L. Ratti a,c , M. Manghisoni b,c , V. Re b,c , V. Speziali a,c , G. Traversi b,c

28
CMOS processes in the 100 nm minimum feature size range for applications to the next generation collider experiments L. Ratti a,c , M. Manghisoni b,c , V. Re b,c , V. Speziali a,c , G. Traversi b,c c INFN Sezione di Pavia b Università degli Studi di Bergamo Dipartimento di Ingegneria Industriale a Università degli Studi di Pavia Dipartimento di Elettronica

description

CMOS processes in the 100 nm minimum feature size range for applications to the next generation collider experiments. L. Ratti a,c , M. Manghisoni b,c , V. Re b,c , V. Speziali a,c , G. Traversi b,c. a Università degli Studi di Pavia Dipartimento di Elettronica. - PowerPoint PPT Presentation

Transcript of L. Ratti a,c , M. Manghisoni b,c , V. Re b,c , V. Speziali a,c , G. Traversi b,c

Page 1: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

CMOS processes in the 100 nm minimum feature size range for applications to the next

generation collider experiments

L. Rattia,c, M. Manghisonib,c, V. Reb,c, V. Spezialia,c, G. Traversib,c

cINFN Sezione di Pavia

bUniversità degli Studi di Bergamo Dipartimento di Ingegneria Industriale

aUniversità degli Studi di Pavia Dipartimento di Elettronica

Page 2: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

2VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Use of CMOS microelectronic processes in the last two decades has had a deep impact on the way HEP instrumentation is conceived

Motivation

Quarter micron technology able to comply with the challenging design requirements of the LHC experiments in terms of

noise figure

power dissipation

radiation tolerance

Luminosity and track densities expected at the next generation colliders (LHC upgrade, ILC, Super B-Factory) set the demand for increased spatial resolution, denser functional packing, higher radiation hardness and better noise/power trade-off

HEP people is considering moving to more scaled CMOS processes

Technology monitoring to

keep design criteria and methodologies up to date

fight process obsolescence

study scaling down effects on the main design parameters

Page 3: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

3VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Investigated technologies and devices

Single transistors from HCMOS9 130 nm and CMOS090 90 nm triple well, epitaxial CMOS technologies by STMicroelectronics

Technology features:– VDD = 1.2 V– tOX= 2 nm– COX=15 fF/μm2

Available geometries– W = 200, 600, 1000 μm– L = 0.13 - 1 μm

Technology features:

– VDD = 1 V

– tOX= 1.6 nm

– COX=18 fF/μm2

Available geometries– W = 100, 200, 600, 1000 μm– L = 0.1 – 0.7 μm

HCMOS9 (Lmin=130 nm) CMOS090 (Lmin=90 nm)

Devices under test are PMOS and NMOS transistors with standard open layout

Page 4: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

4VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Operating regionDrain current in DUTs: from tens of A to 1 mA low power operation as in high density front-end circuits

Characteristic normalized drain current I*Z may provide a reference point to define device operating region

2TOX

*Z nVC2I

• μ carrier mobility

• COX specific gate oxide capacitance

• VT thermal voltage

• n proportional to ID(VGS) subthreshold characteristic

1

10

100

10-9 10-8 10-7 10-6 10-5

gm

/ID [

1/V

]

IDL/W [A]

CMOS 130 nm

CMOS 90 nm

PMOS

NMOS

I*Z,P,130 I*

Z,P,90

I*Z,N,130

I*Z,N,90

Weak inversion lawStrong inversion law

Page 5: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

5VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Operating region

LW

I

IIC

*Z

D0

Inversion coefficient

At the considered drain currents, DUTs work in weak or moderate inversion region

At a given drain current, operation is shifted towards weak inversion region with technology scaling

0.01

0.1

1

10

100

0.1 1

nmos L=0.13 mnmos L=1.00 mpmos L=0.13 mpmos L=1.00 m

Inv

ersi

on

Co

eff

icei

nt

Drain Current [mA]

