DR. KIRSTEN MATHEUS OCTOBER 9, 2018, …...K. Matheus, BMW AG, Ethernet&IP@Automotive Technology Day...

EVOLUTION OF ETHERNET-BASED AUTOMOTIVE NETWORKS: FASTER AND CHEAPER.

DR. KIRSTEN MATHEUSOCTOBER 9, 2018, LONDON

8TH ETHERNET & IP @ AUTOMOTIVE TECHNOLOGY DAY

K. Matheus, BMW AG, Ethernet&IP@Automotive Technology Day 2018

AGENDA.

−How did it all start?

−Where are we today?

−Why is automotive Ethernet more than just another IVN technology?

−What is the role of higher data rates?

−What is the role of lower data rates?

−Conclusion


AUTOMOTIVE ETHERNET STARTED AS AN IVN TECHNOLOGY TO ADDRESS THE DEMANDS FOR HIGHER DATA RATES.

− 1. Use case: To limit the time needed for flash updates to 15minutes for 1GByte of flash data.

− With the existing CAN the updates would have taken 16h.

− With 100BASE-TX Ethernet it was possible to meet the target.

− 2. Use case: To allow the RSE to access the map data stored in the HU for new customer functions.

− MOST25 as such does not allow to transmit 20Mbps of data packets. Additionally the bandwidth of the existing MOST25 was used up in any case

− With 100BASE-TX it was possible to meet the target.

− 3. Use case: To transmit digitized (camera) video data.

− For the introductory use case chosen, it was more cost efficient to compress the video data and to transmit it over 100BASE-T than the LVDS technologies available at the time.


INITIAL IMPROVEMENTS OF AUTOMOTIVE ETHERNET MADE IT MORE POWERFUL AS A TECHNOLOGY FOR HIGH(ER) DATA RATES.

Extension in the infotainment and driver assistance domains−Replacement of MOST25−Adding the antenna module and other (new)

infotainment systems−Adding systems from the driver assist domain,

which require the bandwidth during runtime/for flash or for IP and Ethernet communication

Extended use of AVB−Time synch 802.1AS, in parts also usable over

CAN&Co−1722 to bypass IP in case of audio/video

transmission−Stream reservation, traffic class C−Traffic shaping

Standardization of 1000BASE-T1−Growth possibility needed to be initiated early

in order to be future proof−CFI in March 2012−Publication of 1000BASE-T1 standard in June

2016−SOP at BMW planned for 2021

Standardization of TSN−Addressing safety critical applications−Redundancy−Preemption, time aware shaping−Multiple master clocks …


AGENDA.






−Conclusion


SINCE ITS INTRODUCTION, THE NUMBER OF ETHERNET ECUS HAS CONTINUOUSLY INCREASED IN THE BMW E/E ARCHITECTURE.

GW

I&C1

I&C4I&C2

I&C3I&C5

I&C7

I&C6

I&C8

DA2

DA3

DA4

DA1`DA1

HSFZ

C1 C2 C3 C4

⊗

⊗

HSFZGW

DA2

DA3

DA4

DA1`DA1

I&C1

I&C4I&C2

I&C3

C1 C2 C3 C4

⊗

HSFZGW

DA1

I&C1

I&C2

C1 C2 C3 C4

HSFZ*)GW

I&C1

I&C2

2008 2013 2015

100BASE-TX 100BASE-T1 exor*) High Speed Car Access


TRACTION IS ALSO SEEN IN THE INDUSTRY AS SUCH AS OTHER CAR MANUFACTURERS ARE FOLLOWING SUIT.

OEM B

2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020

100B

ASE-

TX10

0BAS

E-T1

OEM C

OEM DOEM EOEM F

OEM GOEM H

Identity not yet public

OEM b OEM

c

OEM e

OEM fOEM

d

OEM g

OEM h


HOWEVER, WITH ETHERNET BEING JUST ANOTHER IVN, THERE IS AN INCREASING NUMBER OF IVN-TECHNOLOGIES WHICH NEED TO BE SUPPORTED.

