Paul Scherrer Institute - PowerPoint PPT Presentation

About This Presentation
Title:

Paul Scherrer Institute

Description:

Title: Personalrekrutierung: Kandidaten werden bez glich verschiedener Faktoren auf ihre Eignung untersucht Author: Mac04 Last modified by: Stefan Ritt – PowerPoint PPT presentation

Number of Views:120
Avg rating:3.0/5.0
Slides: 35
Provided by: Mac2164
Category:

less

Transcript and Presenter's Notes

Title: Paul Scherrer Institute


1
Paul Scherrer Institute
Stefan Ritt
Limiting factors in Switched Capacitor
ArraysSampling speed, Timing accuracy, Readout
speed
2
Follow-up Optimal Sampling Speed
Theory (Nyquist) 1 GHz signal 350 ps rise-time,
2 GSPS
350 ps
Threshold
500 ps
Reality Noise! (e.g. quantization noise of ADC)
and tail of input power
3
Measured Resol. Mod724, 14 bit, 100 MS/s
5T
50 mV
StdDev (ns)
100 mV
200 mV
500 mV
C. Tintori
4
Switched Capacitor Array
0.2-2 ns
Inverter Domino ring chain
IN
Waveform stored
Out
FADC 33 MHz
Clock
Shift Register
Time stretcher GHz ? MHz
5
Limits on sampling speed
  • vs. technology

6
Inverter chain
RC-delay with TG
Starved inverters
  • Layout more compact
  • Used in DRS4 chip
  • TG in signal path
  • Parasitics of TG counts
  • Used in most designs
  • Starving trans. outside signal path
  • Parasitics do not count

7
Achievable sampling speeds
Cpar 5 fF Transmission Gates Starved Inverters
0.25 UMC 3.7 GHz 13.0 GHz
0.25 UMC GAA 5.9 GHz
0.13 IBM 2.1 GHz 17.9 GHz
0.11 UMC 3.8 GHz 20.3 GHz
0.11 UMC HS 4.1 GHz 23.0 GHz
DRS4
  • Starved inverters better than TG
  • Speed does not linearly scale with
    technology(parasitics limited)
  • High speed (low Vt) option helps

8
Today highest sampling speed
130 IBM, J.-F. Genat, Clermont-Ferrand, Jan. 2011
9
Interleaved sampling
  • Fine tuning delays STURM chip (Gary) O(100
    GSPS)
  • For fixed interleaving, this can also be achieved
    by chip layout
  • Alternative Comparators with different thresholds

10
Limits on analog bandwidth
  • Parasitics, bond wires, Ron of sampling cell

11
Signal Chain
  • Detector (covered in next talks)
  • Connector (LEMO connector has a BW of 500 MHz)
  • Cable (RG58 5 m has a -3db BW of 1 GHz)
  • PCB
  • Preamplifier
  • Chip package
  • On-chip bus
  • Analog cell switch
  • Storage capacitor

PCB
Chip
Det.
Cpar
12
Influence on chip package
  • Bond wire has 2-3 nH and thus limits the BW to
    2-3 GHz
  • Input inductance can be reduced by using bump
    bonding or stud bonding

Stud
Bump
200 mm
Wire
75 mm
13
Effect of write bus
Length 3500 u Widths 4x8u, 4x14u (beginning/end
of bus)
DRS3 300 MHz with 2mm width
14
Influence on parasitics
  • Minimal write switch has 10 fF parasitic
    capacitance
  • Write bus has resistance of 0.05 Ohm/square
    (0.013 Ohm square for 20k top metal option)? 15
    Ohm after 3 mm bus bond wire (1.5 Ohm)? 10 pF
    after 3 mm

cellsbus length 32 cells 256 cells 1024 cells 64k cells
0.1 mm 130 GHz 30 GHz 8 GHz 0.13 GHz
1 mm 7.7 GHz 4.6GHz 1.9 GHz 0.04 GHz
3 mm 1.1 GHz 0.87 GHz 0.52 GHz 0.015 GHz
DRS3 300 MHz with 2u width
DRS4
15
Influence of write switch
-3db Bandwidth GHz
wopt. 6 um
  • Write switch has a finite on resistance
  • Storage cap needs to be gt10 fF forreasonable kTC
    noise
  • Leakage current requires even bigger C
  • Simulation
  • Cstore 50 fF
  • UMC 0.25 um technology
  • Vdd 2.5V
  • Minimal l
  • W 0.25 um N
  • Note Ngt1 adds parasitic to write bus!