Strong inversion

Moderate inversion

Weak inversion

90 nm techW=600 m

10-7

10-6

1 1.5 2 2.5 3 3.5

NMOSPMOS

I* z [A

]

Physical tOX

[nm]

90 nm

180 nm

130 nm

STMicroelectronics

Page 6: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

6VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Transconductance

At small ID (weak inversion), gm fairly independent of the device dimension and polarity

In weak inversion, possible difference between PMOS and NMOS and between CMOS nodes only due to different n values (n1.25 for both polarities and technologies considered here)

T

Dm nV

Ig

0

0.01

0.02

0.03

0 0.0002 0.0004 0.0006 0.0008 0.001

Tra

nsc

on

du

ctan

ce [

A/V

]

Drain Current [A]

NMOSW=600 mV

DS=0.6 V

L=0.35 m

130 nm tech

90 nm tech L=0.13 m

L=0.35 m

0

0.01

0.02

0.03

0 0.0002 0.0004 0.0006 0.0008 0.001

Tra

nsc

on

du

ctan

ce [

A/V

]

Drain Current [A]

90 nm techW=600 m|V

DS|=0.6 V

NMOS

PMOS

L=0.70 m

L=0.35 m

Page 7: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

7VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Noise in CMOS transistors

Noise in the drain current of a MOSFET can be represented through an equivalent noise voltage source in series with the device gate

(f)SS(f)S 21/f

2W

2V

SW - white noise

• channel thermal noise (main contribution in the considered operating conditions)

• other contributions from parasitic resistances

n

g

T4kS

W

m

B2ch ,

• kB Boltzmann’s constant

• T absolute temperature

• αw excess noise coefficient

• γ channel thermal noise coefficient

S1/f - 1/f noise

• technology dependent contribution

• both kf and αf depends on the polarity of the DUT

fWLfC

k(f)S

OX

f21/f

• kf 1/f noise parameter

• αf 1/f noise slope-related coefficient

Page 8: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

8VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

1

10

100

103 104 105 106 107 108

CIN

= 6 pF

ID = 100 A

W/L = 2000/0.45, 0.25 um processW/L = 1000/0.5, 0.13 um processW/L = 600/0.5, 0.09 um process

Noi

se V

olta

ge

Sp

ectr

um

[nV

/Hz1/

2 ]

Frequency [Hz]

NMOS

Noise in different CMOS generations

250 nm TSMC

130 nm STM

90 nm STM

Page 9: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

9VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

1

10

100

103 104 105 106 107 108

L=0.13 m

L=0.35 m

L=1.00 m

No

ise

Vo

lta

ge

Sp

ect

rum

[n

V/H

z1/2 ]

Frequency [Hz]

130 nm techW=1000 mID=0.25 mA

VDS

=600 mV

NMOS

Noise vs gate length – STM 130 nm

High frequency, white noise virtually independent of the gate length L, in agreement with gm behavior

1/f noise contribution decreases with increasing channel length, as predicted by the noise equation

1

10

100

102 103 104 105 106 107 108

L=0.13 m

L=0.35 m

L=1.00 m

No

ise

Vo

ltag

e S

pec

tru

m [

nV

/Hz1

/2]

Frequency [Hz]

130 nm techW=1000 mID=0.25 mA

|VDS

|=600 mV

PMOS

Page 10: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

10VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Noise vs drain current - NMOS

0.1

1

10

100

103 104 105 106 107

Id=0.10 mAId=0.25 mAId=1.00 mA

No

ise

Vo

ltag

e S

pe

ctru

m [

nV

/Hz

1/2 ]

Frequency [Hz]

STM 90 nm

NMOSW/L=600/0.2V

DS=600 mV

0.1

1

10

100

103 104 105 106 107 108

Id=0.10 mAId=0.25 mAId=1.00 mA

No

ise

Vo

lta

ge

Sp

ect

rum

[n

V/H

z1/2 ]

