PDOA Primer

Calculating position directly

from range and phase
difference of arrival at two
antennas

Version 1.3

This document is subject to change without

notice

© Decawave 2018 PDOA Primer 1.3

Decawave Confidential
PDOA Primer

TABLE OF CONTENTS
1 INTRODUCTION ..................................................................................................................... 3
2 PDOA SETUP ......................................................................................................................... 4
3 DISCUSSION .......................................................................................................................... 5
3.1 ADJUSTING PATH DIFFERENCE FOR ANTENNA EFFECTS.......................................................................... 7
3.2 EFFECTS OF TEMPERATURE ON PHASE DIFFERENCE. ........................................................................... 10
3.3 SOME TEST RESULTS .................................................................................................................... 12
4 LIMITATIONS AND FUTURE WORK ....................................................................................... 14
4.1 LIMITATIONS .............................................................................................................................. 14
4.1.1 Phase difference operational range ................................................................................ 14
4.1.2 Tag elevation ................................................................................................................... 14
4.2 FUTURE WORK ............................................................................................................................ 14
4.2.1 New PDOA antenna array designs .................................................................................. 14
4.2.2 Temperature tests ........................................................................................................... 14
APPENDIX A: PDOA POST-PROCESSING METHODOLOGY ............................................................. 15
5 DOCUMENT HISTORY .......................................................................................................... 17
5.1 REVISION HISTORY ...................................................................................................................... 17

LIST OF TABLES

TABLE 1: DOCUMENT HISTORY .................................................................................................................... 17

LIST OF FIGURES

FIGURE 1 APPARATUS TO CALCULATE POSITION ................................................................................................ 4

FIGURE 2 A RADIO SIGNAL ARRIVES AT TWO ANTENNAE SEPARATED BY D .............................................................. 5
FIGURE 3 RANGE VARYING IN 30CM STEPS TO 8M, PHASE DIFFERENCE VARYING IN 20⁰ STEPS ................................. 7
FIGURE 4 PATH DIFFERENCE DISTORTION. PATH DIFFERENCE IS IN UNITS OF CENTIMETRES. ...................................... 9
FIGURE 5 EXAMPLE PATH CALIBRATION CURVES FOR THREE BOARDS................................................................... 10
FIGURE 6 HOW PHASE DIFFERENCE CHANGES AS BOARD 1 WARMS UP ................................................................ 11
FIGURE 7 SMALLER PHASE DIFFERENCE CHANGES AS BOARD 2 WARMS UP ........................................................... 12
FIGURE 8 EXAMPLE RESULTS FROM BOARD 1 TAKEN EVERY 15° IN A CIRCLE ......................................................... 13
FIGURE 9 EXAMPLE RESULTS FROM BOARD 2 TAKEN EVERY 15° IN A CIRCLE ......................................................... 13
FIGURE 10 RAW PDOA DATA WITH JUMPING ................................................................................................ 15
FIGURE 11 RAW PDOA DATA AND CORRECTED PDOA DATA ............................................................................ 16

© Decawave 2018 PDOA Primer 1.3 Page 2

Decawave Confidential
PDOA Primer

1 INTRODUCTION
Using two receiver antennas, the location of a transmitter can be found by using time of flight
to get the distance to the transmitter and using the difference in phase of arrival, to calculate
the relative X/Y position of the transmitter.

The path difference can be found by using the difference in the local oscillator phase at the
time of arrival of the earliest signal path of a received frame at two of the antennas.

© Decawave 2018 PDOA Primer 1.3 Page 3

Decawave Confidential
PDOA Primer

2 PDOA SETUP

Figure 1 Apparatus to calculate position

© Decawave 2018 PDOA Primer 1.3 Page 4

Decawave Confidential
PDOA Primer

3 DISCUSSION

Baseband
Transmitter

Carrier

Tx Antenna C

𝑟𝑟
𝑦𝑦
𝑟𝑟 − 𝑝𝑝

𝛼𝛼 𝛽𝛽
𝑑𝑑 𝑥𝑥 − 𝑑𝑑
Antenna A Antenna B
𝑥𝑥
Figure 2 A radio signal arrives at two antennae separated by d

In this scenario, the signal travels a distance 𝑟𝑟 to get to antenna A and a distance 𝑟𝑟 − 𝑝𝑝 to
get to antenna B. The distance between the two antennas is 𝑑𝑑.

We can use well known methods to find the distance r between antenna A and antenna C,
e.g. measuring the time of flight between A and C.

Provided that the distance d between two antennas is less than or equal to one half
wavelength, 𝜆𝜆/2, of the radio signals transmitted from antenna C, the path difference p will
always be between − 𝜆𝜆/2 and + 𝜆𝜆/2.

