Micro-Doppler and vibrometry at millimeter

and sub-millimeter wavelengths.

Duncan A. Robertson* & Scott L. Cassidy
University of St Andrews, SUPA School of Physics & Astronomy,
St Andrews, Fife KY16 9SS, Scotland
ABSTRACT
Micro-Doppler is an emerging technique for the measurement and analysis of target modulation characteristics, rooted in
the analysis of X-band radar measurements of people. Despite the advantage of higher Doppler sensitivity, there appears
to be little such work reported at (sub-) millimeter wave frequencies. We have developed fully coherent, solid state,
FMCW radar systems operating at 94 and 340 GHz, suitable for micro-Doppler and vibrometry studies (as well as
SAR/ISAR), which make use of DDS chirp generation combined with upconversion and MMIC or Schottky diode
frequency multiplication. Due to the low phase noise architecture, the phase (i.e. displacement) sensitivity can be below
1 micron in distance.
Keywords: Radar, micro-Doppler, vibrometry, millimeter wave, sub-millimeter, phase noise.

1. INTRODUCTION
Micro-Doppler is an emerging technique for the measurement and analysis of target modulation characteristics. The
topic emerged from analysis of X-band radar measurements of people and has led to a rapidly developing field of
research 1. However, despite the advantage of higher Doppler sensitivity at higher carrier frequencies, there appears to be
little work reported on micro-Doppler at millimeter or sub-millimeter wave frequencies.
The Doppler shift is proportional to the carrier frequency so operating at millimeter or sub-millimeter wave frequencies
ought to offer the ability to better resolve fine Doppler features, especially for relatively slow moving targets such as
humans, and this may in turn improve target classification. Additionally, Doppler radars (which must operate coherently)
are able to sense very small displacements in target position to much less than a wavelength, and this can be exploited in
vibrometry to reveal characteristic signatures of targets associated with internal vibrations or modulations e.g. engine
vibrations, breathing, heartbeat, speech, etc. The measurable phase, and hence displacement, error is a function of the
signal to noise ratio (SNR) which in the thermal noise limited case is given in radians by

1
2 SNR

(1)

Figure 1 illustrates these two aspects of the Doppler effect. Figure 1 (left) shows the Doppler shift per 1 ms-1 radial
velocity as a function of radar frequency. Figure 1 (right) shows the displacement error in microns as a function of radar
frequency for different SNR values.
Clearly, very high displacement sensitivities, at the sub-micron level, are possible at higher radar frequencies although
these require very high SNRs and, in practice, physically very stable hardware setups. Over long path lengths,
atmospheric effects may become the limiting factor in the sub-millimeter wave range as the refractivity of air varies due
to turbulence arising from thermal and humidity fluctuations.

*dar@st-and.ac.uk; phone/fax +44 1334 467307; www.st-and.ac.uk/~mmwave

Radar Sensor Technology XVII, edited by Kenneth I. Ranney, Armin Doerry, Proc. of SPIE Vol. 8714, 87141C
2013 SPIE CCC code: 0277-786X/13/$18 doi: 10.1117/12.2015048

Proc. of SPIE Vol. 8714 87141C-1

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

100
Displacement Error [m]

Doppler Freq. for 1ms-1 [Hz]

2500
2000
1500
1000
500

20dB SNR
40dB SNR
60dB SNR

10

0.1

0
0

50

100

150

200

250

300

350

50

Frequency [GHz]

100

150

200

250

300

350

Frequency [GHz]

Figure 1. Doppler shift per 1ms-1 radial velocity versus radar frequency (left) and displacement error versus frequency
for different SNR values (right).

Despite the apparent benefits of performing micro-Doppler experiments at (sub-) millimeter wave frequencies, there
appears to have been relatively little such work reported, perhaps due to the popularity and relatively low cost of systems
at X-band, 2.4 GHz or other low frequencies. Some experimental results have been reported at frequencies of 27GHz 2,
35 GHz 3 and 36 GHz 4 for heartbeat monitoring, human stride rate estimation and intruder detection respectively. FOI,
Sweden, have made use of a 77 GHz radar, derived from automotive radar technology, to investigate micro-Doppler
from humans and its potential for gait parameter estimation and target classification 5, 6, 7. Very little work has been
reported at sub-millimeter wave (sometimes misnamed as THz) frequencies and Jin Li et al report micro-Doppler
simulations which assume a radar frequency of 340 GHz 8.
The Millimetre Wave Group at the University of St Andrews, Scotland, has specialized in the development of advanced
millimeter wave radars since 2002 and we wish report on our current developments of fully coherent solid state FMCW
radar systems operating at 94 and 340 GHz which are suitable for micro-Doppler and vibrometry studies. Additionally,
they can be applied to range-Doppler imaging and SAR/ISAR. The use of DDS chirp generation combined with
upconversion and MMIC or Schottky diode frequency multiplication schemes yields low phase noise and rapid,
contiguous chirps, necessary for Doppler studies and other coherent processing techniques. Our radars use FMCW
modulation and employ a solid state design which yields relatively compact, lower power and field deployable systems,
suitable for field trials and data gathering outside the lab.

