Transcript R cable

B2GM, 14 March 2012
DC/DC Converters
Markus Friedl (HEPHY Vienna)
What’s It All About?
 We need to bring power to the front-end
 There is a significant power drop in the cables
 Limited space and material budget
 Low voltages required by modern electronics
 In CMS and ATLAS Trackers, ~50% of total power (tens of kW) is
burnt in the cables
I
Rcable
Rload
Ploss = 2 x Rcable x I
I
Rcable
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
2
How to Avoid This?
 Lower Rcable: hardly possible 
 Serial powering: deliver constant current at higher voltage,
needs shunt regulator for each serially connected unit
 DC/DC converters: deliver higher voltage, needs converter
for each parallel unit
 Both methods reduce I and thus power loss at constant Rcable
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
3
Serial Powering vs. DC/DC Converters
 Both options have pros and cons and are intensively studied
for S-LHC upgrades
 Serial powering:
 no common ground: requires level translation for I/O
 generally easier for digital output – ATLAS
 DC/DC converters:
 switching device: requires careful design to avoid noise
 generally easier for analog output – CMS
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
4
DC/DC: How Does It Work?
 Buck converter (aka step-down converter):
Essentially a kind of pulse width modulation of primary
voltage, smoothened by LC filter
 L, C: the bigger, the better
 Higher switching frequency allows lower L,C values
T1 open, T2 closed
T1 closed, T2 open
S
+
Vin
−
+
D
Duty cycle D = t1,on/T;
C
Vout
−
1/D = Iout/Iin = Vin/Vout = r
Markus Friedl (HEPHY Vienna): DC/DC Converters
In reality, both
switch and diode
are MOSFET
transistors
(K.Klein, Aachen)
14 March 2012
5
DC/DC: Reality is Not Ideal
 Efficiency is <100% (but much better than linear regulator)
 Losses occur in several places:
 Every coil has an Ohmic resistance
 Ohmic resistance when MOSFET switch is on
 Capacitive switching losses  frequency
 Control circuit also needs power
 Typical DC/DC converters for our application have
 Vin/Vout = 2..10, Iout = O(1A)
 f = 1..5 MHz, switch Ron = O(10mW)
 L = a few 100nH, coil RL = a few 10mW
 Efficiency = 50..90%
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
6
DC/DC Challenges in HEP
 DC/DC converters are very common (e.g. PC power supply), but
we have some special issues in the front-end of HEP experiments
 Radiation
 Commercial integrated circuits are usually not rad-hard,
they typically die after 10…100 kRad
  Needs customized, rad-hard ASICs
 Magnetic field
 Presence of static B field saturates coil core
  Needs air coil  low L  high switching frequency
 Material budget
  Needs small, light-weight materials
 Noise emissions (conductive and radiated)
  Needs proper circuit design, layout and shielding
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
7
R&D at CERN and Aachen
 All those issues (previous page) are addressed, there are no
showstoppers
 CMS (and ATLAS) Tracker group have been studying the use
of DC/DC converters for future S-LHC upgrade
 CERN Microelectronics group is developing rad-hard DC/DC
converter ASICs and air coils
 Intensive tests with existing Tracker modules (APV25) and
rad-hard DC/DC converter prototypes at Aachen
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
8
ASIC: AMIS Family
 Being developed since 2007
 AMIS5 now being produced, expected back in summer 2012
AMIS4 (almost final)
2.56 x 2.87 mm2
0.35µm ON Semiconductor
(S. Michelis, CERN)
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
9
AMIS4 Measurements
Efficiency vs. Iout
80…85%
Markus Friedl (HEPHY Vienna): DC/DC Converters
(S. Michelis, CERN)
Efficiency vs. TID
Rad-hard beyond 100Mrad
Also tested: no SEB or SEGR
14 March 2012
10
CERN DC/DC Module
 Final AMIS chip not yet available, but substitute is
 CERN group developed a DC/DC module (SM01C) based on



13.5mm

commercial converter chip (LTC3605) with similar
performance as AMIS, but not rad-hard
Optimized design & layout SM01C Converter Module
Shielded air core coil
Available for tests
Chip will later be replaced
by rad-hard AMIS ASIC
28.4mm
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
11
Tests with DC/DC Converter and APV25
 Performed in Aachen (K. Klein et al) – references: appendix
TEC petal
2 Converters
(2.5V and 1.25V)
very close to
APV25 chips
Ring 6 modules
6.4
6.3
6.2
6.1
Motherboard
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
12
Impact on APV25 Noise (2009)
• Noise of (old) AC1
converter increased
with conversion ratio r
= Vin / Vout
• AC2-StandardC with
Mini Toroid and -filter
exhibits no significant
additional noise for all
accessible conversion
ratios
(K. Klein, Aachen)
 No converter
AC1 (2008)
 AC2-StandardC with Mini Toroid
AC2-StandardC with Mini Toroid + filter 2
 APV25 + DC/DC work together without noise penalty! 
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
13
Belle II SVD
 Originally (TDR), we have forseen rad-hard voltage
regulators close to the front-end (“Junction box”=DOCK box)
Finesse Transmitter
Board (FTB)
FADC+PROC
~2m Junction ~10m
copper
box
copper
cable
cable
Unified optical
data link (>20m)
COPPER
1748
APV25
chips
Front-end
hybrids
Rad-hard
voltage
regulators
Analog level translation,
data sparsification and
hit time reconstruction
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
Unified Belle II
DAQ system
14
Motivation for DC/DC @ Belle II SVD
 Due to financial constraints, we have to re-use Kenwood
power supplies of Belle I (SVD2)
 Problem:
 Belle I: 864 VA1TA chips, ~200W (+ repeater power)
 Belle II: 1748 APV25 chips, ~600W
 Existing Kenwood supplies are not sufficient for linear
regulators & APV25, but more than enough with DC/DC
 Now designing junction box with CERN DC/DC modules
 Similar space requirements, less cooling needed @ DOCK
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
15
Summary & Outlook
 Cable losses can be reduced by serial powering or DC/DC
 DC/DC is baseline for CMS Tracker upgrade
 CERN & Aachen are developing
 Rad-hard ASIC, air core coils, shields, circuit and layout
 APV25 + DC/DC converter: no noise penalty
 Now designing Junction box with CERN DC/DC modules
 Existing Kenwood power supplies are enough for future SVD
 I will meet CERN guys in late April for discussion
 Anyone (besided SVD) interested in DC/DC converters?
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
16
Some References
 Wikipedia: http://en.wikipedia.org/wiki/Buck_converter
 CERN Microelectronics: http://project-dcdc.web.cern.ch
 K.Klein (Aachen):


https://indico.cern.ch/contributionDisplay.py?contribId=56&confId=49682 (2009)
https://indico.cern.ch/contributionDisplay.py?contribId=47&confId=104062 (2011)
 F. Faccio, S.Michelis (CERN):



https://indico.cern.ch/contributionDisplay.py?contribId=92&confId=83060 (2010)
http://indico.cern.ch/getFile.py/access?contribId=39&sessionId=7&resId=0&materialId=
slides&confId=113796 (2011)
https://indico.cern.ch/contributionDisplay.py?contribId=21&confId=120853 (2011)
Markus Friedl (HEPHY Vienna): DC/DC Converters
14 March 2012
17