powering: status & outlook cec meeting, karlsruhe may 18th, 2011 katja klein 1. physikalisches...

Powering: Status & Outlook

CEC Meeting, KarlsruheMay 18th, 2011

Katja Klein1. Physikalisches Institut B

RWTH Aachen University

Overview

Katja Klein 2Powering - Status and Outlook

Novel power distribution schemes are being / have to be developed for:

(1) Phase-1 pixel detector (target date ~ shutdown 2016/2017)- DC-DC buck converters with conversion factor of ~ 3

(2) Phase-2 tracker upgrade (target year 2020)

a. Readout modules

b. pT modules

c. GBT system- DC-DC conversion scheme with conversion factor of ~ 10

(3) Phase-2 pixel detector??

Overview


• Work is discussed within the Tracker Upgrade Power WG (~ 4x / year)

• Active groups: ITA (Aragon), Aachen ( CEC), CERN, Fermilab, PSI, ...

• Work has settled on two types of converters- Buck converter

- Switched capacitor converter

• ... and can be divided into three categories:- Generic R&D: ASIC development, optimization for high efficiency and low

noise, development of shielding & inductor

- Phase-1 specific R&D: system design & system tests with pixel modules

- Phase-2 specific R&D: development of switched capacitor converter and its test with CMS Binary Chip

Challenges• Radiation tolerance of high voltage (~15V) power transistors

• Magnetic field tolerance air-core inductor radiated emissions

• Conductive switching noise

• Maximization of efficiency & minimization of material and size

DC-DC Buck Converter


High currents (~3A) with high efficiency Comparably simple & compact Output voltage regulation by Pulse Width Modulation logic (not shown)

Idea: on-detector voltage conversion: P = UI = (rU)(I/r) with r > 1

lower power losses in cables & less material

Duty cycle D = t1, on/T

Conversion ratio r = Vin/Vout = Iout/Iin = 1/D

DC-DC buck converters

ASIC

Switched Capacitor Converter


Pro: no air-core inductor needed no magnetic radiation, less material

Con: can provide only relatively small currents (< 100mA), no regulation

• Targeting phase-2 application- Could be combined with buck converter in a 2-step conversion scheme, to reduce

required buck conversion ratio and thus improve efficiency (e.g. 10V 2.5V 1.2V)- Could derive digital (0.9V) from analogue (1.2V) voltage

• Based on capacitors (instead of inductor) as energy storage element, “charge pump“

• Driving logic on-chip (possibly inside ROC), capacitors external

Buck Converter ASICs


• ASIC includes transistors and voltage regulation circuit

• ASIC is being developed within CERN electronics group (F. Faccio et al.)

• Radiation tolerance of many semi-conductor technologies evaluated Best candidate is AMIS I3T80 0.35µm (ON Semiconductor, US) - functional up to dose of 300Mrad & fluence of 51015 p/cm2

- no Single Event Burnout effect

• AMIS prototypes: AMIS1 (2008) AMIS2 (2009), some bugs, but used in system tests AMIS3 (just back from foundry), small improvements wrt AMIS2 AMIS4 (submitted in January 11), full and almost final functionality

• Work with second supplier (IHP, Germany) to improve radiation tolerance - two prototypes in 2010, but development on-hold due to rad. tolerance issues

SEB = Single Event Burnout = ionizing particle in source turns parasitic npn transistor on destructive current

Buck Converter ASICs


F. Faccio, ACES 2011

Aachen DC-DC Converter Development


ASIC: AMIS2 by CERNIout < 3AVin < 12VVout configurable; 2.5V & 3.3Vfs configurable, e.g. 1.3MHz

PCB:2 copper layers a 35µm0.3mm thickLarge ground area on bottom for cooling

Toroidal inductor:L = 450nHRDC = 40m

ShieldA = 28 x 16 mm2

M 2.5g3.8% of a radiation length

“PIX_V7“:

Design guidelines from CERN grouphave been implemented.

Pi-filters at in- and output

The Shield


The shield has three functions:

1) to shield radiated emissions from inductor

2) to reduce conducted noise by means of segregation between noisy and quiet parts of board (less coupling)

3) to provide cooling contact for coil through its solder connection to PCB, since cooling through contact wires not sufficient

We are currently investigating several technologies:

• Aluminium shields of 90µm thickness (milled in our Workshop)

• Plastic shields (PEEK) coated with a metall layer (outside, inside & outside)- Aluminium sputtered (5 - 10µm)- Copper/Nickel sputtered (15 - 30µm)- Copper, galvanic deposition- ...

Efficiency


PIX_V7, Vout = 3.3V

• Phase 1 conditions: Vout = 3.3V or 2.5V, Iout < 2.8A, conversion ratio of 2 - 3 about 75% efficiency: ok

• Phase 2 conditions: Vout = 1.25V, Iout = 3A, conversion ratio of 8 - 10 about 55% efficiency: too low- Improvements expected from new ASICs, bump bonding, IHP?, ...?- If all fails: move to 2-step conversion scheme

PIX_V4_R3, Vout = 1.25V

Conductive Noise


Differential Mode, no shield Common Mode, no shield

Differential Mode, with shield Common Mode, with shield

Further reduction possible for higher switching frequency ( lower efficiency)

PIX_V7output noiseVout = 3.3VVin = 10Vfs = 1.3MHzL = 450nH

Shielding from Radiated Noise


Shielding of magnetic field: Eddy currents in metallic shield

• Need to shield all parts with large current variations, not only the coil

• 90µm milled Aluminium shield works fine (but is very expensive!)