2. COHERENT MILLIMETER & SUB-MILLIMETER WAVE RADARS

Due to the large Doppler shift associated with millimeter wave frequencies, the FMCW-Doppler technique can require a
high chirp repetition frequency (CRF) in order to unambiguously sample the maximum Doppler frequency of a moving
target. A high CRF can be difficult to generate whilst maintaining the other aspects of good chirp fidelity such as high
linearity and low phase noise.
The sequential dechirped signals must be close enough together to sample unambiguously the rate of change of phase
from chirp to chirp due to scatterer motion 9. This forces an upper limit on the duration of the chirp, T:

c
4 f 0 v max

(2)

where f0 is the carrier frequency of the radar, vmax is the maximum target velocity one wishes to measure and c is the
speed of light. For millimeter wave radars, with large f0 and thus high Doppler sensitivity, the limitation on T can be
severe. For example, in order that a 94 GHz radar can unambiguously sample a target moving at 10 ms-1 one requires a
chirp duration of no more than 78 s and thus a CRF of at least 12.5 kHz.
So in general, for millimeter wave FCMW-Doppler measurements, the chirp times are short (~few s) and CRFs are
high (~tens of kHz). Given that the chirps should ideally by contiguous this implies that a sawtooth waveform, with
instant flyback, is preferred. These temporal requirements for the chirp modulation are difficult to achieve with voltage

Proc. of SPIE Vol. 8714 87141C-2

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

controlled oscillators, especially when trying to maintain high linearity and low phase noise. Furthermore, the successive
chirps and the signal acquisition must be coherent in order for the phase history of the target to be recorded correctly.
2.1 Generic DDS-based chirp generator
Our approach is to generate the chirp using a DDS in the MHz range and translate it up to the (sub-) millimeter wave
carrier frequency using a combination of upconversion and MMIC or Schottky frequency multiplication. Specifically, we
upconvert the chirp from the DDS (an AD9910 clocked at 1 GHz) onto a microwave carrier at around ~7 GHz provided
by a stable local oscillator (STALO) prior to multiplication. The generic scheme we use is illustrated in Figure 1 (left).
-60
L(f) [dBc/Hz]

TRIG

-80

BPF

D
D
S

-100
-120
-140

CLK

-160
100

STALO

1000

10000 100000 1E+06 1E+07

Offset Frequency [Hz]

Figure 2. Upconverted DDS chirp generator architecture (left) and its measured CW phase noise at ~7 GHz (right).

This method maintains all the high speed dynamic frequency control and inherently high chirp linearity offered by the
DDS and yields a lower phase noise solution than alternative hybrid DDS/VCO/PLL schemes. In particular, the instant
flyback of the DDS sawtooth chirp modulation is preserved.
The CW single sideband (SSB) phase noise, L(f), of the DDS-based chirp generator was measured using an HP4352B
VCO/PLL Signal Analyzer phase noise measurement system. To access the input frequency range of the analyzer (< 6
GHz), an external mixer and low phase noise reference oscillator (a free running DRO) were used to mix down the chirp
generator output frequency of ~7 GHz. The SSB phase noise is plotted in Figure 1 (right). The phase noise of the radar
transmitter signal will be degraded from this level by the action of the frequency multiplier according to 20log(N).
The measured CW phase noise is extremely good for a source at ~7 GHz and outperforms VCO-based chirp generators
at the same frequency by ~20 dB. The phase noise of the upconverted DDS scheme is essentially governed by the phase
noises of the STALO and the DDS, both of which can be very low since the former is a fixed frequency oscillator and
the latter is operating at a relatively low frequency. For the VCO-based scheme, the achievable phase noise (at least
outside the loop bandwidth) tends to be set by the phase noise of the free-running VCO which, being a tunable oscillator,
will always be higher than that of a fixed frequency STALO of the same frequency.
2.2 Millimeter and sub-millimeter wave radars
We have developed a pair of coherent radars called NIRAD and IRAD which are capable of undertaking micro-Doppler
and vibrometry studies both of which employ the low phase noise, high speed, upconverted DDS chirp generator
architecture discussed above.
NIRAD is a 94GHz FMCW search radar with fast azimuth sweep rate. It is principally intended to provide a real-time
search function in which a narrow antenna beam is rotated rapidly in azimuth (up to 10 Hz rotation rate) and the resulting
high resolution plan position indicator (PPI) display yields a map of the surrounding scene capable of detecting and
tracking moving people and objects. Additionally, the radar can operate in staring mode, pointing in a single line of
sight, yielding a range profile (A-scope) display at very high update rates (up to 50 kHz).
IRAD is a 340 GHz FMCW 3D imaging radar with frame rates up to 10 Hz 10. The radar system is principally intended
to provide a near real-time 3D imaging function for concealed object detection in which an antenna beam, tightly
focused to a small spot at a standoff distance, is rastered in azimuth and elevation over a limited field of view containing

