ultra low power analog integrated circuits for implantable...

F. Silveira, Univ. de la República, Montevideo, Uruguay IEEE CASS DLP 1

Ultra Low Power Analog Integrated Circuits

for Implantable Medical Devices

Fernando Silveira

Universidad de la República, Uruguay

CCC Medical Devices

nanoWattICs

F. Silveira, Univ. de la República, Montevideo, Uruguay IEEE CASS DLP 2 F. Silveira, Univ. de la República, Montevideo, Uruguay IEEE CASS DLP 2

Objectives of this talk

• Introduce the needs and characteristics of Active Implantable Medical Devices (AIMDs) from the circuit designer point of view.

• Present the techniques and circuits applied in Analog ULP

• Device Modeling

• Design Methodology

• Circuit techniques

• Show the current research and development topics and prospects of the area


Engineering School Universidad de la Republica

• Founded 1888, approx. 1k new students / year, 670 teaching staff


• Since 1991

• Under Graduate & Graduate Teaching (MSc, PhD)

• Research

• Design of Analog / RF and Mixed-Signal Integrated Circuits, particularly Ultra-Low Power (ULP)

• Also works on ULP Digital / DC-DC and Embedded

• Strong links with Research groups abroad

• Industrial Experience

• Implantable Medical Devices

• 2007: spin-off: NanoWattICs

Microelectronics Group


Implantable Devices in Uruguay

Feb. 3, 1960: Drs. O. Fiandra and R. Rubio

performed the first effective pacemaker implant to a human being in the world.

1969: Dr. O. Fiandra founded CCC to develop and manufacture pacemakers

1999: CCC develops a pacemaker line based on an DDDR ASIC designed by the Microelectronics Group of Universidad de la Republica

Today: CCC designs and manufactures active implantable devices and complete medical systems for third parties.


Outline

I. System: Active Implantable Medical Devices Today

II. Transistors and Circuits: Analog Design for ULP.

Transistor Modeling.

Design Methodology.

III. Circuit Techniques: Implementation of AIMDs blocks

IV. Conclusions and Prospects


Active Implantable Medical Devices (AIMD)

Implantable: Introduced inside the body by a medical procedure and intended to remain there after the procedure.

Active: Including a Power Source

Not considered here:

Passive implants (e.g. bone prostheses, valves, stents)

Wearable / Portable / Swallowable (!) Active Medical Devices


Cardiac Pacemaker: first implantable device, 1960

AIMDs: Main Historical Milestones (I)

Cochlear Implants (1960s -)


Cardiac Defibrillators (1980)

AIMDs: Main Historical Milestones (II)

Deep Brain Stimulator for

Parkinson (1995)


• Heart Failure

• Obesity

• Diabetes

• Neurostimulators:

• Pain control

• Blood pressure control

• Foot drop correction

• Urinary incontinence

• Sleep Apnea

• …

• Patient monitoring

• Brain – computer interface

AIMDs: Some of the new developments


• Pacemaker:

• Goal: Treat Bradycardia (slow heart rythm) and conduction disorders between atria and ventricles

• How: Stimulating to contract the heart when it does not contract spontaneously (“watchdog”)

• Sensing of:

• cardiac muscle signals that indicate ventricles / atria contraction

• other indicators of physical activity, additionally in some cases

Some system examples


• Stimulation (Open Loop)

• Early Pacemakers

• Cochlear Implants

• Deep Brain Stimulators for Parkinson

• Neurostimulators (sometimes “Man/Woman in the loop”)

• Stimulation and Sensing (Closed Loop)

• Cardiac area (Pacemakers, Defibrillators, Heart Failure)

• Obesity

• Some Neurostimulators

• Only Sensing

• Implanted “long term Holter” (“insertable loop recorder”)

• Sensing + external actuation: Brain-computer interface

Basic Functions


Stimulation: Voltage mode

• E.g.: Pacemakers

• 0.1V … 7.5V

• 50ms … 1.5ms

• Requires battery voltage multiplier.

• RL: 500 Ohms typ.


