IEEE 802.15 subject

1
Project IEEE P802.15 Working Group for Wireless
Personal Area Networks (WPANs)
Submission Title Header Length Comments D
ate Submitted 8 May, 2005 Source Vern Bre
thour Company Time Domain Corp.
Address 7057 Old Madison Pike Suite 250
Huntsville, Alabama 35806 USA
Voice(256) 428-6331, FAX [phone],
E-Mail vern.brethour_at_timedomain.com
Re 802.15.4a. Abstract Companion discus
sion for a spreadsheet contributed as
IEEE802.15-05-0245. Purpose To provoke a dis
cussion of header lengths in 802.15.4a.
Notice This document has been prepared to assis
t 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 contrib
utor acknowledges and accepts that this
contribution becomes the property of IEEE and may
be made publicly available by P802.15.
2
One of the most important decisions we will make
is picking the length of the packet header.
How much time for the header?
Data (to include the time stamp of when the
delimiter was at the antenna of the transmitter.
Channel sounding
Acquisition
A delimiter signaling event separates the header
from the rest of the packet.
3
The length of the packet header plays a huge
roll in determining our long range positioning
performance.

• Our Standard is about the signal on the air.
• The Signal on the air must support our
  performance targets.
  
• Yet our performance is also largely determined by
  the receiver.

4
Simulations are best for predicting performance

• Even simulators are costly, so we need something
  quick and simple to pick an initial direction.
  
• This is a companion document to 0245r0, which is
  a spreadsheet to quickly evaluate the impact of
  architectural trade-offs.

5
The 0245r0 spreadsheet is full of assumptions
about the receiver architecture.

• The receiver is NOT part of the standard.
• I would love to ignore the receiver, but The
  receiver does exist and it has performance
  determining properties.
  
• So this discussion will include a reference
  receiver.

6
Include a reference receiver ..okay, but
first.. disclaimers!

• This is not and will not be part of the
  standard.
  
• There are lots of ways to build a radio.
• There is absolutely no claim here that this
  reference receiver is the best way to build a
  15.4a radio.
  
• This is simply a structure to put the performance
  spreadsheet into some context.
  

