Chirp Spread Spectrum (CSS) PHY Submission

1
Project IEEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs) Submission Title
Chirp Spread Spectrum (CSS) PHY Presentation for
802.15.4a Date Submitted January 04,
2005 Source John Lampe Company Nanotron
Technologies Address Alt-Moabit 61, 10555
Berlin, Germany Voice 49 30 399 954 135, FAX
49 30 399 954 188, E-Mail j.lampe_at_nanotron.com
Re This is in response to the TG4a Call for
Proposals, 04/0380r2 Abstract The Nanotron
Technologies Chirp Spread Spectrum is described
and the detailed response to the Selection
Criteria document is provided Purpose Submitted
as the candidate proposal for TG4a
Alt-PHY Notice This document has been prepared
to assist the IEEE P802.15. It is offered as a
basis for discussion and is not binding on the
contributing individual(s) or organization(s).
The material in this document is subject to
change in form and content after further study.
The contributor(s) reserve(s) the right to add,
amend or withdraw material contained
herein. Release The contributor acknowledges and
accepts that this contribution becomes the
property of IEEE and may be made publicly
available by P802.15.
2
Chirp Spread Spectrum (CSS)PHY Presentationfor
802.15.4a

• by
• John Lampe, Rainer Hach, Lars Menzer
  Nanotron Technologies GmbHBerlin, Germany
• www.nanotron.com

3
General Properties of Chirp Signals

• Simplicity
• Basically a 2 ary baseband transmission
• The windowed chirp is a linear frequency sweep
  with a total duration of 1 us

4
Chirp Properties (cont.)

• Real part of a windowed up-chirp signal with a
  total duration of 1us

5
Chirp Properties (cont.)

• Figure shows the autocorrelation function (acf)
  and cross-correlation function (ccf) of the
  signal described above
• Note that the ccf has a constant low value
  (compared to DS sequences).

6
Chirp Properties (cont.)

• This figure shows the PSD of the signal with a
  power of 10 dBm
• By padding the signal with zeros the frequency
  resolution has been set to 100 kHz so that the
  plot is similar to what a spectrum analyzer would
  measure.
• At 12 MHz offset from the center is below -30dBm
  (which is the ETSI requirement for out of band
  emissions)

7
Chirp Properties (cont.)

• Further processing of the signals Sig A and Sig B
  for symbol detection could be done in coherent
  (real part processing) or non coherent manner
  (envelope filtering).
• Since the analytical results are well known for
  AWGN channels we will mention these
• Simulations over other channels will all refer to
  the non coherent system as drawn below.

8
Chirp Properties (cont.)

• This figure shows the analytical BER values for 2
  ary orthogonal coherent and non coherent
  detection and the corresponding simulation
  results (1E5 symbols) for up down chirp (using
  the chirp signals defined above)
• The performance loss due to the non-orthogonality
  of up and down chirp is very small.

9
Signal Robustness

• The proposed CSS PHY is designed to operate in a
  hostile environment
• Multipath
• Narrow and broadband intentional and
  unintentional interferers
• Since a chirp transverses a relatively wide
  bandwidth it has an inherent immunity to narrow
  band interferers
• Multipath is mitigated with the natural frequency
  diversity of the waveform
• Broadband interferer effects are reduced by the
  receivers correlator
• Forward Error Correction (FEC) can further reduce
  interference and multipath effects.
• Three non-overlapping frequency channels in the
  2.4 GHz ISM band
• This channelization allows this proposal to
  coexist with other wireless systems such as
  802.11 b, g and even Bluetooth (v1.2 has adaptive
  hopping) via DFS
• CSS proposal utilizes CCA mechanisms of Energy
  Detection (ED) and Carrier Detection
• These CCA mechanisms are similar to those used in
  IEEE 802.15.4-2003
• In addition to the low duty cycle for the
  applications served by this standard sufficient
  arguments were made to convince the IEEE 802
  sponsor ballot community that coexistence was not
  an issue.