• Neurostimulators and others

• 0.1mA … 10mA

• 30ms … 300ms

• Load voltages up to 15V => Requires battery voltage multiplier

Stimulation: Current mode

t


Sensing: Medical signals in general

• Low frequency: from < 1 Hz to a few kHz (neural signals)

• Low amplitude: mV to mV

• Variability:

" Most measured quantities vary with time, even when all controllable factors are fixed. Many medical measurements vary widely among normal patients, even when conditions are similar “ (Source:J. Webster, Medical Instrumentation. Application and Design).

Objective of most analog signal processing: qualitative detection for closed loop control.

Traditionally advantage to analog implementation in terms of consumption, process scaling is changing this


• Biopotentials:

• mioelectric signals (mVs, 100s Hz - 1kHz)

• cardiac signals (mVs, 10s Hz – 300Hz)

• neural signals (mVs, up to 8kHz)

• Impedance (tens of mOhms => mVs, few Hz)

• Movement (Physical activity, position) => accelerometer (mVs (sensor dependent), up to 10Hz)

Sensing


• Telemetry

• Inductive (up to 10cms)

• 403 MHz MedRadio Band (a couple of meters)

• Battery Supervision (Voltage / Impedance / Consumed Charge Measurement)

• Lead Impedance Measurement

• Magnet Sensor (Reed Relay / Hall Sensor)

• Battery Recharge (if applicable)

• Control: Microcontroller & Firmware

Auxiliary Functions


Non-implantable System Components

Medical System Components

IPG Leads

Programmer

System

PSA

Patient

wand

Battery

charger

Logger


Example: Implantable Pacemakers

Stimulus

Sense Channel

Micro controller

Battery Supervision

Activity Sensing Lead

Selection (polarity)

Telemetry

Amplification, Filtering and Detection

Programmable Voltage Multiplier

0.1VDD a 2-3 VDD

35% / 6mA

30% / 5mA

18% / 3mA

17% / 2.8mA

Approx. Consumption Distribution


Example: Closed Loop Stimulator for Drop Foot Correction (I)

• Neurostep System (Simon Fraser Univ, Canada, Neurostream Technologies)

• Closed loop operation based on neural signal sensing and neural stimulation

• On clinical trials


Example: Closed Loop Stimulator for Drop Foot Correction (II)

http://www.youtube.com/watch?v=xH2vNu2BbnU


General Requirements: Size

Currently approximately 12 cc (5cm x 4cm x 0.6cm)

Approx. 30 to 40% occupied by the battery

Less consumption = Smaller size @ Equal Service Life

Biotronik

1968- 1998 (Source: M. Wilkinson, course: MST for Medical Devices)


• Lithium-Iodine Battery (Li/I2)

• THE pacemaker battery during almost 30 years

• Beguining of life: 2.8V, Operation down to 2.0V.

• High internal impedance, specially near depletion (several kohms)

• Must be taken into account in complete device (power decoupling) and circuit design (instantaneous consumption, PSRR)

• Common problem with other non-implantable batteries (e.g LiMn02)

• Capacity: In the order of 1 Ah = 114 mA . Year

General Requirements: Batteries (I)

0

0.5

1

1.5

2

2.5

3

0 20 40 60 80 100

Percentage of Nominal Capacity Delivered (%)

Voltage(V

dc)

0

2000

4000

6000

8000

10000

12000

14000

16000

Inte

rnal Im

pedance (

Ohm

s)

Voltage

Internal Impedance


• For higher instantaneous consumption devices, lower impedance required

• Lithium-Silver-Vanadium-Oxide (Li/SVO)

• Lithium-Carbon-Monofluoride (Li/CFx)

• QMR / QHR: Li + combined SVO / CFx

• Being applied also in pacemakers recently.