16
Effect on sampling capacitor
  • Smaller Cstore leads to higher bandwidth
  • But kTC noise ? 20 fF for 11 bits
  • Practical limit 5 fF
  • Important Leakage current!
  • Worse with smaller technologies
  • Non-Gaussian distribution on chip
  • Worse for low Ron switch
  • Temperature dependent

G. Varner
17
Leakage current
  • Leakage current Must be small to get DVltlt1mV
    during readout
  • Distribution has long tail
  • Either make C large orkeep storage time short

G. Varner, 2010, Krakov
18
Comparison between technologies
UMC 0.25 7 kW 6 mm opt. 16 GHz
UMC 0.11 5 kW 180 mm opt. 21 GHz
UMC 0.11 low Vt 4 kW 120 mm opt. 37 GHz
dI/dU
Ron kW
VDS V
BW GHz
N/1000
19
Bandwidth DRS4 (1024 sampling cells)
  • Bandwidth is determined by bond wire and
    internalbus resistance/capacitance
  • 850 MHz (QFP), 950 MHz (QFN), ??? (flip-chip)

finalbus width
QFP package
850 MHz (-3dB)
Simulation
Measurement
20
Bandwidth STURM2 (32 sampling cells)
G. Varner, Dec. 2009
21
Optimal Chip Layout
Bond Pad
write

32 sampling cells
write-
Bond Pad
22
Limits on timing resolution
  • Matching PLL phase jitter Aperture

23
Typical SCA PLL
  • Matching (inverter-to-inverter variation by
    statistical limits in doping) is fixed over time
    and can be corrected
  • PLL phase jitter is typical 25 ps can can be
    corrected for with separate timing channel (DRS4
    81 channels)
  • Residual cell jitter caused by Vdd noise, short
    delay line is better

Inverter Chain
sampling speed control
PLL
T
Q
Phase Comparator
F1
loop filter
up
down
External Reference Clock
F2
24
Residual aperture jitter
  • Vdd (GND) noise causes jitter
  • Effect worse if rise time is slow (starving)
  • Typical values
  • 100 ps rise time for 1.2 V signal
  • 5 mV noise
  • 32 cells
  • Jitter 5 mV/1.2 V 100 ps 32 13 ps
  • Noise can originate off-chip (e.g. running ADC)
  • Solution Differential inverters, LDO on chip
  • Disadvantage More power

Noise
Timing
25
Limits on readout speed
  • Analog-Digital readout, multi-buffer

26
Readout time
N input channels
M output channels
treadout N/M nsamples tsample
Analog tsample 20 100 ns (external ADC 10-50
MHz) Digital tsample 5 10 ns nbits /
nlines 1024 samples, 10 bits, N8, M1 ?
treadout 400 ms 32 samples, 10 bits, N8, M8 ?
treadout 1.6 ms
27
ROI readout mode in DRS4
normal trigger stop after latency
delayed trigger stop
Trigger
stop
Delay
33 MHz
readout shift register
Patent pending!
28
Multi buffering
  • Multi-buffering can reduce dead time for
    Poisson-distributed events

Event occurring during readout of first event is
stored in second buffer
Event is stored in first buffer
R 1 kHz, T400 ms
R event rate Hz T readout time s LT
live time N Number of buffers
N LT
1 67
2 94
3 99.2
4 99.9
Cumulative distribution function for
Poisson-distributed events
29
Storage Depth
  • Has to accommodate trigger delay
  • High energy physics experiments require 100s of
    buffered events
  • CMS Possible hit every bunch crossing at 25 ns,
    155 bunch crossings before L1 trigger
  • ILC 3000 bunch trains 5 Hz
  • TOF-PET gt MHz event rate

CTA
? Deep storage depth ? Many storage segments ?
High event rate
ILC
30
The vision for the future
Low power
Low number of input cells
The Perfect Chip
Highchanneldensity
Deep Sampling Depth
Highevent rate
Manyanalogbuffers
Highevent rate
31
Cascaded Switched Capacitor Arrays
shift register
input
  • 32 fast sampling cells at 10 GSPS
  • 100 ps sample time, 3.1 ns hold time
  • Hold time long enough to transfer voltage to
    secondary sampling stage with moderately fast
    buffer (300 MHz)
  • Shift register gets clocked by inverter chain
    from fast sampling stage

. . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . .
fast sampling stage
secondary sampling stage
32
Typical Waveform
33
Dead-time free acquisition
  • Self-trigger writing of 128 short 32-bin
    segments (4096 bins total)
  • Simultaneous writing and reading ofsegments
  • Quasi dead time-free
  • Data driven readout
  • Ext. ADC runs continuously
  • ASIC tells FPGA when there is new data
  • Possibility to skip segments ? analogbuffer for
    HEP experiments
  • Coarse timing from300 MHz counter
  • Fine timing by waveformdigitizing and analysis
    in FPGA
  • 20 20 ns 0.4 ms readout time? 2 MHz
    sustained event rate (ToF-PET)
  • Attractive replacement for CFDTDC

DRS5planned for 2013
34
Conclusions
  • SCAs will more and more replace Q-ADC and
    CFGTDCs
  • New designs are in the pipeline for gt3 GHz analog
    BW, multi-buffering and fast readout
  • Current limitations areare well known and will
    bepushed further in nextgeneration of chips
Write a Comment
User Comments (0)
About PowerShow.com