10
Support for Interference Ingress

• Example (w/o FEC)
• Bandwidth B of the chirp 20 MHz
• Duration time T of the chirp 1 µs
• Center frequency of the chirp (ISM band) 2.442
  GHz
• Processing gain, BT product of the chirp 13 dB
• Eb/N0 at detector input (BER10-4) 12 dB
• In-band carrier to interferer ratio (C/I _at_
  BER10-4) 12 - 13 -1 dB
• Implementation Loss 2 dB

11
Support for Interference Egress

• Low interference egress
• IEEE 802.11b receiver
• More than 30 dB of protection in an adjacent
  channel
• Almost 60 dB in the alternate channel
• these numbers are similar for the 802.11g
  receiver

12
Regulatory

• Devices manufactured in compliance with the CSS
  proposal can be operated under existing
  regulations in all significant regions of the
  world
• including but not limited to North and South
  America, Europe, Japan, China, Korea, and most
  other areas
• There are no known limitation to this proposal as
  to indoors or outdoors
• The CSS proposal would adhere to the following
  worldwide regulations
• United States Part 15.247 or 15.249
• Canada DOC RSS-210
• Europe ETS 300-328
• Japan ARIB STD T-66

13
Scalability Data Rate

• Mandatory rate 1 Mb/s
• Optional rate 267 Kb/s
• Other possible data rates include 2 Mb/s to allow
  better performance in a burst type, interference
  limited environment or a very low energy
  consumption application
• Lower data rates achieved by using interleaved
  FEC
• Lower chirp rates would yield better performance
• longer range, less retries, etc. in an AWGN
  environment or a multipath limited environment
• It should be noted that these data rates are only
  discussed here to show scalability, if these
  rates are to be included in the draft standard
  the group must revisit the PHY header such as the
  SFD.

14
Scalability Frequency Bands

• The proposer is confident that the CSS proposal
  would also work well in other frequency bands
• Including the 5975 to 7250 MHz band
• Mentioned in the new FCC operating rules SECOND
  REPORT AND ORDER AND SECOND MEMORANDUM OPINION
  AND ORDER released December 16, 2004

15
Scalability Data Whitener

• Additionally, the group may consider the use of a
  data whitener, similar to those used by Bluetooth
  and IEEE 802.11 to produce a more noise-like
  spectrum and allow better performance in
  synchronization and ranging.

16
Scalability Power Levels

• For extremely long ranges the transmit power may
  be allowed to rise to each countrys regulatory
  limit
• For example the US would allow 30 dBm of output
  power with up to a 6 dB gain antenna
• The European ETS limits would specify 20 dBm of
  output power with a 0 dB gain antenna
• Note that even though higher transmit requires
  significantly higher current it doesnt
  significantly degrade battery life since the
  transmitter has a much lower duty cycle than the
  receiver, typically 10 or less of the receive
  duty cycle
• In this manner the averaged transmitter current
  drain will be less than the averaged receiver
  current drain.

17
Scalability Backward Compatibility

• Due to some of the similarities with DSSS it is
  possible to implement this proposal in a manner
  that will allow backward compatibility with the
  802.15.4 2.4 GHz standard
• The transmitter changes are relatively
  straightforward
• Changes to the receiver would include either dual
  correlators or a superset of CSS and DSSS
  correlators
• It is anticipated that this backward
  compatibility could be achieved via mode
  switching versus a dynamic change on-the-fly
  technique
• left up to the implementer
• This backward compatibility would be a
  significant advantage to the marketplace by
  allowing these devices to communicate with
  existing 802.15.4 infrastructure and eliminating
  customer confusion

18
Mobility Values

• Communication
• No system inherent restrictions are seen for this
  proposal
• The processing gain of chirp signals is extremely
  robust against frequency offsets such as those
  caused by the Doppler effect due to high relative
  speed vrel between two devices
• Such situations also occur when one device is
  mounted on a rotating machine
• The limits will be determined by other, general
  processing modules (AGC, symbol
  synchronization,...)
• Ranging
• The ranging scheme proposed in this document
  relies on the exchange of two hardware
  acknowledged data packets
• One for each direction between two nodes
• We assume that the longest time in this procedure
  is the turnaround time tturn between the two
  nodes which will be determined by the respective
  uC performance. During this time the change of
  distance should stay below the accuracy da
  required by the application.

• For da 1m
• tturn 10 ms this yields
• vrel ltlt 100m/s

19
MAC Enhancements and Modifications

• There are no anticipated changes to the 15.4 MAC
  to support the proposed Alt-PHY. Three channels
  are called for with this proposal and it is
  recommended that the mechanism of channel bands
  from the proposed methods of TG4b be used to
  support the new channels. There will be an
  addition to the PHY-SAP primitive to include the
  choice of data rate to be used for the next
  packet. This is a new field.
• Ranging calls for new PHY-PIB primitives that are
  expected to be developed by the Ranging
  subcommittee.

