IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I...

24
IN3170/4170, Spring 2020 Philipp Häfliger [email protected] What to expect

Transcript of IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I...

Page 1: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

IN3170/4170, Spring 2020

Philipp Hä[email protected]

What to expect

Page 2: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Content

Why Application Specific Integrated Circuits?

Why Transistor Level Digital?

Why Analog?

Course Goal

Course Organization

Page 3: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Content

Why Application Specific Integrated Circuits?

Why Transistor Level Digital?

Why Analog?

Course Goal

Course Organization

Page 4: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do an ASIC?

Well, why not?I Costly (development, design iteration time, production)I Inflexible and low level of reusability

Alternatives?I Embedded SystemsI FPGA (pure digital)I Microcontroller (digital, mixed signal)

So why bother?I Ultimate performance (speed, power)I Ultimate miniaturizationI Reliability (fewer points of failure)I Very cheap for high volume production (e.g. CPUs)I For (Mixed-Signal) Systems-on-Chip (SoC)

Page 5: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do an ASIC?

Well, why not?

I Costly (development, design iteration time, production)I Inflexible and low level of reusability

Alternatives?I Embedded SystemsI FPGA (pure digital)I Microcontroller (digital, mixed signal)

So why bother?I Ultimate performance (speed, power)I Ultimate miniaturizationI Reliability (fewer points of failure)I Very cheap for high volume production (e.g. CPUs)I For (Mixed-Signal) Systems-on-Chip (SoC)

Page 6: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do an ASIC?

Well, why not?I Costly (development, design iteration time, production)I Inflexible and low level of reusability

Alternatives?I Embedded SystemsI FPGA (pure digital)I Microcontroller (digital, mixed signal)

So why bother?I Ultimate performance (speed, power)I Ultimate miniaturizationI Reliability (fewer points of failure)I Very cheap for high volume production (e.g. CPUs)I For (Mixed-Signal) Systems-on-Chip (SoC)

Page 7: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do an ASIC?

Well, why not?I Costly (development, design iteration time, production)I Inflexible and low level of reusability

Alternatives?

I Embedded SystemsI FPGA (pure digital)I Microcontroller (digital, mixed signal)

So why bother?I Ultimate performance (speed, power)I Ultimate miniaturizationI Reliability (fewer points of failure)I Very cheap for high volume production (e.g. CPUs)I For (Mixed-Signal) Systems-on-Chip (SoC)

Page 8: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do an ASIC?

Well, why not?I Costly (development, design iteration time, production)I Inflexible and low level of reusability

Alternatives?I Embedded SystemsI FPGA (pure digital)I Microcontroller (digital, mixed signal)

So why bother?I Ultimate performance (speed, power)I Ultimate miniaturizationI Reliability (fewer points of failure)I Very cheap for high volume production (e.g. CPUs)I For (Mixed-Signal) Systems-on-Chip (SoC)

Page 9: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do an ASIC?

Well, why not?I Costly (development, design iteration time, production)I Inflexible and low level of reusability

Alternatives?I Embedded SystemsI FPGA (pure digital)I Microcontroller (digital, mixed signal)

So why bother?

I Ultimate performance (speed, power)I Ultimate miniaturizationI Reliability (fewer points of failure)I Very cheap for high volume production (e.g. CPUs)I For (Mixed-Signal) Systems-on-Chip (SoC)

Page 10: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do an ASIC?

Well, why not?I Costly (development, design iteration time, production)I Inflexible and low level of reusability

Alternatives?I Embedded SystemsI FPGA (pure digital)I Microcontroller (digital, mixed signal)

So why bother?I Ultimate performance (speed, power)I Ultimate miniaturizationI Reliability (fewer points of failure)I Very cheap for high volume production (e.g. CPUs)I For (Mixed-Signal) Systems-on-Chip (SoC)

Page 11: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Content

Why Application Specific Integrated Circuits?

Why Transistor Level Digital?

Why Analog?

Course Goal

Course Organization

Page 12: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do a Digital ASIC?

See previous arguments for and against ASICThe most important is the small price per piece for high volumeproduction particularly for large scale systems-on-chip (SoC), e.g.CPU, but also FPGAs, GPUs, Microcontrollers etc. (mostly not ’fullcustom’ design but automated ’synthesis’), but real understandingon a single transistor level is required for the ultimate performancein speed, miniaturization, power consumption. Analogous in SW ofwhere it’s worth to program in Assembler.