Frequency [Hz]

STM 130 nm

NMOSW/L=1000/0.35V

DS=600 mV

High frequency, white noise decreases with increasing drain current in both technologies, in agreement with gm behavior

1/f noise contribution is to a large extent independent of the drain current

Page 11: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

11VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Noise vs drain current - PMOS

0.1

1

10

100

102 103 104 105 106 107

Id=0.10 mAId=0.25 mAId=1.00 mA

No

ise

Vo

lta

ge

Sp

ect

rum

[n

V/H

z1/2 ]

Frequency [Hz]

STM 90 nm

PMOSW/L=600/0.2|V

DS|=600 mV

0.1

1

10

100

102 103 104 105 106 107

Id=0.10 mAId=0.25 mAId=1.00 mA

No

ise

Vo

ltag

e S

pec

tru

m [

nV

/Hz1

/2]

Frequency [Hz]

STM 130 nm

PMOSW/L=1000/0.35|V

DS|=600 mV

High frequency, white noise decreases with increasing drain current, more markedly so in the 90 nm technology

1/f noise contribution increases with increasing current, more significantly in the STM 130 nm technology

Page 12: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

12VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Noise and inversion region

1

10

100

103 104 105 106 107 108

STM 90 nm

STM 130 nm

No

ise

Vo

lta

ge

Sp

ect

rum

[n

V/H

z1/2]

Frequency [Hz]

NMOSW/L=600/0.35ID=0.1 mA

VDS

=600 mV

1

10

100

103 104 105 106 107 108

STM 90 nm

STM 130 nm

No

ise

Vo

ltag

e S

pec

tru

m [

nV

/Hz1

/2]

Frequency [Hz]

NMOSW/L=600/0.35ID=1 mA

VDS

=600 mV

At low drain current both devices work in the weak inversion region channel thermal noise is roughly the same for both devices

Better 1/f noise performance provided by the STM 90 nm technology

At high drain current, a significant difference in the channel thermal noise can be detected device from the 90 nm technology works closer to weak inversion region

Page 13: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

13VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

0

50

100

150

200

250

300

0 50 100 150 200 250 300

Eq

uiv

ale

nt

No

ise

Res

ista

nce

[

n/gm

[

90 nm techNMOS L>0.13 m

Linear fitoffset = 1.68 +/- 1.45slope = 0.96 +/- 0.02

Channel thermal noise – STM 90 nm

Equivalent channel thermal noise resistance

slope excess noise coefficient w

w close to unity no sizeable short channel effects in the considered operating regions (no data available for channel thermal noise in devices with L ≤0.13 m)

Negligible contributions from parasitic resistances

offset noise contributions from parasitic resistors

T4k

SR

B

2W

Th

Page 14: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

14VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

0

50

100

150

200

250

300

0 50 100 150 200 250 300

Eq

uiv

ale

nt

No

ise

Res

ista

nce

[

n/gm

[

130 nm techNMOS L > 0.13 mLinear fitoffset = 6.85 +/- 1.90slope = 1.01 +/- 0.02

0

50

100

150

200

250

300

0 50 100 150 200 250 300

Eq

uiv

ale

nt

No

ise

Res

ista

nce

[

n/gm

[

130 nm techPMOSLinear fitoffset = 1.82 +/- 2.12slope = 0.97 +/- 0.02

Channel thermal noise – STM 130 nm

αw close to unity no sizeable short channel effects in the considered operating regions also for STM 130 nm technology (except for NMOS with L=0.13 m)

Negligible contributions from parasitic resistances in NMOS devices, larger in PMOS transistors possibly due to different doping levels used in critical layers

Page 15: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

15VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Flicker noise

1

10

100

103 104 105 106 107

NMOSPMOSN

ois

e V

olt

age

Sp

ectr

um

[n

V/H

z1/2]