P2P

Bus

Ethernet

P2P + someEthernet

1985 1990 1995 2000 2005 2010 2015 2020 2025

10k

1k

100

10

1

0,1

0,01LIN

MOST25

Gro

ssbi

trat

e[M

bps]

Year of appearance

SerDes Koax

100BASE-TX

OABR/ 100BASE-T1

PSI5

USB 2.0

CAN-FD

SENT

LVDSAPIX

A2B„PLC“

PWM

FlexRay

1000BASE-T1

CAN High

CXPI

D2B Byteflight

VAN

eMOST50

MOST150

USB 3.0

USB 3.1

K-Line J1850

GEPOF

New developments

Discontinued technologies

HDMI 1.2

APIX 2

APIX 3

HDbaseT

cMOST150

IseLED

MIPI SerDes

10BASE-T1S

2.5GBASE-T1

10GBASE-T15GBASE-T1


WHILE AT THE SAME TIME THE IVN NEEDS TO BE MORE POWERFUL AND ROBUST TO SUPPORT CHALLENGING NEW REQUIREMENTS.

Mobile services and data:• Many new apps and services

during the lifetime of a car. • Regular software updates (also

over the air, also 3rd party software).

• Always on, with the car as just another node in the world wide network.

• More communication in- and outside.Al l t t it t d d

E-mobility: • Regulation/legislation

supported.• Continuous updates on routes,

reach, and charging during runtime.

• Car has a unique identity in the world wide grid (smart billing).

• Cross-domain communication inside the car.

Autonomous Driving:• Large amounts of sensor data.• Processing across different

ECUs (powerful communication network).

• Algorithms enhanced by offsite information and processing.

• Service oriented architecture.• Regular software updates. • Always latest security

standards.


AGENDA.






−Conclusion


AUTOMOTIVE ETHERNET AS DEFINED IN THIS PRESENTATION COVERS THE COMPLETE PROTOCOL STACK WITH STRICT LAYER SEPARATION.

OptimizationIT Standard Automotiv

e

Audio/VideoTransport

Time Sync

ControlCommuni-

cation

Diagnosisand Flash

UpdateService

DiscoveryAddressConfig.

Address Resolution,Signaling,

etc.

AutomotiveNetwork

Management

Layer 3

Layer 4

Layer 5-7

Layer 1

Layer 2 IEEE Ethernet MAC + VLAN (802.1Q)

Automotive Ethernet Physical Layers (100BASE-T1, 1000BASE-T1, …)

UDP-NM

SOME/IP

SOME/IP-SD

DHCP

IEEE802.1

AS

AVB

IEEE1722

AVB

ICMPARP

DoIP

UDP UDPTCP and/or UDP

IPv4/IPv6

AUTOMOTIVE ETHERNET ALLOWS FOR SIGNIFICANTLY IMPROVED SECURITY MEASURES OVER LEGACY NETWORKS.

Auth?Enc?

MulticastBroadcast

100%protected?

# ofkeys?

Dyn.Keys?

MinimumOverhead

Selector Implementation ConfigComplexity

MACsec

Yes/Yes

Yes Yes Lowest Yes ~2% L1/L2 State-of-the-artHardware + Software

Low

IPsec Yes/Yes

No No Low Yes ~2% L3 + L4 State-of-the-artSoftware

Low-Medium

(D)TLS Yes/Yes

No No Medium Yes ~2% L4 only State-of-the-artSoftware

Low-Medium

SecOCEth

Yes/*1 somewhatpossible

No High *2 ~2% Depends

NewSoftware + *4

High

SecOCCAN


somewhatpossible

High *2 ~100%*3 *5

ID NewSoftware + *4

High

SecOCCAN-FDFR


somewhatpossible

High *2 ~25% (CAN-FD)

ID NewSoftware + *4

High


*1 Possible to integrate in standard*2 No standardized solution exists.*3 Only reduced security.

CAN…

Ethernet

*4 Additional new Hardware might be needed to reduce group key trust limitatio*5 Avoiding Transport Protocol due to safety reasons.