20
Channel Models

• Since this proposal refers to the 2.4GHz ISM
  band, only channel models with complete parameter
  set covering this frequency range can be
  considered
• At the time being these are LOS Residential (CM1)
  and NLOS Residential (CM2).
• The 100 realizations for each channel model were
  bandpass filtered with -15MHz around 2.437GHz
  which corresponds to the second of the three
  sub-bands proposed.
• The filtered impulse responses were down
  converted to complex baseband.
• The magnitudes over time are shown in the
  following plots
• Furthermore some graphs of the function H_tilde
  as described and required in the SCD are shown
• For now we assume that the neighbor sub-bands
  will not differ significantly from the center
  sub-band and that we restrict simulations on the
  center sub-band
• The SCD requirements on the payload size to be
  simulated seem to be somewhat inconsistent. At
  some point 10 packets with 32 bytes are mentioned
  which would be a total of 2560 bits. On the other
  hand a PER of 1 is required which mean
  simulating more than 100 packets or 25600 bits.
• Since the delay spread and thus the time in which
  subsequent symbols can influence each other of
  all given channel impulse responses are well
  below the symbol duration of 1us suggested by
  this proposal we believe that we get the best
  results when we simulate a large number of
  independent transmissions of symbols.
• Assuming an equal probability of error for all
  bits of a packet we can give the relationship
  between the BER and PER by
  With N being the number of payload
  bits.
• Thus we can calculate the BER which is required
  for any PER
• For PER1 and N256 we get BER3.9258E-5

21
Channel Model 1
22
Channel Model 2
23
Size and Form Factor

• The implementation of the CSS proposal will be
  much less than SD Memory at the onset
• following the form factors of Bluetooth and IEEE
  802.15.4/ZigBee
• The implementation of this device into a single
  chip is relatively straightforward
• As evidenced in the Unit Manufacturing
  Complexity slides

24
PHY SAP Payload Bit Rate Data Throughput

• The PPDU is composed of several components as
  shown in the figure below

• The following figure shows in greater detail, the
  component parts of each PPDU.

25
PHY SAP Payload Bit Rate Data Throughput (cont)

• The SFD structure has different values for, and
  determines, the effective data rate for PHR and
  PSDU
• The Preamble is 32 bits in duration (a bit time
  is 1 us)
• In this proposal, the PHR field is used to
  describe the length of the PSDU that may be up to
  256 octets in length
• In addition to the structure of each frame, the
  following shows the structure and values for
  frames including overhead not in the information
  carrying frame

26
PHY SAP Payload Bit Rate Data Throughput (cont)

• The figures show the structure, as defined in the
  IEEE Standard 802.15.4 and the SCD
• cases of acknowledged transmissions as used in
  section 3.2.1 values and for unacknowledged
  transmissions.
• For this proposal, the value of Tack and SIFS are
  retained from the IEEE Standard 802.15.4 and are
  each 192 microseconds
• The value of LIFS is also shown as 192
  microseconds.
• Additional revisions of this proposal may show a
  different value as the authors discuss the need
  for longer LIFS values with members of TG4b
• The values of SIFS and LIFS have a MAC dependency
  above the value of 192uS required for PHY turn
  around
• SIFS has a value of 192us (12 symbols) in the
  current standard and LIFS has a value of 40
  symbols.

27
Simultaneous Operating Piconets

• Separating Piconets by frequency division
• This CSS proposal includes a mechanism for FDMA
  by including the three frequency bands used by
  802.11 b, g and also 802.15.3
• It is believed that the use of these bands will
  provide sufficient orthogonality
• The chirp signal defined earlier has a rolloff
  factor of 0.25 which in conjunction with the
  space between the adjacent frequency bands allows
  filtering out of band emissions easily and
  inexpensively.

