A First Comparison Tracker Upgrade Power WG Meeting October 7 th 2008 Katja Klein 1 Physikalisches Institut B RWTH Aachen University Katja Klein Serial Powering vs DCDC Conversion 2 Outline ID: 784437
Download The PPT/PDF document "Serial Powering vs. DC-DC Conversion -" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Serial Powering vs. DC-DC Conversion -
A First Comparison
Tracker Upgrade Power WG Meeting
October 7
th
, 2008
Katja Klein
1. Physikalisches Institut B
RWTH Aachen University
Slide2Katja Klein
Serial Powering vs. DC-DC Conversion
2
Outline
Compare Serial Powering & DC-DC conversion under various aspectsPower loss in cablesLocal efficiencyCompatibility with servicesPower suppliesBias voltage
SafetySlow controlStart-up ScalabilityFlexibilityPotential
to deliver different voltagesProcess considerations & radiation hardnessInterplay with FE-chipInterplay with readout & controlsNoise
Material budgetSpaceTest systemsDiscussion
Slide3The Basic IdeasKatja Klein
Serial Powering vs. DC-DC Conversion3
Conversion
ratio r:
r = Vout / Vin ! << 1 Pdrop
= RI02n2
r2
Vdrop = R
I0Pdrop = RI0
2
Serial powering
Powered from constant
current source
Each module is on different ground potential
AC-coupled communication
Shunt regulator and transistor to take
excess current and stabilize voltage
Voltages are created locally via shunt
and linear regulators
Parallel
powering with
DC-DC
conversion
Need
radiation-hard magnetic field tolerant
DC-DC converter
One converter per module or parallel scheme 1-step or 2-step conversion
Slide4Katja Klein
Serial Powering vs. DC-DC Conversion
4
The Buck Converter
Convertion ratio g > 1
:g = Vin / Vout
Switching frequency fs:fs = 1 / Ts
The “Buck converter“ is simplest inductor-based step-down converter:
Slide5Katja Klein
Serial Powering vs. DC-DC Conversion
5
The Charge Pump
Capacitor-based design
Step-down: capacitors charged in series and discharged in parallel Conversion ration = 1 / number of parallel capacitors
Low currents
Slide6Implementation ExamplesKatja Klein
Serial Powering vs. DC-DC Conversion6
PP
with DC-DC conversion:
Serial powering:
Atlas pixels, Tobias Stockmanns
Stefano Michelis, TWEPP2008
Two-stage system Diff. technologies proposed for the two stages
Analogue and digital power fully separated
Power for optical links ~ integrated
HV not integrated
Regulators on-chip or on the hybrid
AC-coupled communication with off-module
electronics
Power for optical links not integrated
HV not integrated
Slide7What Conversion Ratio do we need?Katja Klein
Serial Powering vs. DC-DC Conversion7
Total tracker current estimate
Current strip tracker: 15kA; current pixel: 1.5kA Geoffs Strawman: strips: 25kW/1.2V = 21kA; pixels: 3.2kA; trigger layers: 10kA
Currents increase roughly by factor of 2 in this strawman Power loss in cables
Goes with I2 increase by factor of 4 for same number of cables (2000) T
otal power loss inverse proportional to number of power groups Can compensate with (conversion ratio)2 Material budget
Saving in cable x-section scales with I Total material independent of segmentation
Of course want to reduce as much as possibleConversion ratio needed for parallel powering with DC-DC converters?With conversion ratio of ¼ we would be as good as or better than today.SP: current fixed; cable material & power loss depends only on # of cables!
Slide8Power Losses in CablesKatja Klein
Serial Powering vs. DC-DC Conversion8
Consider
system
with n modules: P
det = n
I0V0
Voltage drop on cables & power loss P
cable calculated within each scheme Efficiency = Pdet
/ Ptotal = Pdet
/ (Pdet + P
cable)
Power losses in cables lead to decrease of overall power efficiency
expensive
... increase the heat load within the cold volume cooling capacity must be higher
SP
DC-DC,
r = 1/10
DC-DC, r = 1/5
Serial powering
Eff
. increases with n.
Since
10-20
modules can be chained
,
efficiency
can be very high!PP with DC-DC
conversion
E
ff
. goes down with n. Need
more
cables
or lower
conversion
ratio
Equal to SP if
conversion ratio
=
1/n
Slide9Local EfficiencyKatja Klein
Serial Powering vs. DC-DC Conversion9
Serial powering
Constant current source total power consumption is contant!