AUTOMOTIVE ETHERNET OFFERS THE RIGHT FLEXIBILITY AND TOPOLOGY CHOICES TO SERVE AS A HOMOGENEOUS IVN.


Automotive Ethernet E.g. CAN(-FD), FlexRayNumber of ECUs in network 256 unlimited, bandwidth can

be addedCAN e.g. 16 (SAE), FR e.g. 4/x, limited by bandwidth

Different speed grades supported in the same network

Various (different PHYs but same protocol stack)

One (change over to new IVN if data rate no longer sufficient)

Bandwidth efficiency 50-90% 20-55%Topologies Extremely flexible limitedPriority and timing schemes AVB-TSN Message or ID/timeslot basedAddressing MAC and IP NoneSecurity State of the art LimitedDomain separation Physical or virtual PhysicalExtendibility Switch port, no changes to existing

nodesCAN: no changes to existing nodes, FR: might need active star port

Service orientation Possible Not really possible

Automotive Ethernet E.g. CAN(-FD), FlexRayNumber of ECUs in network 256 unlimited, bandwidth can

be addedCAN e.g. 16 (SAE), FR e.g. 4/x, limited by bandwidth

Different speed grades supported in the same network

Various (different PHYs but same protocol stack)

One (change over to new IVN if data rate no longer sufficient)

Bandwidth efficiency 50-90% 20-55%Topologies Extremely flexible LimitedPriority and timing schemes AVB-TSN (various) Message or ID/timeslot based

(one)Addressing MAC and IP NoneSecurity State of the art LimitedDomain separation Physical or virtual PhysicalExtendibility Switch port, no changes to existing

nodesCAN: no changes to existing nodes, FR: might need active star port + timing

Service orientation Possible Not really possible


AGENDA.






−Conclusion

10k

1k

100

10

1

0,1

0,01LIN

MOST25

SerDes Koax

100BASE-TX

OABR/ 100BASE-T1

PSI5

USB 2.0

CAN-FD

SENT

LVDSAPIX

A2B„PLC“

PWM

FlexRay

1000BASE-T1

CAN High

CXPI

D2B Byteflight

VAN

eMOST50

MOST150

USB 3.0

USB 3.1

K-Line J1850

GPOF

HDMI 1.2

APIX 2

APIX 3

HDbaseT

cMOST150

IseLED

MIPI SerDes

10BASE-T1S

2.5GBASE-T1

10GBASE-T15GBASE-T1


THE LARGE NUMBER OF TECHNOLOGIES ADDRESSING HIGHER DATA RATES SHOWS THAT THE DEMAND IS THERE TODAY.

P2P

Bus

Ethernet

P2P + someEthernet

1985 1990 1995 2000 2005 2010 2015 2020 2025

Gro

ssbi

trat

e[M

bps]

Year of appearance

New developments

Discontinued technologies

10k

1k

1001985 1990 1995 2000 2005 2010 2015 2020 2025

Year of appearance

SerDes Koax

100BASE-TX

OABR/ 100BASE-T1

USB 2.0

LVDSAPIX

1000BASE-T1

USB 3.0

USB 3.1

GEPOF

HDMI 1.2

APIX 2

APIX 3

HDbaseT

MIPI SerDes

2.5GBASE-T1

10GBASE-T1

5GBASE-T1


FOR AUTOMOTIVE ETHERNET AUTONOMOUS DRIVING IS DRIVING THE NEED FOR HIGHER DATA RATES.

The main drivers are−Higher resolution for sensor data & displays.−Fast communication between high end computing platforms.−The need for using the Ethernet protocols.

The existing technologies for higher data rates are unsuitable because they are−P2P (i.e. they do not support networking functions/Ethernet protocols) and/or−Consumer technologies (i.e. their cabling is not efficient in the car) and/orP i t (i th l k th t )


THE MAIN CHALLENGES FOR HIGHER DATA RATES ARE COSTS, POWER CONSUMPTION AND TIMING.

Technical challenges:−Robustness of the PHY technology ( BER < 10 -12).−EMC behavior at high frequencies ( affects cabling and implementation choices).−Power consumption ( data rate and error avoidance/correction).