• Plastic shield coated with 30µm Cu worse and adds ~ 40% more material (but probably cheaper...)

No shield 90µm Alu 30µm Cu

Switched Capacitor Converter ASIC


• Work by Michal Bochenek (CERN PH-ESE)

• ASIC in IBM 0.13µm, conv. ratio ~ 2

• Simulated efficiency = 97% for 60mA output current

• Large voltage spikes due to wire bond inductance

• Back since February, results not yet presented

• Implemented in CMS Binary Chip (CBC)

Simulation without bonds

Simulation with bonds

CBC Power Aspects


• 130nm chip for outer tracker, under test since February

• Power consumption < 0.3mW/channel for 5pF strips 80mW for 256 channels @ 1.2V

• Three powering options implemented

1) Direct (1.1V for both voltages or lower voltage for digital part)

2) Via Low DropOut (LDO) regulator (on-chip, but bypassable) improve PSRR

3) Via switched capacitor converter (on-chip, but bypassable) (2.5V 1.2V)

• Effect of switching noise will be investigated by IC

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0V

olts

43210

microseconds

DC-DC output LDO input LDO output

Buck Converters for Pixel Upgrade

Katja Klein 15Powering - Status and OutlookKatja Klein 15

DC-DC converters

2.2m

Integration for pixel barrel detector onto supply tube Pseudorapidity ~ 4 material budget not critical

8 layer bus board and Alu cooling bridges possible! ROC has fast on-chip regulators insensitive to noise Large distance of converters to pixel modules Sufficient space

CO2 cooling pipesd

2 000 DC-DC convertersrequired in 2014(Aachen responsibility!)

Phase-2 Specific Challenges


Goals: Deliver large currents with existing cable plant, and save as much copper as possible in cables and mother boards inside the detector volume

• Conversion ratio as high as possible (Vin < 12V) potential efficiency issue

• DC-DC converters to be installed close to detector modules

very little space available, in particular for pT modules

material budget of DC-DC converters is very critical

• Have to provide up to four different operating voltages (readout chips: Vana = 1.2V & Vdig = 0.9V?, GBT: 1.5V & 2.5V)

find power distribution scheme with maximal benefit for material budget

Phase-2 Specific R&D Topics


• Understand possible efficiency gain with new ASICs (CERN, Aachen)

• Explore potential use of switched capacitors within readout chip (IC)

• Development of power distribution scheme

- Requires input from electronics group: chip power consumption, module & tracker strawman, etc. only paper exercises up to now

• System tests with new readout chips (first candidate: CBC)- Need access to new readout chips, plus DAQ system etc.

• Study mechanical and electrical integration of DC-DC buck converters into readout and pT modules or staves- Modularity: how many converters are required per module- Where to put them: on the module or on the support structure, top or back side etc.- How to fix them: separate PCB or part of hybrid etc.- Slow controls- How to cool them- Electro-magnetic interference: coupling to sensor & readout electronics- Study of material budget gain

Readout Modules


• 2 x 8 CBCs with 256 channels: P = 1.2W or I = 1A ( converter is under-used)

• Module design basically by Duccio

• Thermal behaviour and electromagnetic interference to be studied

• No motherboards foreseen

Stereo Modules


• 4 x 8 CBCs with 256 channels: P = 2.4W or I = 2A 1 buck converter per module 12 buck converters per rod

• Thermal behaviour and electromagnetic interference to be studied

Strip-based pT Modules


• 2 x 8 CBCs with 256 channels: P = 1.2W or I = 1A

• Could power 2-3 modules from one buck converter 6 bucks per rod

• 1 GBT per module required additional DC-DC converters

Pixellated pT Modules


• 3D approach by Lipton et al.

• Power estimates range from 3.5W to 10W for 10 x 10 cm2 module stack

• DC-DC converters to be integrated onto “beam“ buck converter delivered to FNAL

• Alternative proposals by Sandro Marchioro and Geoff Hall

beam super-module

Summary & Outlook


• R&D on buck converters makes good progress

• Focus is currently on phase-1 application

- Commitment from Aachen group to develop & produce bucks for phase-1

• For phase-2, there are additional requirements which still have to be adressed

• Switched capacitors are being studied with CBC very important input for phase-2

• Integration of converters into modules or support structures is currently mainly a paper exercise; main obstacles are:

- Still too many module concepts

- Availability of module and stave prototypes

- Man power

powering: status & outlook cec meeting, karlsruhe may 18th, 2011 katja klein 1. physikalisches...

Documents

powering status

detector voltage conversion

required buck conversion

ddcdc buck convertersasic

high efficiency

conversion factor

pixel modules phase

overviewkatja klein