• For higher average consumption devices:

Rechargeable lithium batteries (since approx. year 2000, capacity in the order of 0.3Ah)

Direct powering from RF energy transmitted transcutaneously

General Requirements: Batteries (II)


General Requirements: Battery and Consumption

• Capacity: In the order of 1 Ah = 114 mA . Year

• Consumption

Service Life: 6 to 12 years => consumption: 19 mA to 9 mA

Average consumption due to stimulation (pacemakers): 3 mA to 12 mA=> Unavoidable

State of the art: average consumption internal to the circuit around 5 to 10mA

Consumption internal to the circuit: 50% to 75% of total consumption

There is room and need for improvement


General Requirements: Safety and Reliability

This is not acceptable !!!

Source: Quino



Reliability => Frequency and probability of faults

Safety: Involves many aspects, particularly:

=> A single fault must not provoke a catastrophic event

High Reliability => Probability of single fault is low and double fault is virtually impossible

+ Safety

=> Probability of malfunctioning is low

=> Catastrophic Failure: virtually impossible



Involves all the stages:

System and Circuit Design

System and Circuit Verification, Qualification and Medical Validation

Medical Device Application, Configuration and Use

Strongly conditions design: E.g. Limiting DC leakage towards the heart under single fault conditions => external capacitors

Importance of paying attention from the very beginning to applicable standards on AIMD safety, risk analysis and applicable regulations.


Outline




Design Methodology.




• Bipolar (BJT) Analog Design:

gm = Ic/UT

– Basically 1 degree of freedom: Ic

• MOS Analog Design:

gm = f(ID, W, L)

– 3 degrees of freedom

– Traditionally: only part of the design space: the strong inversion (above threshold) region

Introduction


-0.5 0 0.5 1 1.5 2 2.510

-14

10-12

10-10

10-8

10-6

10-4

10-2

VG(V)

ID(A

)

Subthreshold

Current

Leakage Current

VT0

-1 0 1 2 30

2

4

6

ID(m

A)

VG(V)

VT0

MOST Inversion Regimes (1)

Strong Inversion (S.I.)

ID(VG-VT)2

Weak inversion:

IDeVG/(n.UT)

Moderate inversion


0 0.5 1 1.5 2 2.5

10-10

10-5

VG(V)

ID(A

)

Diffusion Current

Drift Current

VT0

MOST Inversion Regimes (2)

Strong Inversion (S.I.)

ID(VG-VT)2

Weak Inversion (W.I.)

IDeVG/(n.UT)

UT=k.T/q

n: slope factor

Moderate Inversion (M.I.)


All regions, continuous MOST models

• EKV (Enz, Krummenacher, Vittoz, EPFL, AICSP 1995): originally mathematic interpolation between strong and weak inversion equations, now physical

• ACM (Advanced Compact Model, A. Cunha, C. Galup-Montoro, M. Schneider, UFSC, IEEE JSSC 1998): Physical model.

• … or experimental / simulation curves

0 0.5 1 1.5 210

-15

10-10

10-5

VG(V)

ID(A

) Strong Inversion

Approximate Limit

Weak Inversion

Approximate Limit

VT0

Weak Inversion

ModelStrong Inversion

Model

General Model

Leakage

Current


“Intrinsic” MOS Amplifier

A0=gm/gd A (dB)

f(Hz)

fT=gm/(2..CL)

V

vi

vo

CL

ID

DDV

vi

vo

CL

ID

DD

vi

vo

CL

ID

DD

D

d

D

mA

D

m

d

mom

I

g

I

gV

I

g

g

grgA ..0

T

0

0

L

mT

f2

s.A1

AA ,

C2

gf

• Consumption: ID

• Speed gm/CL

• CL: total: external + parasitics

• Speed - Consumption trade-off : gm/ID

OTA: Operational Transconductance Amplifier


gm/ID vs. VG

G

DGD

DD

m

V

)Ilog(VI

I

1

I

g

• gm/ID is the slope of ID vs. VG in log scale

Maximum in WI

Equal to 1/(n.UT)

n typ: 1.3 a 1.5 -0.5 0 0.5 1 1.5 2 2.5

10-14

10-12

10-10

10-8

10-6

10-4

10-2

VG(V)

ID(A

)

Subthreshold

Current

Leakage Current

VT0


gm/ID vs. ID

10-15

10-10

10-5

100

0

5

10

15

20

25

30

35

40

ID(A)

gm

/ID

(1/V

)

gm/Ic, transistor bipolar

• As the current increases, the “gm generation efficiency decreases”

• To reach the maximum frequency allowed by the technology:

=> high gm => high current => strong inversion => low efficiency

W/L =100 and 0.8mm technology.