28
Signal Acquisition

• The signal acquisition is basically determined by
  the structure and duration of the preamble
• In contrast to always on systems like DECT or
  GSM
• Low duty rate systems must be able to acquire a
  signal without any prior knowledge about that
  signals level or timing
• While the IEEE 802.15.4-2003 uses a preamble
  duration of 32 symbols (128 µs at 2.4 GHz) other
  commercially available transceiver chips (e.g.
  nanoNET TRX from Nanotron) use 30 symbols at
  1MS/s (i.e. 30µs).
• For consistency with IEEE 802.15.4-2003 this CSS
  proposal is based upon a preamble of 32 symbols
  which at 1MS/s turns out as 32 µs
• Existing implementations demonstrate that
  modules, which might be required to be adjusted
  for reception (Gain Control, Frequency Control,
  Peak Value Estimation, etc.), can be setup in
  such a time duration
• The probability of missing a packet is then
  simply determined by the probability that the SFD
  is received correctly
• As shown before, the BER required for a PER of
  1, is BER3.9258E-5

29
Clear Channel Assessment

• A combination of symbol detection (SD) and energy
  detection (ED) has proven to be useful in
  practice. The duration of the preamble can be
  used as upper bound for the duration for both
  detection mechanism. By providing access to the
  threshold for ED the system allows the
  application to adjust its behavior (false alarm
  vs. miss probability) according to its needs.

30
System Performance

• Simulation over 100 channel impulse responses (as
  required in the SCD) were performed for channel
  model 1 and channel model 2.
• No bit errors could be observed on channel model
  1 (simulated range was 10 to 2000m). This is not
  really surprising because this model has a very
  moderate increase of attenuation over range
  (n1.79)
• The results for channel model 2 are displayed
  below. The parameter n4.48 indicates a very high
  attenuation for higher ranges. The results were
  interpreted as BER and PER respectively and for
  convenience were plotted twice (linear and log y
  scale).

31
System Performance
32
System Performance
33
System Performance
34
TOA Estimation for Ranging
Noise and Jitter of Band-Limited Pulse Given a
band-limited pulse with noise su we want to
estimate how the jitter (timing error) st with
which the passing of the rising edge of the pulse
through a given threshold can be detected is
effected by the bandwidth B.
35
Symmetrical Double-Sided Two-Way Ranging (SDS
TWR)
Tround ... round trip time Treply ... reply
time Tprpp ... propagation of pulse
Double-Sided Each node executes a round trip
measurement. Symmetrical Reply Times of both
nodes are identical (TreplyA TreplyB). Results
of both round trip measurements are used to
calculate the distance.
36
Ranging - Effect of Time Base Offset Errors
Assuming offset errors eA, eB of the timebases
of node A and B we get
On the condition that the two nodes have almost
equal behavior, we can assume
This has the effect that timebase offsets are
canceled out
Calculations show that for 40ppm crystals and 20
us max difference between TroundA and TroundB
and between TreplyA and TreplyB an accuracy
below 1 ns can easily be reached!
37
Influence of Symmetry Error Calculation of an
SDS TWR Example System

• Example System EtA 40 ppm, EtB 40 ppm
  (worst case combination selected)

• Conclusion Even 20 µs Symmetry Error allows
  excellent accuracy of distance ! Symmetry Error
  below 2 µs can be guaranteed in real
  implementations !

38
Principle of Dithering and Averaging

• Dithering

• Averaging

39
Simulation of a SDS TWR System
Example system Simulates SDS TWR Dithering
Averaging Crystal Errors 40 ppm Single shot
measurements _at_ 1 MBit/s data rate
(DATA-ACK) Transmit Jitter 4 ns
(systematic/pseudo RN-Sequence) Pulse detection
resolution 4 ns Pulses averaged per packet
32 Symmetry error 4 µs (average) Distance 100
m Results of Distance Error ?d ?dWC lt 50
cm ?dRMS lt 20 cm (ideal channel without noise)
40
Link Budget

• footnotes
• 1 Rx noise figure in addition the proposer can
  select other values for special purpose (e.g. 15
  dB for lower cost lower performance system)
• 2 The minimum Rx sensitivity level is defined
  as the minimum required average Rx power for a
  received symbol in AWGN, and should include
  effects of code rate and modulation.