So, if we can measure the phase of arrival at each antenna, the phase difference going from
-180⁰ to +180⁰ can give us a path difference varying from -𝜆𝜆/2 to +𝜆𝜆/2. We wish to find the
𝑥𝑥, 𝑦𝑦 location of the transmitter. So, we want to find the distances 𝑥𝑥 and 𝑦𝑦.
© Decawave 2018 PDOA Primer 1.3 Page 5
Decawave Confidential
PDOA Primer

From cosine rule:

𝑏𝑏 2 + 𝑐𝑐 2 − 𝑎𝑎2
𝑐𝑐𝑐𝑐𝑐𝑐(𝐴𝐴) =
2𝑏𝑏𝑏𝑏

𝑟𝑟 2 + 𝑑𝑑2 − (𝑟𝑟 − 𝑝𝑝)2

𝑐𝑐𝑐𝑐𝑐𝑐(𝛼𝛼) =
2𝑟𝑟𝑟𝑟

𝑥𝑥 𝑟𝑟 2 + 𝑑𝑑2 − 𝑟𝑟 2 + 2𝑟𝑟𝑟𝑟 − 𝑝𝑝2

=
𝑟𝑟 2𝑟𝑟𝑟𝑟
𝑑𝑑 2 +2𝑟𝑟𝑟𝑟−𝑝𝑝2
𝑥𝑥 = 2𝑑𝑑
(1)

𝑝𝑝 𝑝𝑝 𝑑𝑑
𝑥𝑥 = �𝑟𝑟 − � + (1a)
2 𝑑𝑑 2

𝑟𝑟 2 = 𝑥𝑥 2 + 𝑦𝑦 2

𝑦𝑦 = ±√𝑟𝑟 2 − 𝑥𝑥 2 (2)

Alternatively:
2
2
𝑑𝑑2 + 2𝑟𝑟𝑟𝑟 − 𝑝𝑝2
2
𝑥𝑥 + 𝑦𝑦 = � � + 𝑦𝑦 2 = 𝑟𝑟 2
2𝑑𝑑

2
2 2
𝑑𝑑2 + 2𝑟𝑟𝑟𝑟 − 𝑝𝑝2
𝑦𝑦 = 𝑟𝑟 − � �
2𝑑𝑑

�(𝑑𝑑2 − 𝑝𝑝2 )((2𝑟𝑟 − 𝑝𝑝)2 − 𝑑𝑑2 )

𝑦𝑦 = ±
2𝑑𝑑

�(𝑑𝑑2 − 𝑝𝑝2 )(4𝑟𝑟 2 − 4𝑟𝑟𝑟𝑟 + 𝑝𝑝2 − 𝑑𝑑2 )

𝑦𝑦 = ±
2𝑑𝑑

2
�(1 − �𝑝𝑝� )(4𝑟𝑟 2 − 4𝑟𝑟𝑟𝑟 + 𝑝𝑝2 − 𝑑𝑑2 )
𝑑𝑑
𝑦𝑦 = ±
2

We cannot tell if the 𝑦𝑦 coordinate is positive or negative, so there is a front/back ambiguity.

Simplifying this, we note that 𝑑𝑑 is small compared to 𝑟𝑟 so 𝑑𝑑2 is very small compared to 𝑟𝑟 2 and
can be neglected.

2
�(1 − �𝑝𝑝� )(4𝑟𝑟 2 − 4𝑟𝑟𝑟𝑟 + 𝑝𝑝2 )
𝑑𝑑
𝑦𝑦 ≈ ±
2

© Decawave 2018 PDOA Primer 1.3 Page 6

Decawave Confidential
PDOA Primer

𝑝𝑝 𝑝𝑝 2
𝑦𝑦 ≈ ±(𝑟𝑟 − )�1 − � � (3)
2 𝑑𝑑

The maximum error by using the approximation is 0.22mm for a 6.5GHz carrier and a
receiver antenna separation of 𝜆𝜆/2.

Thus, using equations (1) and (2), or (1) and (3), we can calculate the 𝑥𝑥, 𝑦𝑦 coordinates of the
transmitter. We just need to know:

(1) 𝑟𝑟 – the range to one of the antennas

(2) 𝑑𝑑 – the distance between the two antennas
and
(3) 𝑝𝑝 – the path difference for the signal arriving at the antennas

One of the most accurate ways to get the path difference is to get the phase difference of
arrival of a signal in fractions of a cycle and then multiply by the wavelength.

Figure 3 Range varying in 30cm steps to 8m, Phase Difference varying in 20⁰ steps

We can see from this figure that the position uncertainty at phase differences near +/- 180 is
quite large. A very small change in phase gives a large change in 𝑦𝑦 position.