Bipolar Transistor:

gm/IC

independent of current in a wide range


gm/ID and transistor size

ID=mCox(W/L).f(VG, VS, VD)

S

Dfnorm

ox

Dnorm

Dnormnorm

D

m

I

IiI

LWC

II

LW

IIf

I

g

)/(

)/( I

m

When short channel effects are significant: L,If

I

gnorm

D

m

When short channel effects are not significant:

10-10

10-8

10-6

10-4

10-2

0

5

10

15

20

25

30

ID/(W/L)(A)

gm

/ID

(1/V

)


Design Methodology: gm/ID key variable

0

10

20

30

40

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04

ID(A)

gm

/ID

(1

/V)

10-12

10-10

10-8

10-6

weak inv.

mod. inv.

strong inv.

1/nUT 2/GVO

Circuit Performance Transistor Operating Mode

Transistor Sizing

gm / ID

ID=mCox(W/L).f(VG, VS, VD)

g

If I

I

W L

II

C W L

m

Dnorm norm

D

normD

ox

I

( / )

( / )m

[Silveira, Flandre, Jespers, IEEE JSSC 1996, P.Jespers, Springer, 2010]

gm.vi gdVi

VO

CL

ID

VDD

CL

Vi VO

Ag

IV

g

C C

g

IIm

D

A

m

L L

m

D

D0

1

2

1

2 fT


W/L C gm

Optimum of Power Consumption

10-10

10-8

10-6

10-4

10-2

0

5

10

15

20

25

30

ID/(W/L)(A)

gm/ID

(1/V

)g

m/I

D (

1/V

)

ID/(W/L) (A)

Weak inversion: IDeVG/(n.UT)

Moderate inversion

Strong inversion (ID(VG-VT)2)

gm/ID ID

Usually an optimum exists in moderate inversion

• Working towards WI


Example Intrinsic Amplifier: Power Optimum

0 5 10 15 20 2510

-5

10-4

gm/ID (1/V)

ID (

A) 0 5 10 15 20 25

100

101

102

103

104

gm/ID (1/V)

W1 (

um

)

fT=10MHz, CL=3pF, L=2mm, tech: 0.8mm

V

vi

vo

CL

ID

DDV

vi

vo

CL

ID

DD

vi

vo

CL

ID

DD

0 5 10 15 20 252

4

6

8

10

12

gm/ID (1/V)

CLto

t (p

F)


Example Intrinsic amplifier 0.8um, CL 3pF

1.00E-08

1.00E-07

1.00E-06

1.00E-05

1.00E-04

1.00E-03

1.00E-02

1.00E-01

1.00E+00

0 5 10 15 20 25 30

gm/ID(1/V)

ID(A

)

100kHz

1MHz

10MHz

100MHz

1GHz

M.I. optimum

W.I. optimum

S.I. optimum

10-10

10-8

10-6

10-4

10-2

0

5

10

15

20

25

30

ID/(W/L)(A)

gm

/ID

(1/V

)g

m/I

D (

1/V

)

ID/(W/L) (A)


gm/ID in the nanometer and post CMOS era

10-11

10-10

10-9

10-8

10-7

10-6

10-5

10-4

10-3

0

5

10

15

20

25

30

35

ID/(W/L) (A)

gm

/ID

(1/V

) L=0.10,0.11,0.12,0.13,0.14

L=0.5L=1.0

L=4.0

W=10um,L(um)VDS=0.6V

B. Sensale-Rodriguez et al, accepted for presentation at IEEE SubVt 2012, joint work with U. Notre Dame, Indiana.

Derived based on data from P. Jespers, The gm/ID Methodology a sizing tool for low-voltage analog CMOS circuits, Springer, 2010, extras.springer.com


Outline




Design Methodology.




