traces and on the FE board.  The coupling is mitigated with the addition of a shielding case that segregates the filtered areas from the noisy areas on the DCDC board.  To be effective, eddy currents must develop in the shield conductor material. At 2 MHz, δ = 50 µm of copper. The shielding effectiveness for Cu thickness from 10 µm to 100 µm will be studied.  The shield will reduce the conducted noise and also the couplings in the FE board. 13

14. Radiated Magnetic Field TWEPP 2010 G. Blanchot, PH/ESE The radiated magnetic field is measured along X, Y and Z axes with a 1cm loop probe over a grid. The vector magnitude is computed. 13 5 1 PROTO5 AMIS2 SM01B Switching freq. = 1 MHz Switching freq. = 1 MHz L = 350 nH L = 350 nH Load = 1A. Load = 1A. dBµA/m] [dBµA/m]UnshieldedShieldedComment PROTO5115 Shielded coil only SM01B>120<100Non EMC optimized layout AMIS2110<100EMC optimized layout 14

15. AMIS2DCDC Conducted Noise TWEPP 2010G. Blanchot, PH/ESE Compared to Proto5: Compared to Proto5: AMIS2_DCDC has more than 20dB less CM noise, of about 300 nA at 3 MHz. AMIS2_DCDC has more than 20dB less CM noise, of about 300 nA at 3 MHz. Switch Frequency is now 3 MHz: there are less harmonic peaks in the sensitive band. Switch Frequency is now 3 MHz: there are less harmonic peaks in the sensitive band. Now complies with Class B with more than 20dB of margin. Now complies with Class B with more than 20dB of margin. The DM noise has also been reduced. The DM noise has also been reduced. Barely visible. Barely visible. Two peaks at 3MHz and 6 MHz only, with less than 300 nA amplitude. Two peaks at 3MHz and 6 MHz only, with less than 300 nA amplitude. Now complies with Class B with more than 25dB of margin. Now complies with Class B with more than 25dB of margin. ATLAS Limit Class A Limit Class B Limit To further mitigate the radiated fields, electric and magnetic near field couplings that take please within the DCDC board and its components were modeled. Based on this, noise cancelling routing and placement topologies were implemented onto a new generation of converters using the AMIS2 ASIC. On this, a shield is added. 15

16. Optimized AMIS2 DCDC on UniGe Module TWEPP 2010G. Blanchot, PH/ESE Conducted noise test Radiated noise at corner Radiated noise on top of hybrid AMIS2_DCDC, shield: ENCSigma AMIS2_DCDC, shield: ENCSigma Reference: 560 32 Reference: 560 32 Conducted: 56033 Conducted: 56033 Radiated Corner: 55834 Radiated Corner: 55834 Radiated Top: 61438 Radiated Top: 61438 VCC and VDD are each powered from two different DCDC converter, without regulator on VCC. The AMIS2-PIB, induces 10% more noise with respect to the reference configuration, when two converters are place straight on top of the hybrids. The improvement is very significant, and is in line with the noise reduction observed on the reference test stand (CM and DM noise). VCC and VDD are each powered from two different DCDC converter, without regulator on VCC. The AMIS2-PIB, induces 10% more noise with respect to the reference configuration, when two converters are place straight on top of the hybrids. The improvement is very significant, and is in line with the noise reduction observed on the reference test stand (CM and DM noise). 16

