Download Mixed-Signal ASIC Design for Space Communications - amicsa

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Mixed-Signal ASIC Design for
Space Communications
Presented by Dr. Rajan Bedi
ESA AMICSA 2006
Slide 1 of 34
Mixed-Signal ASIC Design for Space
Communications
UK EXPORT CONTROL SYSTEM EQUIPMENT & COMPONENTS RATING:
3A001a1a, 3A001a2, 3A001a5a1, 3A001a5a2, 3A001a5a3, 3A001a5a4,
3A001a5b, 3A101a, 3C001a, 3C001b, 4A002b2, 4A002b1 & 5E001c1.
UK EXPORT CONTROL TECHNOLOGY RATING: 3E001, 4E001 5E001b1 &
5E001c2e.
Rated By : Rajan Bedi with reference to UK Export Control Lists (version
INTR_A12. DOC 13 August 2003) which contains the following caveat:
“The control texts reproduced in this guide are for information purposes
only and have no force in law. Please note that where legal advice is
required, exporters should make their own arrangements”.
Export licence : Not required for EU countries. Community General Export
authorisation EU001 is valid for export to : Australia, Canada, Japan,
New Zealand, Norway, Switzerland & USA.
ESA AMICSA 2006
Slide 2 of 34
Presentation Overview
 Introduction & Motivation
 Research into mobile payload ADCs
 Research into broadband payload ADCs
 Research into space-grade DACs
 Mixed-Signal Processing
 Conclusion
ESA AMICSA 2006
Slide 3 of 34
Introduction
 Increasing amounts of analogue circuitry and digital logic
are now being integrated on die and flown.
 The availability of:
– Low-power, high-performing BiCMOS technologies, fT (peak
cut-off frequency) ~ 300 GHz!
– The inherent total-dose tolerance of the SiGe HBT.
– The addition of a SOI layer to SiGe BiCMOS.
– The ability to individually size of NPNs and PNPs with trench
isolation.
– Integrated EDA tools and common design & verification flows.
 All of these offer the potential to advance mixed-signal
microelectronics for space applications.
ESA AMICSA 2006
Slide 4 of 34
Motivation
 The benefits of integrating analogue circuitry with digital
logic include:
 Higher performance
 Lower power consumption
 Less mass
 Reduced costs
 Improved reliability
 Greater levels of reusability
 Enhanced system testing and quality
 All of these benefits contribute to the spirit of “Faster,
Better, Cheaper”!
ESA AMICSA 2006
Slide 5 of 34
Motivation
 Telecommunication satellites with on-board digital
processors require ADCs at the receiver to digitize
the IF/baseband carrier information.
 Future mobile missions will be required to digitize
information bandwidths around 50 MHz – this will
necessitate baseband sampling at a rate greater
than 100 MSPS.
 Future broadband missions will be required to
digitize information bandwidths around 500 MHz –
this will necessitate baseband sampling at a rate
greater than 1 GSPS.
ESA AMICSA 2006
Slide 6 of 34
Motivation
 This performance will only be delivered by advances in
mixed-signal processing.
 Experience has shown that accessing the latest deep sub-
micron technologies is difficult due to the low volumes
required by space companies.
 When the major semiconductor vendors are willing to
develop hardware, the NRE costs are exorbitant.
 Moreover, the suitability of the fabrication technology for
space flight first needs to be assessed, with vendors often
adopting a low-risk approach to radiation testing and
qualification.
 As a satellite manufacturer, how do we support the
development of future digital payloads for our customers?
ESA AMICSA 2006
Slide 7 of 34
Motivation
 This variation in the range of sampling frequencies
greatly impacts the micro-architecture, the circuit
design, the resolution and power consumption of mixedsignal converters that can be considered for use on
satellite payloads.
 This range of payload types, limited power budget,
restricted fabrication options and operation in a harsh
environment, all combine to make ADC/DAC design for
space communications extremely challenging.
 This objective of this presentation is to share with
vendors current research on low-power, highperformance, space-grade converters for use with
future mobile and broadband payloads.
ESA AMICSA 2006
Slide 8 of 34
ADC Design
 The
design of low-power ADCs capable of
sampling around 150 MSPS differs from those
required to sample above 1 GSPS.
 The development of an ADC involves a number of
trade-offs, e.g. as the required sampling frequency
or dynamic range increase, so does the
complexity, size, weight, cost and power
dissipation of devices.
 As a general rule, effective resolution decreases
by one bit for every doubling of the sampling
frequency.
ESA AMICSA 2006
Slide 9 of 34
Low-Power Narrowband ADC
 For low-power ADCs, a sampling frequency of
150 MSPS is achieved by implementing microarchitectures that combine elements of the
successive approximation and flash techniques
with pipeline-type signal processing.
ESA AMICSA 2006
Slide 10 of 34
Pipelined ADC Micro-Architecture
First stage of the pipeline performs a coarse quantisation on
the input while subsequent stages are concurrently
processing previously acquired input samples.
ANALOGUE PIPELINE
INPUT
STAGE 0
STAGE 1
…
STAGE N-1
+
+
LATCH