ULP Analog Signal Processing: 1) RC Active Filter

Limited by values possible to integrate: up to k o M (special process) and 100s pF

Large spread in absolute values of integrated components

=> Drawback: need for external components ..., but ....

Low pass filter

s.C.R1

RR

v

v

2

1

2

in

o

VoVin

+

-

C

R1

R2


ULP Analog Signal Processing: External components should always be avoided ?

A few are required to avoid single fault risks.

Full integration might be paid in terms of consumption in order to accomodate less precise components

1 Ah = 114 mA.year 11.4mA 44mm3/ 100nA

10 year service life 5cc battery

0805

0603

0201

0402 SMT components up to 10M and 15mF

0402 SMT => 1mm x 0.5mm x 0.35mm = 0.18mm3

Considering PCB, IC package pin, routing => 2 - 5 mm3

Size and Consumption are linked through Battery size


ULP Analog Signal Processing: 2) Switched capacitor filters

1/R2.C = fclk.C2/C

+ precise, + large time constants possible, + fully integrated

- op amps consumption, - antialiasing

VoVin

+

-C

R1

R2

VinVo Vin

Vo

+

-

R1

1

+

-

C1

2

21

VinVo

2

+

-

C11

C

2 1

1

C2 22

1


ULP Analog Signal Processing: 3) Gm-C Filters

v+

v-

+

-

gm io = gm(v+ - v-)

vinVo

+

-

gm

C

+-

gm

VoVin

+

-

C

R1

R2

R=1/gm => 1/R2.C = gm2/C => spread

=> on-chip automatic tuning

+ large time constants possible

- input linear range requirements for transconductors


Example of modules: Pacemaker Activity Sense

Objective:

E.g. Activity indicator: 3s Average of the absolute

value of acceleration in the 0.5 - 7 Hz band.

Sensor 3s Averaging

Amplifier / filter

Ideal Rectifier

Amplitude: tens to hundreds of mV


Accelerometer Signal Conditioning (1): Amplifier / Bandpass filter

High pass

characteristic

Vs

Vf

Input signal

Feedback signal

Double input symmetrical OTA (DDA)

Vo

Vo=A1Vs+A2Vf

ISCAS 1998


Accelerometer Signal Conditioning (2): Layout and results

Gain 2900

Equivalent input noise (mVrms)

18

Consumption (mA)

3.4

Area (mm2)

1.82


Accelerometer Signal Conditioning (3): Results

0

20

40

60

80

100

120

140

160

180

200

0

20

40

60

80

100

120

140

160

180

200

0 50 100 150 200

dig

itiz

ed

ou

tpu

t o

f cir

cu

it

ca

rd

iac freq

.(p

pm

)

time(s)

actual cardiac frequency of healthy patient

simulated pacemaker frequency

circuit output


Accelerometer Signal Conditioning (4) : Gm-C implementation

A. Arnaud (UR), C. Galup (UFSC), ISCAS 2004

FiltroFiltro--Amplificador 0.5Amplificador 0.5--7Hz7Hz

CC11=550p=550p

VVBiasBias==

700mV700mV CC22=50p=50p

VVOutOut11

++

GGm1m1

GGm2m2

GGm3m3

VVININ

SensorSensor

VVlinlin==5mV5mV

CC33=50p=50p

VVOut2Out2

++

GGm4m4

GGm5m5

GGm6m6

VVlinlin==500mV500mV

CC44=250p=250p

G=385G=385

Gm4=21nS

Gm5=2.5nS

Gm6=89pS

Ganancia 2a:

G2=8.3

Gm4=21nS

Gm5=2.5nS

Gm6=89pS

Ganancia Preamplificador:

G1=46.4

IDD= 290nA

Equivalent

Input noise:

2.1mVrms

Gain: 390

Fully

integrated


Example of modules: Neural Recording Amplifier

• Objective: Signal detection from e.g: cuff electrodes or cortical electrodes arrays