7
Reference Receiver for 0245r0
(in phase data stream)
RF Front End LNA, Band definition filter
s, etc.
A2D (1024 M Sa/s)
Low pass filter
I
Rectangular to polar transform
Magnitude (to Acquisition Ranging)
Local Oscillator _at_ Tx center frequency
Phase (to tracking)
90 degree phase shift
A2D (1024 M Sa/s)
Low pass filter
Q
(quadrature data stream)
8
This is the spreadsheet contributed as 05-0245r0
The cover sheet is not interesting. Go to the
sheet named Computation
9
How do we use the spreadsheet?
We make decisions and trade off numbers in this
part of the spreadsheet
While we keep our eye on these two answers These
are the projected preamble lengths needed to
satisfy the conditions
10
So, where do the numbers come from?
Band sketch from Welborn 0240r0
-40
-45
-50
dBm/MHz
-55
-60
Center Frequency matters the performance can be
different in each band
-65
-70
2
2.5
3
3.5
4
4.5
5
5.5
6
9
Frequency
x 10
11
More numbers
Band sketch from Welborn 0240r0
-40
-45
-50
dBm/MHz
-55
Depending on how much pulse shaping we do, the
3dB Tx bandwidth might only be 63 of the 10 dB
bandwidth. We must convert, because most of the
calculations are with respect to a 3 dB bandwidth.
-60
-65
-70
2
2.5
3
3.5
4
4.5
5
5.5
6
9
Frequency
x 10
12
More numbers
The pass band of the low pass filter is often
wider than the incoming signal envelope
bandwidth. That allows in more noise, which we
account for with this cell.
13
More numbers This one is the Biggie!
For performance in long links, in simulations, in
spreadsheets, and in real life, this number is
dominant. A free space channel has a path loss
exponent of 2. A moderately nasty indoor
channel has a path loss exponent of 3. A
really nasty channel can have a path loss
exponent much higher. I would feel better if
this cell contained the value 3. But if it
did, the spreadsheet would say that the 50 meter
link does not work in 4 ms. The value 2.6
represents a benign channel. This needs
discussion.
14
More numbers Acquisition S/N
For the purposes of Mr. Boltzmann, this number
really is just the system ambient temperature,
not the junction temperature of the devices in
the LNA. (Thank goodness! We capture that other
nasty stuff separately in the Noise Figure two
cells below. The Noise Figure is not a simple
function of temperature, so we just make it a
number and dont even try to model thermal
effects in the NF.)
15
More numbers
By the time we get done building the Tx and take
it to the compliance test lab, the output
spectrum will never be as smooth as the blue
curves. We must then back off the Tx power
across the entire band to keep the worst little
spike below the FCC emissions mask.
Band sketch from Welborn 0240r0
-40
-45
-50
dBm/MHz
-55
-60
-65
-70
2
2.5
3
3.5
4
4.5
5
5.5
6
9
Frequency
x 10
16
More numbers
A 7dB noise figure will sound high to people used
to narrow band radios. This is UWB , and wer
e targeting a system we can build in CMOS
17
More numbers Acquisition S/N
This number is an estimate of the
post-integration S/N needed to acquire with a
high probability of detect as well as a low
probability of false alarm. Even after extensive
simulations, it is often hard to get consensus on
this number. Its certainly more than 6 dB. 9
dB is a reasonable guess. Others are free to
make their own guesses.
18
More numbers Acquisition S/N
This number is hard for me to distinguish from
the S/N in the cell directly above. Some people
like to manage issues like degradation due to
oscillator drift during the integration period
with a separate number. This spreadsheet is
organized to please those people. In this
spreadsheet, this number and the one in the cell
above are never used separately but rather always
used as a summed pair.
19
Another key number S/N for leading edge.
The channel sounding is characterized by looking
at magnitude information. But what algroithm is
used for this is another issue with the reference
receiver.
What algorithm?
20
Algorithm for characterization of LOS.
Were trying to find this leading edge energy in
the channel sounding.
An indoor channel sounding.
21
Lets think about the problem in free space
Artists concept of a raised cosine envelope
Base band envelope (500 MHz) mixed to DC.
About 5 ns for 500 MHz
22
Consider finding the leading edge in free space
only one arriving pulse envelope.
Base band envelope (500 MHz) mixed to DC.
Sample times (1 GHz)
Actual Samples
Correct answer for position of leading edge
23
How do we find the green arrow?
500 MHz base band envelope mixed to DC. and
sampled at 1 GHz
Correct answer for position of leading edge
(The elusive green arrow)
One popular algorithm simply finds the first
non-zero (in practice, above some threshold)
value and calls that sample position the location
of the leading edge. In this example, that
algorithm would say the leading edge is here.
24
Alternative algorithm Find the green arrow!
Do some math calculate this position.
Correct answer for position of leading edge
Another algorithm uses the first two non-zero (in
practice, above some threshold) values and does
trig computations knowing that they are samples
of a known length cosine to calculate the
location of the leading edge.
25
Find the LOS path we have choices!
Algorithm 1
Pick the first value above a threshold and call
the leading edge position here.
26
Find the LOS path we have choices!
Algorithm 2
Do some math calculate this position.
27
Leading edge algorithms and ranging performance.
This is a receiver design issue. This is NOT a
recommendation about which algorithm to pick.
Pick 1st big one
Pick an algorithm 2 choices are shown here.
There are other choices as well.
Trig. look up table
28
Algorithm selection determines the S/N value.
The modeling of this algorithm is where this
particular number comes from.
trigonometry
29
Allowance for attenuation of the leading edge.
How much attenuation of the Line of Sight energy
will our algorithm tolerate? We make allowance
for that here.
30
LOS algorithm implementation loss.
This number captures stray effects like imperfect
tracking of oscillator drift during the channel
sounding and round-off errors in trig tables and
such distractions. I find it useful to
characterize the S/N needed for the algorithm
(two cells above) as if everything about the
implementation of the algorithm were perfect and
then account for imperfections separately here.
31
Chip time is an element of the computation.
32
Symbol time is an element of the computation.
The Barker 13 sequence is chosen as an example
10
12
13
1
2
3
4
5
6
7
8
9
11
33
The signaling scheme from Zafer 0223r0
One Bit
Always Empty
Always Empty
Always Empty
Time Hop freedom
This is our channel multipath tolerance.
13 chip times
The Other Bit
Always Empty
Always Empty
Always Empty
Time Hop freedom
This is our channel multipath tolerance.
10
12
13
1
2
3
4
5
6
7
8
9
11
34
Time hopping and multipath tollerance.
The Time Hop Freedom cell is to optionally
support the time hopping proposed by Zafer. I
set it to zero, to keep it out of the way in this
analysis.
Signaling from Zafer 0223r0
35
The Symbol repetition rate computation.
These numbers make up the inputs for the
computation of the symbol rep rate over here.
This gets multiplied by the integration count to
get the needed header times for acquisition and
channel sounding.
The final green cell number is the total of the
acquisition time and the channel sounding time.
36
The Path loss computation.
The path loss is determined as if the first meter
is free space. The antenna capture crossection
with different center frequencies comes in to
play here. For all ranges beyond the first
meter, the Path loss exponent is in effect.
37
The integration count computations.
These are the areas where a link budget style
computation is done to find the link margin
(negative at long ranges) that must be bought
back with integration gain for acquisition and
Line of Sight path arrival characterization.
38
What does the spreadsheet say?Bad news, mostly!
This number is small for conservative design
practice. (Would rather it be 3.)
These numbers are crowding our target sounding
times.
39
Conclusion We should stick with our ranging
performance targets, for now.

• 50 meter positioning to 1 meter accuracy in under
  10 ms (per round trip, with a small allowance for
  overhead) will be hard.
  
• 20 meter positioning to 1 meter accuracy in under
  2 ms (per round trip, with a small allowance for
  overhead) will be hard.
  
• These performance targets are only hard, not
  impossible.
  
• There are other positioning solutions in the
  marketplace, but if we hit these targets (or get
  close) we will bring unique value to our
  customers.