STAGE N
+
LATCH
…
+

+
LATCH
+

DIGITAL PIPELINE
ESA AMICSA 2006
Slide 11 of 34
Pipelined ADC Micro-Architecture
Each stage comprises a S&H, low-resolution
ADC and DAC and a subtractor.
SAMPLE & HOLD
+
VIN

A
--
ADC
ESA AMICSA 2006
DAC
Slide 12 of 34
Pipelined ADC Micro-Architecture
 High-throughput, concurrent design.
 For pipelined ADCs, power dissipation can be
minimised through the appropriate choice of perstage resolution and selection of suitable op-amp
architectures.
 Thermal noise, comparator offsets, sample-and-
hold offsets, gain errors and non-uniform levels
within the interstage DACs all affect the
performance of pipelined ADCs.
 Simulations and circuit design are continuing that
investigate trade-offs between thermal noise,
speed, linearity and power dissipation.
ESA AMICSA 2006
Slide 13 of 34
Pipelined ADC Micro-Architecture
 For low power operation, CMOS converters are
preferred as digital consumption scales with
sampling frequency.
 Bipolar converters developed to operate at a
particular frequency but used at lower rate, will
incur a fixed penalty and dissipate the same power
as clocking at the higher frequency due to
constant current structures.
ESA AMICSA 2006
Slide 14 of 34
Pipelined ADC Micro-Architecture
 The digital capability of CMOS easily allows the
inclusion of error correction, calibration and other
computational features to improve overall
converter performance and accuracy.
 Fabrication using deep submicron CMOS results
in higher performance due to reductions in
parasitics.
 However, concomitant with the migration to
smaller feature sizes, lower supply voltages make
it more difficult to maintain input SNR – Continual
technology scaling impacts ADCs!
ESA AMICSA 2006
Slide 15 of 34
Low Power Broadband ADC
 For low-power broadband ADCs, sampling
frequencies above 1 GSPS are achieved by
implementing interleaving micro-architectures or
variations of Flash techniques:
· Folded Flash
· Folded & Interpolative Flash
· Pipelined, Folded & Interpolative Flash
ESA AMICSA 2006
Slide 16 of 34
Folded Flash ADC
Folding is a technique that reduces hardware while maintaining
the one-step nature of a full Flash ADC.
An analogue pre-processing circuit generates a residue which
is digitised to obtain the LSBs.
The MSBs are resolved using a coarse ADC that operates in
parallel with the folding circuit.
INPUT
FOLDING
CIRCUIT
COARSE
ADC
FINE
ADC
DIGITAL ADDER &
CORRECTION CIRCUIT
DIGITAL OUTPUT WORD
ESA AMICSA 2006
Slide 17 of 34
Folded Interpolative Flash ADC
 Interpolation reduces the number of comparator
preamplifiers at the input of a Flash converter.
 Interpolation substantially reduces the input
capacitance, power dissipation and area of flash
converters, while preserving the one-step nature of
a Flash architecture.
 Simulations show that offset, gain and timing
mismatches result in SNR degradation at high
input frequencies.
ESA AMICSA 2006
Slide 18 of 34
DAC Design
 We have started investigating current-steering
DAC micro-architectures with the goal of achieving
low power with high speed and low distortion.
 Traditionally current-steering DACs were
fabricated using bipolar technology, however, the
ability to generate matched CMOS current mirrors
will result in lower power consumption.
 A segmented architecture provides a good
balance between performance vs. area and
complexity.
ESA AMICSA 2006
Slide 19 of 34
VLSI
 To achieve the required performance and low power,
silicon-based technologies are being targeted for
fabrication.
 CMOS converters are preferred as digital power
dissipation scales with sampling frequency.
 SOI CMOS typically offers 30 to 50% higher
performance for the same power compared to bulk
CMOS (30% less power at the speed).
 SOI offers reduced cross-talk for mixed-signal design,
immunity to latch-up, better tolerance of SEEs and the
absence of radiation-induced leakage between devices.
ESA AMICSA 2006
Slide 20 of 34
VLSI
 BiCMOS processes combine faster, higher-current
driving bipolar transistors with smaller, lower-power,
high impedance CMOS devices.
 This makes BiCMOS attractive for mixed-signal design.
 The bipolar transistor can be either Si or SiGe.
 For the same operating current, a SiGe HJT has higher
speed, increased gain, lower RF noise and less 1/f
noise compared to a Si BJT – ft ~ 300 GHz.
 SiGe BiCMOS allows the integration of analog, digital
and RF using existing CMOS foundries.
 Currently investigating SOI and BiCMOS to minimise
noise coupling through the wafer substrate.
ESA AMICSA 2006
Slide 21 of 34
Space Microelectronics
 Heavy-ion strikes can trigger latchup in CMOS and
bipolar devices with the potential to damage a circuit
permanently.
 In ADC/DACs, a transient generated in the analogue