• Requirements:

• 0.5mVrms - 2mVrms noise

• BW: 300Hz – 8kHz

• High CMRR (particularly in Cuff)

• Block high DC offsets (100mV or more) due to electrode/tissue contact

• Negligible DC input current


Example: Cuff Electrode Recording in Neurostep

Hoffer et al, IFESS 2005

Hoffer et al, US Patent 5,824,027 / 1998


Example: Cortical Recordings

Harrison, Proc. IEEE, July 2008


Neural Amplifier Front End (1): Capacitive Feedback

• Inversion region for noise / power optimization: e.g. input pair weak inversion, current mirror active load: strong inversion

• CMRR limited by capacitor matching.

Harrison et al, IEEE JSSC, June 2003

Gain 40 dB

BW 0.13 Hz / 7.5 kHz

Itotal 16 mA

NEF 3.8

vnoise rms 2.1 mV

CMRR > 42 dB

MOS –Bipolar Pseudoresistor (100s Mohms equivalent)


Neural Amplifier Front End (2): DDA Based

J. Sacristán, T. Oses, IFESS 2002,

Another DDA Based Scheme: M. Baru, U.S. Patent 6.996.435, 2006

• High CMRR

(Given by Input Differential Pair)

• Both Differential

Pairs contribute equally to Input Noise (hence to area and consumption)


Neural Amplifier Front End (3): “Asymmetrical” DDA Based (I)

P. Castro, F. Silveira, ISCAS 2011

• Effect of noise (and

hence consumption and area) of Gm2 greatly reduced while keeping high CMRR (given by input differential pair)

• Gm2 less effective in compensating input offset and DC components => Output DC and high pass characteristic fixed by local feedback at the output


Neural Amplifier Front End (4): “Asymmetrical” DDA Based (II)

Spec Castro,

ISCAS 2011

Harrison,

JSSC 2003

Wattapanitch,

TR. BIOCAS

2007

Sacristan,

IFESS 2002

Architecture Asym. DDA Capacitive Capacitive DDA

A (dB) 48 40 41 80

NEF 4.2 3.8 2.7 53.4

Itotal(mA) 16.5 16.0 2.7 180

vi noise

(mVrms)

2.4 2.1 3.1 7.6

CMRR > 107 > 42 > 66 90


Outline




Design Methodology.




Prospects: Digital vs. Analog

• Theoretical limits of power consumption for Analog and Digital Signal Processing

• Analog better for low S/N, but the border is moving ...

• “Digital” pacemaker already present in marketing

Scaling

Source: E. Vittoz


Prospects: Analog ULP and AIMD

• Intense growth of applications / therapies on development and reaching the market

• Broad Analog / Circuit research area:

• Sensing

• Stimulation, Power Management / Battery Recharge, Communication, …

• Once very specific area, now wider (wireless sensor networks, body area networks, portable devices, energy scavenging devices, RFID, ...).


Prospects: AIMDs Brain Computer Interface

Set. 2000, Nicolelis, Duke University


Prospects: AIMDs Brain Computer Interface

Mar. 2002, M. Serruya, J. Donoghue, Brown University


Prospects AIMDs: Brain Computer Interface

July 2004: Pilot FDA trial started by spin/off company of Brown Univ., several tetraplegic patients implanted.


Some Conclusions

•ULP ICs for AIMDs: Each nA counts => Methodology and Optimization

•AIMDs: Very broad field in strong expansion Many R & D opportunities Microtechnology is often the enabling factor.

•AIMDs: Price is not the main concern, but application and performance Suitable for developments with lower volume

productions than in other areas High investment associated with long

development cycles, qualification, clinical testing and regulatory aspects.


Microelectronics Group – IIE Universidad de la República

More Information

iie.fing.edu.uy/vlsi

[email protected]

Acknowledgements

• IEEE CASS

• CCC Medical Devices, NanoWattICs

• Members (present and past) of Microelectronics Group, UR


Thank you !

ultra low power analog integrated circuits for implantable...

Documents