Des De s i gn gnii n g A Poss Pos s i bl ble e 5G PH PHY Y Wi th GFDM Gerhard P. Fettweis Ivan Gaspar, Luciano Mendes, Maximilian Matthé, Nicola Michailow, Andreas Festag, Rohit Datta, Martin Martin Danneberg, Dan Dan Zhang Zh ang coordinator
serial seria l entrepreneur entrepreneur
Wir eless >20 >2020 20 Outl Outl ook 100Tb/s 10 Tb/s 1 Tb/s 100Gb/s 802.11ac/ad
10Gb/s 802.11n
1Gb/s 802.11ag
100Mb/s
802.11b
10Mb/s 1Mb/s 100Kb/s
LTE Advanced HSPA HSDPA
802.11
10Kb/s GSM
1995
?
WLAN (10m)
Cellular (100m)
LTE
3G R99 / EDGE GPRS
2000
2005
2010
2015
2020
2025
2030
Wir eless >20 >2020 20 Outl Outl ook 100Tb/s 10 Tb/s 1 Tb/s 100Gb/s 802.11ac/ad
10Gb/s 802.11n
1Gb/s 802.11ag
100Mb/s
802.11b
10Mb/s 1Mb/s 100Kb/s
LTE Advanced HSPA HSDPA
802.11
10Kb/s GSM
1995
?
WLAN (10m)
Cellular (100m)
LTE
3G R99 / EDGE GPRS
2000
2005
2010
2015
2020
2025
2030
The Th e Tact Tactile ile Int Internet ernet And Its Millisecond
The Tact Tactii le Int nternet ernet Moving from 25ms RTT
1ms tomorrow
Reinventi ng “ Latin Classes”
© MPI Saarbrücken
Platooning 1ms examples of today’s cars: ESC, ABS
Tomorrow: platooned ESC & ABS
Revolution Ahead: The Tacti le Internet
5G: Steering & Control Communications
≤ 4G:
Content Communications
Health & Care Traffic & Mobility Sports & Gym Edutainment Manufacturing Smart Grid …
5G – “ Massi ve” Requirements Massive throughput t r a e h t f o e t a t S
Massive low latency
5G
Massive sensing Massive resilience Massive safety and security
The Tactile Internet
Massive fractal heterogenity
> 10Gbit/s per user
< 1ms RTT
> 10k sensors per cell < 10−8 outage
< 10−12 security 10x10 heterogeneity
5G Research on f our Tracks
Tacti le Internet app li cations Wir eless & Netwo rk
5G L A B GERMANY
Sili con systems
Mobi le edge clou d
Members on Tracks Silicon systems track
Wireless track
Mobile edge cloud track
Tactile Intern et application track
Gerhar d Fettweis Rene Schüffny
Leon Urbas
Eduard Jorswieck
Wolfgang Nagel Uwe Aßmann
Christof Fetzer Frank Ellinger Frank Fitzek
Wolfgang Lehner
Dirk Plett emeier Michael Schröter
Ercan Alt in so y
Thors ten Strufe Hermann Härtig Silvia Santin i
Klaus Janschek
Team of 500+ Researchers
!!!
Christel Baier
Relevant Startups Generated by Team Silicon sys tems track
Wireless track
freedelity
Mobile edge cloud track
Tactile Intern et application track
st a ctuators s ensors & t ransceivers f aInhalt GerhardFettweis),TUDresden Starting2014, fast fast
value chain sales & service systems, networks, software circuits components semiconductors
network of states Berlin Brandenburg Mecklenburg Vorpommern Saxony-Anhalt Saxony Thuringia Baden-Württ. Lower Saxony Bavaria
1ms Impact: 100µs PHY Latency Software Ecosystem Sensor
Embedded Computing
Trans mitter
100 s
Receiver Hosted Computing (decider)
1ms Actuator
Embedded Computing
Latency Goals: = 0.3 ms Terminal
Receiver
100 s
= 0.2 ms Air Interface
Trans mitter
= 0.5 ms Base Station & Compute
Network Config. Manager (SON)
Requir ements / Chall enges scalable bandwi dth
fragmented spectrum f
f
New Air Interface
async. operatio n
LTE clo cking scheme t packet latency 100µs
f
The Subcarrier Time/Frequency Dilemma frequency
R ×
1/R
time
GFDM Signal Properties Multi-carrier scheme
Block based approach Circular signal structure (time and frequency)
GFDM Signal Properties Multi-carrier scheme Block based approach Circular signal structure (time and frequency)
Pulse shaped sub-carriers
Overlapping sub-carriers
Multi -Carrier Revisited OFDM
GFDM
s r e i r r a c b u s N
N=KM
SC-FDM s e l p m a s . q e r f M
s e l p m a s y c n e u q e r f
s r e i r r a c b u s K
N
K time samples N time s amples M sub-symbols
N symbols
Realtime 5G Research Testbed: GFDM With -45dB to -65dB Notches
!