Current of chain is fixed to highest current needed by any member Current not used by a module flows
through shunt regulator Linear regulator: voltage difference between dig. & analog drops across it
Local power consumption is increased! Estimated increase for - Atlas pixels (NIM A557): 35%
- Atlas strips (NIM A579, ABCD): 18%
PP with DC-DC conversion All DC-DC converters have inefficiencies switching losses ESR of passive components
Ron of transistor etc. Typical values (e.g. comm. buck): 80-95%
Efficiency goes down for low conv. ratio! Trade-off betw. eff. & switching frequency
In two-step schemes, efficiencies multiply Estimates (St. Michelis, TWEPP2008):
Step-1: 85-90% Step-2: 93%
Total: 80-85% This needs to be demonstrated
Slide10Compatibility with LIC Cables
Katja Klein
Serial Powering vs. DC-DC Conversion
10PP
with DC-DC conversion 30V is largely enough
For any reasonable segmentation and conv. factor currents should be lower e.g. 20 chips a 53mA per module 1.2A / module
20 modules per rod 24A /rod r = ¼ I = 6A looks compatible
Serial powering
Current is small 30V allows for chains with more than 20 modules looks compatibleConstraints from recycling of current services:
2000 LICs with two LV conductors & common return each Not realistic to split return to obtain 4000 lines
Stay with 2000 LV power lines (“power groups“) LV conductors certified for 30V and 20A Twisted pairs (HV/T/H/sense) certified for 600V
256 PLCC control power cables Adapt at PP1 to (lower mass) cables inside tracker
Slide11Power SuppliesKatja Klein
Serial Powering vs. DC-DC Conversion11
PP with
DC-DC conversion
Standard PS: ~15V, ~10A (radiation & magnetic field tolerant?) Any sensitivity of converter to input voltage ripple?
No sensing needed (local regulation)?
Serial powering Constant current source Not so common in industry (e.g. CAEN) Atlas: PSs developed by Prague group
(developed already their current PSs) No sensing
Assume that power supplies will be exchanged after 10 years
Slide12Bias VoltageKatja Klein
Serial Powering vs. DC-DC Conversion12
PP
with DC-DC conversion
Same options as for SP
Serial powering Not yet well integrated into concept
Derive on-module via step-up converters? In Atlas, piezo-electric transformers are discussed. Or independent delivery using todays cables
Power is not a problem (currents are very low)
Up to now: independent bias lines for 1-2 modules Might not be possible anymore when current cables are re-used Note: T/H/sense wires are equal to HV wires
Slide13Safety (I)Katja Klein
Serial Powering vs. DC-DC Conversion13
PP with
DC-DC conversion
Open connections Converter itself can break Shorts between converter and module If PP of several mod.s by one converter:
risk to loose several modules at once
Serial powering Open leads to loss of whole chain Shunt regulators/transistors to cope with this
Several concepts are on the market (next page) Connection to module can break
bypass transistor on mothercable - high V, high I rad.-hardness? - must be controlable from outside Real-time over-current protection? Real time over-voltage protection?
Fermilab expressed interest to perform a systematic failure analysis
Slide14Safety (II)Katja Klein
Serial Powering vs. DC-DC Conversion14
One shunt regulator + transistor per module
+ no matching issue- no redundany- needs high-current shunt transistor- must stand total voltage
One reg. per module + distributed transistors
+ no matching issue+ some redundancy- feedback more challenging
Shunt regulators + transistors parallel on-chip
(Atlas pixels)
+ redundancy- matching issue at start-up Regulator with lowest threshold voltage conducts first all current goes through this regulator
spread in threshold voltage and internal resistance must be small
Slide15Slow ControlKatja Klein
Serial Powering vs. DC-DC Conversion15
PP
with DC-DC conversion
Slow control IC or block on hybrid For on-chip charge pump: would be useful to have SC information
from individual chips Could be used to set converter output voltage and switch on/off converters
Serial powering Slow control IC or block on hybrid
Could be used to communicate with linear regulator and turn to stand-by Ideas to sense module voltage in
Atlas pixels: - sense potential through HV return - sense through AC-coupled data-out termination - sense from bypass transistor gate Module voltage(s) Module current(s)?