Frequency [Hz]

90 nm techW/L=600/0.2ID=1 mA

|VDS

|=600 mV

f=0.84

f=1.12

Slope f of the 1/f noise term is significantly smaller than 1 in NMOS transistors and larger than 1 in PMOS devices

Page 16: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

16VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

0.6

0.8

1

1.2

1.4

1.6

0 0.2 0.4 0.6 0.8 1 1.2

f

Drain Current [mA]

PMOS

NMOS

130 nm tech

0.6

0.8

1

1.2

1.4

1.6

0 0.2 0.4 0.6 0.8 1 1.2

f

Drain Current [mA]

PMOS

NMOS

90 nm tech

Slope coefficient f

In the examined operating region, f does not exhibit any clear dependence on the drain current or on the channel length

f between 1 and 1.3 for PMOS devices, between 0.8 and 1 for NMOS devices

Page 17: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

17VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Slope coefficient f

Very similar behavior of f was detected through different CMOS generations and different foundries

0.6

0.8

1

1.2

1.4

1.6

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

PMOS

NMOS

f

Minimum Feature Size [m]

STM 90 nm

STM 130 nm

STM 180 nm

STM 350 nm

TSMC250 nm

Page 18: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

18VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

1/f noise coefficient kf vs gate length

In the case of the 130 nm technology, short channel devices (L<0.5 m) exhibit a flicker noise coefficient larger than for NMOSFETs with longer channels

The same behavior concerns devices with L<0.2 m in the case of the 90 nm technology

0

5

10

15

20

25

30

35

40

0 0.2 0.4 0.6 0.8 1 1.2

Id=0.10 mAId=0.25 mAId=0.50 mAId=0.75 mAId=1.00 mA

Kf

[J 1

0 -2

5 H

z (a

lph

a-1

) ]

Channel Length [m]

NMOS W=1000 m

STM 130 nm

0

5

10

15

20

25

30

35

40

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Id=0.10 mAId=0.25 mAId=0.50 mAId=0.75 mAId=1.00 mA

Kf

[J 1

0 -2

5 H

z (a

lph

a-1

) ]

Channel Length [m]

NMOSW=600 m

STM 90 nm

Page 19: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

19VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

1/f noise coefficient kf vs Vov

0

20

40

60

80

100

120

-0.05 0 0.05 0.1 0.15 0.2

k f [J

10 -2

5 H

z (a

lph

a-1

) ]

Gate Overdrive Voltage [V]

90 nm

130 nm

PMOS

0

20

40

60

80

100

120

-0.05 0 0.05 0.1 0.15 0.2

k f [J

10 -2

5 H

z (a

lph

a-1

) ]

Gate Overdrive Voltage [V]

NMOS

PMOS

STM 130 nm

In PMOS devices, flicker noise coefficient is clearly bias dependent (dependence is weaker in STM 90 nm technology)

In NMOS transistors kf is to a large extent independent of the overdrive voltage VOV

Page 20: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

20VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

60Co -rays effects on device performances

0.13 µm technology (open-structure layout)

ID increase in the

subthreshold region of NMOS: edge effects due

to radiation-induced charge at the shallow trench isolation (STI)

oxide. The effect is larger in devices with a shorter channel, affecting ID

regions of interest for low-power applications

(ID = 100 A). 10-10

10-8

10-6

10-4

10-2

-1.2 -0.8 -0.4 0 0.4 0.8 1.2

NMOS

I D [

A]

VGS

[V]

PMOS

W/L = 1000/0.5

before irradiation

10 MRad

Page 21: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

21VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

In short-channel NMOS, at low ID (around 100 A) 1/f noise increases by a

much larger extent than at higher drain currents.This may be correlated with the ID increase in the subthreshold region,

meaning that the shallow trench oxide contributes in determining the 1/f noise properties of irradiated open-structure devices.