part of a device can propagate into the digital section
causing logic level shifts.
 The migration to smaller geometries has helped CMOS
transistors to become inherently more resistant to TID
radiation as thinner gate oxide layers trap less positive
charge.
 Radiation hardening by design techniques are being
used to mitigate the damage, functional upsets and
data loss caused by radiation.
ESA AMICSA 2006
Slide 22 of 34
Radiation Hardening By Design
 The use of enclosed layout NMOS transistors.
 The design of portions of a circuit using PMOS FETs
only. As PMOS transistors are not prone to edge
leakage, there is no need for annular layout.
 Such structures are immune to latchup due to the
absence of stray silicon-controlled rectifier structures.
 Latchup can be avoided by completely enclosing NMOS
and PMOS FETS using guard rings or through the
addition of an epi layer.
 Design-hardened versions of integrated circuits typically
require more die space than their soft counterparts!
ESA AMICSA 2006
Slide 23 of 34
Analogue Section of 12-bit Segmented
DAC Incorporating RHBD
63 Unary Current Sources
ESA AMICSA 2006
6 Binary Current Sources
Slide 24 of 34
Radiation Hardening By Design
 All NMOS transistors drawn as Enclosed Layout
Transistors (Annular) – this makes them immune
to edge leakage effects.
 Guard rings isolate all n+ diffusions at different
potentials.
 Current sources are all PMOS.
 Solely PMOS structures are immune to latch-up.
ESA AMICSA 2006
Slide 25 of 34
PMOS Current Source – (LSB Binary)
VDD
bias_P1
bias_P2
W = 1.1 μm
L = 0.8 μm
M=1
W = 1.1 μm
L = 0.8 μm
M=1
ILSB
out
ESA AMICSA 2006
Slide 26 of 34
PMOS Current Source – (Unary)
VDD
W = 1.1 μm
L = 0.8 μm
M = 64
bias_P1
bias_P2
W = 1.1 μm
L = 0.8 μm
M = 64
64 * ILSB
out
ESA AMICSA 2006
Slide 27 of 34
Mixed Signal Processing
 A number of signal processing techniques
exist that potentially could ease the hardware
implementation of the data converters:
– Complex Baseband Sampling
– RF/IF Bandpass Sampling
ESA AMICSA 2006
Slide 28 of 34
Complex Baseband Sampling
Analogue Generation of I/Q Samples
xi(t)
i(t)
i[n]
A/D
x(t)
i[n] – jq[n]
-90°
xq(t)
q(t)
q[n]
A/D
sin ωct
Analogue
Digital
Complex baseband sampling at a rate greater than the highest
frequency component – allows access to lower sampling
frequencies – lower power if using CMOS ADCs.
Difficult to match frequency response of both paths
ESA AMICSA 2006
Slide 29 of 34
Complex Baseband Sampling
Digital Generation of I/Q Samples
x(t)
A/D
i[n]
i’[n]
q[n]
q’[n]
cos ωct
sin ωct
Analogue
Digital
Complex baseband sampling at a rate four times the
highest frequency component.
Single ADC, no matching issues, simple mixing
ESA AMICSA 2006
Slide 30 of 34
RF Bandpass Sampling of L-band User Uplink
Band-pass AAF
Amp.
Fs/2
70.25
Zone
1
Fs
140.5
2
2 Fs
281
1.5 Fs
210.7
3
4
11.5 Fs
1615.7
Frequency (MHz)
23
12 Fs
1686
24
The availability of wideband ADCs, e.g. bandwidths
approaching 3 GHz, makes direct sampling of L/S-band RF
carriers a real possibility.
Bandpass sampling achieves digitisation and
downconversion in a single operation, without the use of
analogue mixers, local oscillators and image-reject filters.
Intentionally alias the RF information to baseband/IF.
ESA AMICSA 2006
Slide 31 of 34
RF Bandpass Sampling
 To realise bandpass sampling, the ADC must have
the necessary analogue input bandwidth to process
the highest frequency component within the input
signal.
 Low distortion and good linearity at this frequency
are essential.
 ADCs intended for undersampling applications are
more sensitive to the amount of noise at the input
than a traditional converter.
 Board
layout, decoupling considerations and
minimising jitter on the sampling clock are critical.
ESA AMICSA 2006
Slide 32 of 34
Conclusions
 I have shared with you current investigations that research the
development of low-power, high-performance, space-grade
converters for use with future mobile and broadband payloads.
 Satellite manufactures have intimate knowledge of payloads.
 Different types of telecommunication satellites have unique
mixed-signal requirements, e.g. sampling frequency, resolution,
dynamic performance and power consumption.
 Pipelined ADCs should be targeted for future mobile payloads
and folding, interpolative flash architectures for broadband
processors.
 Segmented, current-steering DACs should be targeted for
future satellites.
ESA AMICSA 2006
Slide 33 of 34
Conclusions
 A re-usable design and verification flow and
design trade-offs at the systems-level, the
micro-architectural level, the circuit level and
at a technology level combine to enable
mission-specific, mixed-signal IP.
ESA AMICSA 2006
Slide 34 of 34