Page 13: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Why do a Digital ASIC?

See previous arguments for and against ASICThe most important is the small price per piece for high volumeproduction particularly for large scale systems-on-chip (SoC), e.g.CPU, but also FPGAs, GPUs, Microcontrollers etc. (mostly not ’fullcustom’ design but automated ’synthesis’), but real understandingon a single transistor level is required for the ultimate performancein speed, miniaturization, power consumption. Analogous in SW ofwhere it’s worth to program in Assembler.

Page 14: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Content

Why Application Specific Integrated Circuits?

Why Transistor Level Digital?

Why Analog?

Course Goal

Course Organization

Page 15: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

The world is analog

Analog electronics for sensor/actuator interfaces

⇔⇔ user⇔

Page 16: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Ubiquitous Sensors Interfaces

Trend to ‘Cyberphysical Systems’

1970

1980

1980

2000

2010

5

Computational Infrastructure• Stationary/backend• Wired• High end computing

Mobile access devices• Human interaction• Portable • Mostly wireless• Battery

Sensory swarm• Miniature• Wireless• Autonomous/self-contained• Controlling and sensing natural processes

1

10

>100

Driven by Moore’s Law

Beyond Moore

Page 17: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Even Computers are Analog ;-)

Where the digital abstraction breaks down

• Gates• Increasing speed

– Why this degradation?– How do we improve

performance?– Digital → analog

• Going for speed…

• Noise/interference

– Where does this noise originate?– How do we reduce this noise/interference?– Digital → analog

• When scaling down size and scaling up complexityTSL inf3410

10

100MHz

1GHz

Page 18: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Content

Why Application Specific Integrated Circuits?

Why Transistor Level Digital?

Why Analog?

Course Goal

Course Organization

Page 19: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Course Goal (1/3)

B

Y

A)

A

A B

C

C

V+ V

-

Vb2

Vb3

Vb1

Vout

B)

Understand these two circuits thoroughly!

Understand: analysis, properties, applications, limitations,tweaking, high level descriptions ...

Page 20: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Course Goal (1/3)

B

Y

A)

A

A B

C

C

V+ V

-

Vb2

Vb3

Vb1

Vout

B)

Understand these two circuits thoroughly!Understand: analysis, properties, applications, limitations,tweaking, high level descriptions ...

Page 21: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Course Goal (2/3)

And thereby understand the basic building blocks of analog anddigital circuits:

B

A

CY

A)

V+

V-

+

-Vout

B)

Y = ¬((A ∧ B) ∨ C )

Vout = A(V+ − V−) (1)

Page 22: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Course Goal (3/3)

... starting from modelling the basic active element of CMOSelectronics, the field effect transistor (FET)

G

S

D

G

S

D

G

S

D

I=g (V -V )m G SR={ 0 if V > V

if V < Vswitch

G

G

switch

G

D

S

R={ 0 if V > V

if V < Vswitch

G

G

switch

nFET symbol Digital Abstractions Analog Linear Abstraction

G

D

R={ 0 if V > V

if V < Vswitch

G

G

switch

S

Page 23: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Content

Why Application Specific Integrated Circuits?

Why Transistor Level Digital?

Why Analog?

Course Goal

Course Organization

Page 24: IN3170/4170, Spring 2020 · Why do an ASIC? Well,whynot? I Costly(development,designiterationtime,production) I Inflexibleandlowlevelofreusability Alternatives? I EmbeddedSystems

Teaching 18 lectures (Mondays 10-12 in Shell), lecture foils,podcast (no guarantee!), book: ‘MicroelectronicCircuits’ by Sedra & Smith, International (!) 7thEdition, selected papers

Labs 3 tasks (counting 40% towards final mark, task 1 isonly pass/not pass), lab assistant: Sebastian Wood,workgroups with up to 3 students

Paper exercises exercises in preparation for exam (!), Tuesdays14-16, teaching assistant: Zhijian Zhou

Tools Cadence, matlab, solder iron/bread board, labequipment

Skills electronics, maths, physics, programmingExam written exam, counting 60% towards final mark, early

in June