Bias current
Slide16Start-up & Selective PoweringKatja Klein
Serial Powering vs. DC-DC Conversion16
PP
with DC-DC
conversion If converter output can be switched on/off, then easy and flexible: - controls can be switched on first - bad modules (chips?) can be switched off
- groups of chips/modules can be switched on/off for tests This should be a requirement!
Serial powering If controls powered from separate line,
it can be switched on first Devices in chain switched on together (both module controller and FE-chips)
Can take out modules only by closing bypass transistor from outside
Slide17ScalabilityKatja Klein
Serial Powering vs. DC-DC Conversion17
Serial powering
Current is independent on # of modules
Number of modules reflected in maximal voltage within chain; relevant for capacitors for AC-coupling
constant current source bypass / shunt transistors
PP with DC-DC conversion
If one converter per module: perfect scalability PP of
several mod. by one converter: current depends on # of modules, must be able to power largest group Should specify soon what we need current per chip # of chips per module
# of modules per substructure Otherwise we will be constraint by currents that devices can provide
Consequences if more modules are powered per chain or in parallel? E.g. barrel vs. end caps: different # of modules per substructure
FlexibilityKatja Klein
Serial Powering vs. DC-DC Conversion18
PP
with DC-DC conversion
If one converter per module: very flexible, do not care! If PP of several modules by one converter: distribution between modules arbitrary
Serial powering
Current of chain is equal to highest current needed by any member chains with mixed current requirements are inefficient!
Flexibility with respect to combination of devices with different currents E.g. trigger vs. standard module (or 4 / 6-chips)
Slide19Potential to Provide Different VoltagesKatja Klein
Serial Powering vs. DC-DC Conversion19
PP
with DC-DC
conversion With charge pumps, only integer conversion ratios are possible With inductor-based designs, arbitrary Vout
< Vin can be configured (but feedback circuit optimized for a certain range)
Only hard requirement: Vin >= Vopto Analogue and digital voltage can be supplied independently
no efficiency loss
Serial powering Needed voltage created by regulators ~1.2V by shunt regulator Lower voltage derived from this via linear regulator efficiency loss Technically could power opto-electronics
and controls via own regulators, but inefficient to chain devices with different current consumption D
ecouple from chain (Atlas: plan to power separately from dedicated cables)
Chip supply voltage(es): ~ 1.2V (Atlas: 0.9V for digital part to save power) Opto-electronics supply voltage: 2.5 – 3V
Slide20Process Considerations & Radiation HardnessKatja Klein
Serial Powering vs. DC-DC Conversion
20
Serial powering Regulators must be rad.-hard
Standard CMOS process can be used; but... HV tolerant components (up to nU0):
- capacitors for AC-coupling - bypass transistor Shunt transistors must stand high currents (~2A) if one per module
PP with
DC-DC conversion Commercial devices are not rad.-hard
Apparent exception: Enpirion EN5360 (S. Dhawan, TWEPP2008) Standard 130nm CMOS: 3.3V maximal For high conversion ratio transistors must tolerate high Vin , e.g. 12V Several “high voltage“ processes exist
Rad.-hard HV process not yet identified This is a potential show stopper
For r = ½ (e.g. charge pump) can use 3.3V transistors - radiation hardness?