0.1

1

10

100

103 104 105 106 107 108

before irradiation10 Mrad

Noi

se V

olta

ge S

pect

rum

[nV

/Hz

1/2 ]

f [Hz]

NMOSID

= 100 A

W/L = 1000/0.35

0.1

1

10

100

103 104 105 106 107 108

before irradiation10 Mrad

Noi

se V

olta

ge S

pect

rum

[nV

/Hz

1/2 ]

f [Hz]

NMOSID

= 1 mA

W/L = 1000/0.35

60Co -rays effects on device performances

Page 22: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

22VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Conclusions

Static, signal and noise measurements have been performed on devices belonging to two different CMOS technology nodes, namely the 130 nm and 90 nm STM processes

Channel thermal noise equations developed to describe the device behavior in the considered operating regions provide a reliable model, with short channel effect playing a minor role in both the considered processes

1/f noise results confirm the behavior detected in previous submicron processes as far as the dependence on device polarity and bias and gate geometry is concerned

Extracted noise parameters show that using the 90 nm process may ensure an improvement in the noise performances in applications where large signal dynamic range is not needed while miniaturization can be an asset

Characterization of the 90 nm technology will be completed with radiation hardness tests (open structure vs enclosed layout, study of possible STI effects)

Page 23: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

23VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Backup slides

Page 24: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

24VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Channel thermal noise coefficient

WI

LIu ,u

3

2

2

1

u1

1*z

D

Page 25: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

25VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Low noise charge preamplifier design

f

f

1pOX

ff2

pm

B1gD t

1

WLC

kA2

t

1

g

T4kACCENC

• CD detector capacitance

• CG preamplifier input capacitance

• tp peaking time

• A1 A2 shaping coefficients

Circuit designers can take advantage of single device characterization to predict noise behavior of charge sensitive amplifiers

Data extracted from single transistor characterization can be used to plot minimum ENC as a function of the main design parameters (peaking time, power dissipation, polarity and dimensions of the preamplifier input device)

Equivalent noise charge is the figure of merit to be minimized:

Channel thermal noise contribution

Flicker noise contribution

It is interesting to assess whether (and if so to what extent) using a more scaled technology may improve noise performances

Page 26: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

26VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

100

1000

10 102 103

Op

tim

um

EN

C [

e- r

ms]

Peaking time [ns]

STM 90 nm

CD=20 pF

PD.max

=0.25 mW

L=0.20 m

NMOS

PMOSSTM 130 nm

ENC vs peaking time

In the explored peaking time and power range, PMOS input device always provides better noise performances than NMOS input (except for the 130 nm process at tp close to 10 ns)

Using the 90 nm process may yield quite significant improvement with respect to the 130 nm technology, especially when NMOS input charge preamplifiers are considered

ENC was evaluated in the case of a second order, unipolar (RC2-CR) shaping processor

Page 27: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

27VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

400

600

800

1000

3000

5000

0.01 0.1 1

STM 90 nm

STM 130 nm

Op

tim

um

EN

C [

e- r

ms]

Dissipated power [mW]

PMOSC

D=20 pF

L=0.20 mtp=20 ns

ENC vs dissipated power

At tp=20 ns, noise performances provided by NMOS and PMOS input devices in the 90 nm technology are comparable

400

600

800

1000

3000

5000

0.01 0.1 1

STM 90 nm

STM 130 nm

Op

tim

um

EN

C [

e- r

ms]

Dissipated power [mW]

NMOSC

D=20 pF

L=0.20 mtp=20 ns

Better noise-power trade-off can be achieved by using the 90 nm technology

Page 28: L. Ratti a,c , M. Manghisoni b,c , V. Re b,c ,   V. Speziali a,c , G. Traversi b,c

28VI International Meeting on Front-End Electronics, Perugia, May 18th 2006

Transconductance in all inversion regions

WI

LIu ,

14u1

2

nV

Ig

*z

D

T

Dm