Proc. of SPIE Vol. 8714 87141C-3

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

a target of interest. IRAD can also operate in staring mode, pointing in a single line of sight, yielding a range profile (Ascope) display at very high update rates (up to 50 kHz).
Both NIRAD and IRAD employ a heterodyne architecture, single transmit / dual receive polarization, and a single
antenna. In both radars, the chirp generation is done at around ~7 GHz and the signals are then frequency multiplied up
to the carrier frequency. In NIRAD the multiplication factor is x12, done in MMIC multipliers, and in IRAD the
multiplication factor is x48, done in Schottky diode multipliers. The basic architecture of both radars is shown in Figure
3 and they are shown in photographs in Figure 4.
Tx Chirp

X12

LO1 Chirp

X12

Horn-Lens
Antenna

Duplexer

Mixer

IQ Demodulator

LO2

Beam
Scanner

Tx Chirp

x48

LO1 Chirp

x24

Beam
Scanner

QO
Duplexer

Focussing
Optics

SHM

IQ Demodulator

LO2

Range
Offset

LO3

ADC
ADC

Figure 3. NIRAD 94 GHz radar architecture (left) and IRAD 340 GHz radar architecture (right).

owr

Figure 4. Photographs showing 94 GHz NIRAD (left) and 340 GHz IRAD front view (centre) and rear view (right).

3. RADAR PHASE MEASUREMENTS

To evaluate the capabilities of NIRAD and IRAD for making micro-Doppler and vibrometry measurements we have
characterized their phase stability and undertaken a series of proof-of-principle measurements.
3.1 Measured phase errors
The measurable level of phase error was investigated with NIRAD by recording coherent range profiles of a target (a
+20 dBsm trihedral) at ~45 m range. These measurements were done out of the window of our lab with the target
situated outside whilst the weather was fair with little wind (<10 kmh-1). To ensure maximum stability, the target was
supported on a sturdy tripod which was also weighted down to minimize vibrations and the radar was braced from above
against the ceiling of the lab. A 102.4 s, 600 MHz bandwidth chirp was used at a CRF of 6.94 kHz.
An example of the measured phase history for a bright target with an SNR of 70 dB is shown in Figure 5 (left). The plot
shows the raw phase which exhibits some low frequency characteristics attributed to residual motion of the target tripod
and the same data after high pass filtering to suppress these effects. The peak phase deviation is less than 5 mrad and
the rms phase error is 0.75 mrad which corresponds to an rms displacement error of ~200 nanometers!

Proc. of SPIE Vol. 8714 87141C-4

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

To investigate the measurable phase error as a function of SNR, NIRAD was rotated in azimuth in small steps to vary the
received signal power, whilst the noise level remained constant. For each step, the SNR was calculated along with the
measured rms phase error. The results presented in Figure 5 (right) show extremely good agreement with the theoretical
response calculated from equation (1) and are consistent with results obtained with other millimeter wave radars 11.
40

Raw phase
HPF phase

10
5
0
-5
-10

RMS phase error [mrad]

Target phase [mrad]

15

Measurement

35

Theory

30
25
20
15
10
5
0

-15
0

0.2

0.4
Time [s]

0.6

25

0.8

35

45
55
SNR [dB]

65

75

Figure 5. Phase history for 70 dB SNR target (left). Measured & theoretical rms phase as a function of SNR (right).