3.1 Adjusting path difference for antenna effects

In a real system, when the two antennas are closer than a few wavelengths apart, the
antennas interact through an effect known as mutual coupling. This causes the
electromagnetic waves to behave differently than they would in free space, which in turn
causes the effective path difference to be different then the geometric path difference.

Another effect seen in the real world is that the feed wires to the antennas can have slightly
different lengths or the paths from the down-mixer generators to the two separate down-
mixers can have slightly different delays. These two effects, and others, add a constant
© Decawave 2018 PDOA Primer 1.3 Page 7
Decawave Confidential
PDOA Primer

offset to the path difference. This difference can be quite large in practice, up to +/- half a
wavelength.

These effects can be calibrated out by having a conversion function between measured and
geometric path difference. A system calibration task is carried out, where a number of
measurements are taken from a number of different known coordinates with a wide range of
geometric path differences. Since we know the true 𝑥𝑥, 𝑦𝑦 coordinates, we can calculate this
true geometric difference. By measuring the phase difference, we can also calculate the
effective or measured path difference. In this way we can build a calibration function e.g. by
having a look-up table or by using a piecewise linear function or by using a polynomial fitting
function.

If using a polynomial fit, as we use in the PDOA evaluation kit, the best results are obtained
if the path difference is measured with the transmitter straight in front of the two receive
antennas, i.e. 𝑝𝑝 = 0 ,because the transmitter is equidistant from both antennas. This
constant offset, 𝑝𝑝0 , is then always subtracted from the measured path difference.

Then we can use this calibration function to correct the residual path difference before
applying the formulas to find the 𝑥𝑥, 𝑦𝑦 coordinates. An example of a possible function is
shown in figure 7 and is best fit by the following 4th order polynomial.

𝑝𝑝𝑔𝑔 = − 0.0222𝑝𝑝𝑚𝑚 4 + 0.0328𝑝𝑝𝑚𝑚 3 3 + 0.0729𝑝𝑝𝑚𝑚 2 + 0.854𝑝𝑝𝑚𝑚 + 0.0111

Where pg is the geometric path difference and pm is the measured path difference, both in
centimetres.

© Decawave 2018 PDOA Primer 1.3 Page 8

Decawave Confidential
PDOA Primer

Figure 4 Path difference distortion. Path difference is in units of centimetres.

For the integrated dual antenna supplied with the PDOA kits, the best calibration function
was found to be a straight line, 𝑦𝑦 = 𝑥𝑥.

© Decawave 2018 PDOA Primer 1.3 Page 9

Decawave Confidential
PDOA Primer

Figure 5 Example path calibration curves for three boards

3.2 Effects of temperature on phase difference.

The phase difference can vary with temperature, this is board/device dependent effect. The
effect can be seen by comparing Figure 6 and Figure 7, the board used to gather the data
for Figure 6 is more sensitive to temperature. Some of the tests were run starting from the
receiving node being turned off and some continuing from a previous measurement.

© Decawave 2018 PDOA Primer 1.3 Page 10

Decawave Confidential
PDOA Primer

Figure 6 How phase difference changes as board 1 warms up

Figure 6 shows the phase difference, after the constant offset, 𝑝𝑝0 , of 97° was subtracted for
one particular board. The transmitter was kept at a constant distance from the receiver node
and the node was rotated by 15° between each set of measurements (rotation shown in
coloured typeface).

© Decawave 2018 PDOA Primer 1.3 Page 11

Decawave Confidential
PDOA Primer

Figure 7 Smaller phase difference changes as board 2 warms up

Figure 7 shows the same thing as Figure 6 but for a different board. In this case after a
constant offset, 𝑝𝑝0 , of 23° was subtracted.

Experiments have confirmed that, for a given device, at a given temperature, 𝑝𝑝0 is a
constant, but that it varies from device to device, that it varies with temperature, and that the
rate of change with temperature also varies from device to device.

3.3 Some test results

In the following tests, the transmitter was kept at a constant distance from the receiver node
which was rotated by 15° between each set of measurements. Results for each rotation are
colour coded and the white circles in the centre of the scatter plots is the actual relative
position of the transmitter.
© Decawave 2018 PDOA Primer 1.3 Page 12
Decawave Confidential
PDOA Primer

Figure 8 Example results from board 1 taken every 15° in a circle

Figure 9 Example results from board 2 taken every 15° in a circle

It can be seen from these example results that the spread gets larger the further away the
transmitter is from being directly in front of the receiver node.

© Decawave 2018 PDOA Primer 1.3 Page 13

Decawave Confidential
PDOA Primer

4 LIMITATIONS AND FUTURE WORK

4.1 Limitations

4.1.1 Phase difference operational range

The absolute phase difference range for which the 𝑥𝑥, 𝑦𝑦 coordinates can be successfully
estimated is up to 160°. The PDOA antenna array is not designed to measure PDOA in its
back half-hemisphere, which faces its ground plane.