17. SM01B (LT3605) Performance TWEPP 2010G. Blanchot, PH/ESE Compared to Proto5: Compared to Proto5: CM of SM01B is comparable to that one of Proto5. Beyond 10 MHz it is 10 dB lower. CM of SM01B is comparable to that one of Proto5. Beyond 10 MHz it is 10 dB lower. Switch Frequency is now 2 MHz: there are half less harmonic peaks in the sensitive band. Switch Frequency is now 2 MHz: there are half less harmonic peaks in the sensitive band. Complies with class B except for the 4 MHz peak. Complies with class B except for the 4 MHz peak. The DM noise has also been reduced. The DM noise has also been reduced. 15 dB reduction at 2 MHz. 15 dB reduction at 2 MHz. 10 dB attenuation beyond 10 MHz. 10 dB attenuation beyond 10 MHz. 4 MHz peak still exceeds Class B. 4 MHz peak still exceeds Class B. ATLAS Limit Class A Limit Class B Limit The LT3605 chip is used in SM01B. It integrates the switches and the control circuitry like in AMIS2. It includes two PLLs to improve the voltage tracking but that result in broader peaks. Regular layout applied, with the addition of a shield. 17

18. SM01B DCDC on UniGe Module TWEPP 2010G. Blanchot, PH/ESE SM01B shielded: ENCSigma SM01B shielded: ENCSigma Reference: 560 34 Reference: 560 34 3cm on bonds: 56333 3cm on bonds: 56333 Radiated Top: 57336 Radiated Top: 57336 VCC and VDD are each powered from the same DCDC converter, with regulator on VCC for the analog power of the ABCN chips. The SM01B, induces less than 2% more noise with respect to the reference configuration, when the converter is placed straight on top of the hybrids. SM01B radiates more than AMIS2DCDC but less noise is observed:  2 AMISDCDC vs 1 SM01B.  Setups are slightly different: distance from DCDC to hybrid probably larger for SM01B.  No analog regulator for the AMIS2DCDC test.  It is in both cases an excellent performance for an extreme placement of converters. VCC and VDD are each powered from the same DCDC converter, with regulator on VCC for the analog power of the ABCN chips. The SM01B, induces less than 2% more noise with respect to the reference configuration, when the converter is placed straight on top of the hybrids. SM01B radiates more than AMIS2DCDC but less noise is observed:  2 AMISDCDC vs 1 SM01B.  Setups are slightly different: distance from DCDC to hybrid probably larger for SM01B.  No analog regulator for the AMIS2DCDC test.  It is in both cases an excellent performance for an extreme placement of converters. Reference noise 3cm from bonds Radiated noise on top of hybrid 18

19. Conclusions TWEPP 2010G. Blanchot, PH/ESE  The system tests performed at Liverpool and Geneva modules brought the necessary information to understand the critical noise sources and their coupling mechanisms.  As a result of this, an optimized PCB layout has been designed for the AMIS ASIC, with the addition of a shield enclosing the identified noise sources.  The noise optimized AMIS converters present a negligible level of noise on the reference test stands, with more than 20 dB improvement with respect to the previous prototype (Proto5).  The noise reduction observed on the test stand is confirmed when powering a front-end system: the noise increase in a critical placement is less than 10 % now with two AMIS2 and less than 2% using the SM01B and the ABCN regulator.  In similar conditions, the old Proto5 DCDC was inducing unacceptable levels of noise.  This validates the optimization methodology applied to the design of the new prototypes.  It validates the need for the shield.  The optimized design will be integrated onto supermodules, frame modules and stavelets currently under development.  The system tests performed at Liverpool and Geneva modules brought the necessary information to understand the critical noise sources and their coupling mechanisms.  As a result of this, an optimized PCB layout has been designed for the AMIS ASIC, with the addition of a shield enclosing the identified noise sources.  The noise optimized AMIS converters present a negligible level of noise on the reference test stands, with more than 20 dB improvement with respect to the previous prototype (Proto5).  The noise reduction observed on the test stand is confirmed when powering a front-end system: the noise increase in a critical placement is less than 10 % now with two AMIS2 and less than 2% using the SM01B and the ABCN regulator.  In similar conditions, the old Proto5 DCDC was inducing unacceptable levels of noise.  This validates the optimization methodology applied to the design of the new prototypes.  It validates the need for the shield.  The optimized design will be integrated onto supermodules, frame modules and stavelets currently under development. 19