41
Sensitivity

• The sensitivity to which this CSS proposal refers
  is based upon non-coherent detection
• It is understood that coherent detection will
  allow 2 - 3 dB better sensitivity but at the cost
  of higher complexity (higher cost?) and poorer
  performance in some multipath limited
  environments
• The sensitivity for the 1 Mb/s mandatory data
  rate is -92 dBm for a 1 PER in an AWGN
  environment with a front end NF of 7 dB
• The sensitivity for the optional 267 kb/s data
  rate is -97 dBm for a 1 PER in an AWGN
  environment with a front end NF of 7 dB

42
Power Management Modes

• Power management aspects of this proposal are
  consistent with the modes identified in the IEEE
  802.15.4 2003 standard
• There are no modes lacking nor added
• Once again, attention is called to the 1 Mbit/s
  basic rate of this proposal and resulting shorter
  on times for operation

43
Power Consumption

• The typical DSSS receivers, used by 802.15.4, are
  very similar to the envisioned CSS receiver
• The two major differences are the modulator and
  demodulator
• The modulator of the CSS is much simpler than the
  DSSS however since the major power consumption is
  the transmitter and the difference is negligible
• The power consumption for a 10 dBm transmitter
  should be 198 mW or less
• The receiver for the CSS is remarkably similar to
  that of the DSSS with the major difference being
  the correlator
• The correlator for the CSS uses a frequency
  dispersive mechanism while the DSSS uses a chip
  additive correlator
• The difference in power consumptions between
  these correlators is negligible so the power
  consumption for a 6 dB NF receiver should be 40
  mW or less
• The power consumed during the CCA is basically
  similar to the receiver power consumption
• All of the receiver circuits are being used
  during the CCA (correlator is used for the
  carrier detect function)

44
Power Consumption (cont.)

• Power save mode is used most of the time for this
  device and has the lowest power consumption
• Typical power consumptions for 802.15.4 devices
  are 3 µW or less
• Energy per bit is the power consumption divided
  by the bit rate
• The energy per bit for the 10 dBm transmitter is
  less than 0.2 µJ
• The energy per bit for the receiver is 40 nJ
• As an example, the energy consumed during an
  exchange of a 32 octet PDU between two devices
  (including the transmission of the PDU, the tack,
  and the ack) would be 70.6 µJ for the sender
  device while the receiving device consumed 33.2
  µJ
• As a reference point it should be noted that
  according to the Duracell web site, a Duracell AA
  alkaline cell contains more than 12,000 Joules of
  energy.

45
Antenna Practicality

• The antenna for this CSS proposal is a standard
  2.4 GHz antenna such as widely used for 802.11b,g
  devices and Bluetooth devices. These antennas
  are very well characterized, widely available,
  and extremely low cost. Additionally there are a
  multitude of antennas appropriate for widely
  different applications. The size for these
  antennae is consistent with the SCD requirement.

46
Time To Market

• No regulatory hurdles
• No research barriers no unknown blocks, CSS
  chips are available in the market
• Normal design and product cycles will apply
• Can be manufactured in CMOS

47
Unit Manufacturing Complexity

• Target processRF-CMOS, 0.18 µm feature size

48
Unit Manufacturing Complexity

• Target processRF-CMOS, 0.13 µm feature size

49
Summary

• Chirp Spread Spectrum (CSS) is simple, elegant,
  efficient
• Combines DSSS and UWB strengths
• Adds precise location-awareness
• Robustness multipath, interferers, correlation,
  FEC, 3 channels, CCA
• Can be implemented with todays technologies
• Low-complexity
• Low-cost
• Low power consumption
• Globally certifiable
• Scalability with many options for the future
• Backward compatible with 802.15.4-2003
• Mobility enhanced
• No MAC changes (minimal for ranging)
• Size and Form Factor meets or exceeds
  requirements
• Excellent throughput
• SOPs FD channels
• Signal Acquisition excellent
• Link Budget and Sensitivity excellent
• Power Management and Consumption - meets or
  exceeds requirements

50
Summary

• Chirp Spread Spectrum (CSS) Proposal Meets the
  PAR and 5C
• Precision ranging capability accurate to one
  meter or better
• Extended range over 802.15.4-2003
• Enhanced robustness over 802.15.4-2003
• Enhanced mobility over 802.15.4-2003
• International standard
• Ultra low complexity (comparable to the goals for
  802.15.4-2003)
• Ultra low cost (comparable to the goals for
  802.15.4-2003)
• Ultra low power consumption (comparable to the
  goals for 802.15.4-2003)
• Support coexisting networks of sensors,
  controllers, logistic and peripheral devices in
  multiple compliant co-located systems.