From these results we can conclude that with a high SNR and a stable measurement configuration NIRAD is capable of
resolving phase to a few milliradians which corresponds to displacement sensitivities well below 1 micron in range. We
have also made this type of measurements with IRAD and found similar behavior although the phase measurements are
corrupted with a ~1 kHz sinusoidal modulation whose origin is uncertain but is believed to arise in the Schottky diode
multiplier bias circuits. When the sinusoid is suppressed from the data the rms phase error for a high SNR target is of
order a few mrad. Work is ongoing to correct this problem.
3.2 Vibrometry
An obvious exploitation of such high phase / displacement sensitivity is vibrometry and we have undertaken some basic
experiments with NIRAD to prove this capability. The reflection phase was recorded from a 6 diameter loudspeaker,
whose aperture had been covered with aluminum foil. The speaker was tripod mounted at ~45 m range and driven to a
clearly audible level by a 1300 Hz sine wave from a signal generator. The sound pressure level was not recorded. A
102.4 s, 600 MHz bandwidth chirp was used at a CRF of 8.065 kHz.
Figure 6 (left) shows a segment of the target phase history for the range bin containing the loudspeaker which has been
high pass filtered to remove wind effects below a few tens of Hz and shows that the sine wave is clearly resolved. Figure
6 (right) displays the Doppler spectrum (without high pass filtering) and confirms the 1300 Hz frequency, the high SNR
available and reveals the low frequency wind noise. The rms phase for the segment displayed in Figure 6 (left) is 12.6
mrad which is consistent with the 35 dB SNR evident in Figure 6 (right) and the theoretical curve in Figure 5 (right).
This corresponds to an rms displacement of the speaker foil of 3.2 m.
30
20

15
10
5

10
Amplitude [dB]

Target phase [mrad]

25
20

0
-5
-10
-15

0
-10
-20
-30
-40

-20
-25

-50

0.3

0.31

0.32

0.33

500

Time [s]

1000

1500 2000 2500

Fre quency [Hz]

3000

3500

Figure 6. Target phase from aluminum foil covered 6 loudspeaker playing a 1300 Hz sine wave acquired at 45 m range
with the NIRAD 94 GHz radar, showing high pass filtered time series (left) and raw Doppler spectrum (right).

Proc. of SPIE Vol. 8714 87141C-5

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

4000

We are continuing to explore vibrometry and once IRAD has its phase corruption problem resolved it should prove to be
a valuable test-bed for investigating vibration signatures of targets, not least because its beam is focused to a 30 mm
diameter spot at a range of 20 m, enabling us to isolate the location within a larger target from which vibrations emanate.
3.3 Micro-Doppler
The high phase sensitivity of our millimeter and sub-millimeter wave radars also make them ideal for exploring the
micro-Doppler signatures of humans and we have undertaken some proof-of-principle measurements of this type using
the 94 GHz NIRAD radar. NIRAD has a 0.8 pencil beam such that at the ranges accessible from our lab window (up to
~50 m) the beam only covers about half a person. Whilst this illuminates only the arms and torso or the legs, it has the
advantage of isolating the target from ground clutter. We aim to perform micro-Doppler measurements at longer ranges
at another location which will allow the full illumination of a person. Data were acquired with 102.4 s chirps of 120
MHz bandwidth (1.25 m range bins) at a CRF of 8 kHz, yielding a maximum unambiguous Doppler velocity of 6.4 ms-1.
Figure 7 presents time series received power (left) and unwrapped phase (right) for a person standing at a range of ~40 m
pushing both hands backwards and forwards. Behind the person there is a wall of a building at ~50 m. The received
power traces clearly show that the wall reflection is very stable in power but the received power for the person fluctuates
substantially. The phase data reveals a more significant result which is that the wall phase is very stable whilst the
persons phase exhibits a clear sinusoidal modulation, indicative of the motion of the arms. Figure 8 presents this same
data as a range-Doppler plot (left) and a micro-Doppler plot (right) for the range bin containing the person.
40

50

30

20

7,

-10

-200

-20

250

-30

-300

40

0.5

1.5

2
2.5
Time (seconds)

3.5

45

-3500

0.5

1.5

2
2.5
Time (seconds)

3.5

A5

Figure 7. Received power (left) and unwrapped phase (right) recorded from a wall at ~ 50 m (green) and a person
standing at ~40 m pushing both hands backwards and forwards (blue), acquired with the 94 GHz NIRAD radar.