Slide21Interplay with FE-ChipKatja Klein
Serial Powering vs. DC-DC Conversion21
Serial powering
Several options for shunt - Regulator and transistor on-chip
- Only shunt transistor on-chip - Both external Linear regulators typically on-chip Next Atlas strip FE-chip (ABCnext):
- linear regulator - shunt regulator circuit - shunt transistor circuit Next Atlas pixel chip (FE-I4):
- Shunt regulator - LDO DC-balanced protocol
PP with DC-DC
conversion Ideally fully decoupled Not true anymore in two-step approach with on-chip charge pump Next Atlas strip FE-chip (ABCnext): - linear regulator to filter switching noise
Next Atlas pixel chip (FE-I4): - LDO - Charge pump (r = ½) No influence on protocol
Slide22Readout & ControlsKatja Klein
Serial Powering vs. DC-DC Conversion22
PP
with DC-DC conversion
Nothing special: electrical transmission of data and communication signals to control ICs No DC-balanced protocol needed
Serial powering
Modules are on different potentials AC-coupling to off-module electronics needed Decoupling either on the hybrid
(needs space for chips & capacitors) or at the end of the rod (Atlas strips, P. Phillips, TWEPP08) Needs DC-balanced protocol
increase of data volumeAtlas pixels, NIM A557
Slide23NoiseKatja Klein
Serial Powering vs. DC-DC Conversion23
Serial powering
Intrinsically clean - current is kept constant
- voltages generated locally Main concerns: - pick-up from external source - pick-up from noisy module in chain Tests by Atlas pixels (digital) and strips
(binary) revealed no serious problems - noise injection - modules left unbiased - decreased detection thresholds - external switchable load in parallel to one
module (changes potential for all modules): some effect (Atlas pixels, NIM A557)
PP with DC-DC conversion
Switching noise couples conductively into FE Radiated noise (actually magnetic near-field) is picked up by modules Details depend on FE, distances, filtering, coil type & design, switching frequency, conversion ratio, ... Shielding helps against radiated noise,
but adds material, work and cost LDO helps against conductive noise, but reduces efficiency
Surprises might come with bigger systems Not good to start already with shielding and system-specific fine-tuning
Slide24Material BudgetKatja Klein
Serial Powering vs. DC-DC Conversion24
Serial powering
Regulators ~ one add. chip per hybrid
Components for AC-coupling - HV-safe capacitors (might be big!) - LVDS chip Flex for discrete components
Cable cross-section from PP1 to detector (rest stays) scales with current - One cable must carry I0 - Total mass depends on modules / cable
Motherboard/-cable: power planes can be narrow, small currents & voltages created locally
PP with DC-DC
conversion Converter chip(s) Discrete components - air-core inductor (D = 1-2cm!) - output filter capacitor(s) Flex for discrete components
One cable must carry I0nr total mass depends only on conv. ratio
Motherboard/-cable - buck converter can tolerate certain voltage drop since input voltage must not be exact low mass
- charge pumps have no output regulation: need exact Vin
Shielding?
Slide25SpaceKatja Klein
Serial Powering vs. DC-DC Conversion25
Serial powering
Different options are discussed, but regulators + shunt transistors are either
in readout chip or in a separate chip ~ one additional chip per hybrid Components for AC-coupling - LVDS buffers - HV-safe capacitors (might be big)
Bypass transistor?
PP with DC-DC conversion
Charge pump in readout chip or in a separate chip
Buck converter: - controller chip - discrete air-core inductor (D = 1-2cm!) - discrete output filter capacitor(s) - more? very unlikely to be ever fully on-chip In all other inductor-based topologies more components (inductors!) needed
Slide26Test Systems for Construction PhaseKatja Klein
Serial Powering vs. DC-DC Conversion26
PP
with DC-DC
conversion Electrical readout of single modules possible with adapter PCB needed
Serial powering If AC-coupling at end of stave, a decoupling board is necessary to read out single modules
Adapter PCB needed anyway for electrical readout
Slide27RWTH Aachen (L. Feld)
– proposal acceptedSystem
test measurements with commercial and custom DC-DC (buck) converters
Simulation of material budget of powering schemesRad.-hard magnetic-field tolerant buck converter
in collaboration with CERN groupBristol university (C. Hill)
– proposal accepted
Development of PCB air-core toroid DC-DC converter designs with air-core
transformerPSI (R. Horisberger)
– no proposal, but private communicationDevelopment of on-chip CMOS step-down converter (charge pump)IEKP Karlsruhe (W. de Boer) – proposal under review
Powering via cooling pipesFermilab / Iowa / Mississippi (S. Kwan)
– proposal under review System test measurements focused on pixel modules (DC-DC conversion & SP)
Power distribution simulation software
Katja Klein
Serial Powering vs. DC-DC Conversion
27
Work on Powering within CMS Tracker
Slide28Both schemes have their pros and cons – how to weigh them?
SP is complicated, but I do not see a real show stopper DC-DC conversion is straighforward, but two potential show stoppersnoise, radiation-hardness of HV-tolerant process
Need to understand SP better
In particular safety, slow controlsUp to now, we focus on DC-DC
conversion – should we start on SP? Who?Both Atlas pixels and strips integrate power circuitry in their new FE-chips: shunt regulators, charge pump, LDO
Seems to be a good approach - can we do the same?
Katja Klein
Serial Powering vs. DC-DC Conversion
28Summary