Other challenges:−High data rate Ethernet has to compete with proprietary solutions (cost, timing). −It is a constant race: The data rate provided is never enough. The need for data rates beyond 10Gbps is envisioned, while the standardization process needs time


AGENDA.






−Conclusion


90% OF IVN COMMUNICATION IS BELOW 10MBPS. TO USE 100MBPS ETHERNET FOR THOSE CASES IS NOT COST AND POWER EFFICIENT.

Rel. system costs

10 Gbps

1 Gbps

100 Mbps

10 Mbps

1 Mbps

100 kbps

10 kbps

Data rate

1000BASE-T1

HS-CAN

CAN-FD

LIN ? 1 (log.scale) ?

?

NGAUTO

FlexRay

100BASE-T1

10BASE-T1S

7.3% of nodes*)

Ethernet,MOST, SerDes

92.7% of nodes*)

Safety Bus, LIN, CAN (-FD)

Missing element

*) forecast for 2020, source: Strategy Analytics


COST REDUCTION FOR 10BASE-T1S IS ACHIEVED BY SIMPLIFIED PHY SPECIFICATION AND MULTIDROP CAPABILITY.

Low frequency channel and “low” data rate allow for• Simple equalization• No forward error correction Because of the half duplex transmission 10BASE-T1S needs• No echo cancellation• No hybridMultidrop (bus topology) allows for• One PHY per ECU instead of two per link

.

.

.

P

P

P

P

P

PXIVN

Passive linear bus (daisy chain)


MULTIDROP REQUIRES A (NEW) MEDIUM ACCESS SCHEME, WHICH USES THE BANDWIDTH EFFICIENTLY.

PLCA is based on a fairness per packet bases.It can be used with the existing CSMA interface in MAC and switches.It guarantees a maximum latency.

• Every node receives (minimum 1) access ID

• Every node is allowed to transmit one packet in order of the IDs

• If a node does not transmit, the next node in line can use the transmit slot

000

1

22

3

0 01 2 2 3 0

Max delay = (users-1)*max packet length*), min delay can be immediate

Que

ues

Out

put *) max packet length for

an Ethernet packet (1500 bytes payload) on a 10Mbps link ~1.2ms.


THE MAIN CHALLENGE FOR 10BASE-T1S IS COST EFFECTIVENESS.

IVN decisions within car manufacturers are often based on hardware costs only.Ethernet thus has to compete with legacy networks like CAN, CAN-FD, FlexRay,

etc.The possibility to fully integrate PHY/MAC with processors offers further cost

saving potential. However, the relevance of modern communication paradigms (security, service

based architecture, …) will decide on the use.

PHY MAC

Processor

e.g. MIIPHY MA

C

Processo

r

e.g. SPI

PHY MAC

Switch

e.g. MIIPHY MA

CSwitc

h

Standard stand alone PHYs MACPHY *) (new)

PHY MAC

Processor

*) Allows the use of smaller processors typically used in CAN


AGENDA.






−Conclusion


CONCLUSION.

−Automotive Ethernet started as yet another IVN technology in order to address the demand for higher data rates in the IVN. −But, Automotive Ethernet can do a lot more: It can change the way we design the EE-Architecture and IVN, i.e. the way we develop cars. −With an extensive Ethernet IVN we can address security, service based architecture, unambiguous addressing, software updates etc. more efficiently than with legacy IVN technologies.−However, we need to fully exploit the potential of Ethernet, we need to use Ethernet as a growing system that also addresses demands of lower AND higher data rates while becoming more cost efficient.

Dr. Kirsten Matheus

THANK YOU FOR YOUR ATTENTION

DR. KIRSTEN MATHEUS OCTOBER 9, 2018, …...K. Matheus, BMW AG, Ethernet&IP@Automotive Technology Day...

Documents

Transcript of DR. KIRSTEN MATHEUS OCTOBER 9, 2018, …...K. Matheus, BMW AG, Ethernet&IP@Automotive Technology Day...