The LTE Frame Str ucture
...
frame n-1
frame n-1
frame n-1
...
10 ms subframe 0
...
subframe 9
1 ms slot 0
slot 1
0.5 ms C P
symbol 0 66.7
C P
symbol 1
C P
symbol 2
...
C P
symbol 6
μs
LTE is clearly one order of magnitude above the target latency. The 5G goal: Reduce the 1 ms based frame structure to something in the order of 50-100
μs.
Problems wi th OFDM Every symbol requires a cyclic prefix, which depends on the channel. a) It reduces the spectral efficiency. b) It prevents reducing the latency by shortening the symbols.
OFDM requires frequency flat subcarriers Subcarrier bandwidth coupled to symbol duration. Reducing symbol duration imposes constraint on frequency selectivity of the channel.
coherence bandwidth
Proposed Frame Str uct ure Step 1: Add a single prefix for entire frame. Step 2: Reduce the frame duration to 66.7 μs.
Introducing subsymbols allows to decouple the frequency resolution from the bandwidth of the subcarriers. Coherence bandwidth can be the same as in OFDM (given same sampling frequency).
One LTE Symbol Can Host One 5G GFDM Packet LTE Symbol (e.g. 4 PRBs)
GFDM TTI (e.g. 7 Subsymbols)
TTI Framing Opportunity
100µs
LTE 1ms OFDM TTI (here with 4 PRBs)
5G 70µs GFDM TTI (here with 7 sub-symbols per symbol)
OQAM in GFDM Analyzing the structure of interference for a circular RC pulse
OQAM in GFDM Modulator
complex symbols
I&Q selecto r
GFDM Modulato r1
GFDM Modulato r2
Demodulator
+
modulated s amples
GFDM Demodulat or 1
GFDM Demodulat or 2
complex
I&Q select or
symbols
Flexible Spacing Between Subcarriers & Subsymbols FTN
e m i t
y c n e u q e r f
SE-FDM
FMT
GFDM fo r Virt ualization of 5G Wavefo rm Candidates The design space of GFDM can be explored to obtain other waveforms classical waveforms
filterbank based waveforms
nonorthogonal multicarrier
Machine Type Communications (M2M SIG)
Asynchr onous transmi ssion Time offsets •
In M2M, time shifts between the users cannot be controlled
•
Users cannot be orthogonal
•
Increased protection can be achieved by re-allocating the CP
non-orthogonal signaling inevitable
0
-20
Frequency offsets •
Should be handled by a subcarrier filter
2
| ) f ( G |
-40
-60
with low spectral leakage •
1 SC guard between users should be sufficient
-80
-100
sinc RC, a=0.5
TTI Framing Opportunity
M2M Sensors
LTE 1ms OFDM TTI (here with 4 PRBs)
5G 1ms GFDM TTI 25Bytes generated by a modified SC-FDMA HW!
Requirements / Challenges for 5G PHY scalable bandwi dth f fragmented spectrum need d eep notch es f packet l atency 100µs
Multi Carrier New Air Interface: GFDM Subcarrier Filters Generalized Frequency Division Multiplexing Compact Packet
GFDM a Prime Candidate f or 5G Flexibility/Parameterization Spectrum Engineering OFDM Compatibility -
Evolution
of OFDM and SC-FDE
Synchronization / Equalization / Clocking
Fully MIMO capable (… OFDM and SC-FDE are special cases) TEAM EFFORT B Y Ivan Gaspar, Luc iano Mendes, Maximil ian Matt hé, Nico la Michail ow, An dreas Festag, Rohi t Datta, Martin Danneberg, Dan Zh ang
5G Realti me Lab Sponsor ed by Realtime GFDM With -45dB to -65dB Notches
Cellular Roadmap of USPs
5G – 2022 + Tactile Internet + massi ve M2M + Tb/s
4G – 2012 3G – 2002 2G – 1992 Voice Messages
+ Data + Positioning
+ “ carrier grade” + safe & secure
+ Video everything + 3D Graphics
GFDM
Thank you
5G Lab Germany Coordinators: Gerhard Fettweis Frank Fitzek
dresden5GLab.org
[email protected]
Thank You
or 20 years of continued support !