4.1.2 Tag elevation

The supplied PDOA antenna array has two elements. As such, it can measure the 𝑥𝑥, 𝑦𝑦
coordinates in only two dimensions. These two dimensions are the plane defined by three
points; the positions of the two PDOA node array elements and position of the tag. When the
tag is elevated above the azimuth plane of the antenna, the calculated 𝑥𝑥, 𝑦𝑦 coordinates in
this plane will not correspond to the 𝑥𝑥, 𝑦𝑦 coordinates in the azimuth plane of the antenna. To
estimate the 𝑥𝑥, 𝑦𝑦, 𝑧𝑧 coordinates of the tag, more complex antenna array geometries are
required with at least three elements.

4.2 Future work

4.2.1 New PDOA antenna array designs

Further antenna optimization is on-going to reduce the limitations mentioned.

4.2.2 Temperature tests

The DW1000 has a temperature monitor. The performance of the kit is being tested at
different temperatures on many different devices with the goal of relating the indicated
temperature to the constant offset, 𝑝𝑝0 .

© Decawave 2018 PDOA Primer 1.3 Page 14

Decawave Confidential
PDOA Primer

APPENDIX A: PDOA POST-PROCESSING METHODOLOGY

Let us consider a scenario, in which the kit is used at phase differences with absolute values
higher than 160 degrees. In this region, due to individual measurements exhibiting variations
from the mean, some of the PDOA measurements will go over the ±𝜋𝜋 phase discontinuity,
as shown in Figure 10. This will appear as jumping between positive and negative
𝑥𝑥 locations in the GUI. Furthermore, jumping will bias the mean of the measurements and
considerably increase the standard deviation.
In order to get the mean and the standard deviation of the PDOA on the circle, the point of
the phase discontinuity can be moved away from ±π using some current estimate of PDOA.
For example, if the current estimate of PDOA is close to π, allowed phases can be π/2 to
3π/2. The formula proposed to achieve this is
� /2�mod 2𝜋𝜋 − 𝜋𝜋 + 𝑃𝑃𝑃𝑃
𝑃𝑃𝐷𝐷𝑐𝑐 = �𝑃𝑃𝑃𝑃 + 𝜋𝜋 − 𝑃𝑃𝑃𝑃 � /2.
Here, phase difference in radians are:
• 𝑃𝑃𝑃𝑃 – current PDOA measurement, as calculated in Sec. Error! Reference source
not found..
� – current (filtered) estimate of 𝑃𝑃𝑃𝑃.
• 𝑃𝑃𝑃𝑃
• 𝑃𝑃𝐷𝐷𝑐𝑐 – corrected value of current PDOA measurement.
Note that the above formula allows prevention of jumping of PDOA and thus proper
calculation of PDOA mean and standard deviation on the circle. However, it does not
prevent jumping of the 𝑥𝑥 locations, since any given 𝑃𝑃𝑃𝑃 and corresponding 𝑃𝑃𝐷𝐷𝑐𝑐 result in the
same 𝑥𝑥 location estimate. Furthermore, the above method is not successful if jumping is
severe, i.e. if the numbers of positive and negative PDOA measurements are similar.

1
PDOA (rad)

-1

-2

-3

0 10 20 30 40 50 60 70 80
Measurement index

Figure 10 Raw PDOA data with jumping

�
Figure 11 shows PDOA jumping prevention of the raw data given in Figure 10, with the 𝑃𝑃𝑃𝑃
calculated as mean of the previous eight 𝑃𝑃𝑃𝑃 values.

© Decawave 2018 PDOA Primer 1.3 Page 15

Decawave Confidential
PDOA Primer

In laboratory measurements, where the 𝑥𝑥 location is known, instead of the above formula,
the following can be used:
𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(𝑥𝑥) 𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠(𝑥𝑥)
𝑃𝑃𝐷𝐷𝑐𝑐 = �𝑃𝑃𝑃𝑃 + 𝜋𝜋(1 − )� mod 2𝜋𝜋 − 𝜋𝜋(1 − )
4 4

2 Raw PDOA
Corrected PDOA

1
PDOA (rad)

-1

-2

-3

0 10 20 30 40 50 60 70 80
Measurement index

Figure 11 Raw PDOA data and corrected PDOA data

© Decawave 2018 PDOA Primer 1.3 Page 16

Decawave Confidential
PDOA Primer

5 DOCUMENT HISTORY
5.1 Revision History
Table 1: Document History

1.3 30-JUL-2018 Logo Update KD

1.2 30-May-2018 Tidy Up. MMcL
1.1 30-May-2018 Added Section 4 and Appendix A MMcL
1.0 29-May-2018 Initial release MMcL

© Decawave 2018 PDOA Primer 1.3 Page 17

Decawave Confidential