Figure 8. Range-Doppler plot (left) and micro-Doppler plot (right) for person standing at ~40 m pushing both hands
backwards and forwards, acquired with 94 GHz NIRAD radar.

Proc. of SPIE Vol. 8714 87141C-6

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

The range-Doppler plot shows the wall as a bright target at zero velocity and longer range whilst the person is closer and
exhibits both positive and negative velocities due to the arm and torso motion. This is more clearly revealed in the microDoppler plot which shows that torso wobbles slightly in anti-phase to the arm motion.
Figure 9 shows a range-Doppler plot (left) and a micro-Doppler plot (right) for a person marching towards the radar with
arms swinging alternately in a conventional fashion. Once again the range-Doppler plot shows the wall as a bright target
at zero velocity but the person appears at a net positive velocity due to them approaching the radar. Their velocity
signature is quite spread in both directions due to the alternately swinging arms. The micro-Doppler plot reveals this
classic behavior of the arm swing and the slight bounding motion of the torso occurring between footsteps.

Figure 9. Range-Doppler plot (left) and micro-Doppler plot (right) for person marching towards the radar from ~45 m,
acquired with 94 GHz NIRAD radar.

The micro-Doppler data which can be acquired with the 94 GHz NIRAD radar is rich in detail due to the high Doppler
sensitivity afforded by the millimeter wave carrier frequency. We aim to use this to investigate a range of target motion
phenomena. Once the 340 GHz IRAD radar is fully operational for phase measurements it should be invaluable for
investigating localized micro-Doppler effects within complex targets and clutter.

4. CONCLUSIONS
Micro-Doppler and vibrometry require coherent measurements of the Doppler signature of targets. Improved Doppler
sensitivities are available at (sub-) millimeter wave frequencies which in turn require fast, low noise, high linearity chirp
generators for FMCW measurements. We have demonstrated a chirp generator scheme using an upconverted DDS chirp
which meets these requirements and is employed in heterodyne FMCW radars operating at 94 and 340 GHz. The
measured phase sensitivity of the 94 GHz NIRAD radar has been shown to be extremely high and in agreement with
theory, given sufficient SNR and a stable set up, with sub-micron displacements being detectable.
The high phase sensitivity of our radars has been exploited in demonstrations of vibrometry for the remote detection of
target vibration signatures and the micro-Doppler measurement of moving humans. The high fidelity results obtained are
proof of the benefits afforded by making such measurements at millimeter or sub-millimeter wavelengths.

5. ACKNOWLEDGEMENTS
The authors thank the UK Engineering and Physical Sciences Research Council and the Natural Environment Research
Council for financial support. The authors gratefully acknowledge the substantial contribution to this work received from
Patrick Beasley, QinetiQ, Malvern, UK.

Proc. of SPIE Vol. 8714 87141C-7

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx

REFERENCES
[1] Chen, V.C., The Micro-Doppler Effect in Radar, Artech House (2011).
[2] Yanming Xiao, Changzhi Li, Jenshan Lin, Accuracy of A Low-Power Ka-Band Non-Contact Heartbeat
Detector Measured from Four Sides of A Human Body, Microwave Symposium Digest, 1576 1579, (2006).
[3] Tahmoush, D. & Silvious, J., Stride Rate in Radar Micro-Doppler Images, Int. Conf. Systems, Man, and
Cybernetics, 4218 4223 (2009).
[4] Nanzer, J.A., Anderson, M.G., Josserand, T.M., Kuan Li, Olinger, G.A., Buhl, D.P. & Rogers, R.L., Detection
of moving intruders from a moving platform using a Ka-band continuous-wave doppler radar, Antennas and
Propagation Society International Symposium, 1 4 (2009).
[5] Bjorklund, S., Petersson, H., Nezirovic, A., Guldogan, M.B., & Gustafsson, F., Millimeter-wave radar microDoppler signatures of human motion, International Radar Symposium (IRS), 167 174 (2011).
[6] Guldogan, M.B., Gustafsson, F., Orguner, U., Bjorklund, S., Petersson, H. & Nezirovic, A., Human gait
parameter estimation based on micro-doppler signatures using particle filters, Int. Conf. Acoustics, Speech and
Signal Processing, 5940 5943 (2011).
[7] Bjorklund, S. ; Johansson, T. ; Petersson, H., Evaluation of a micro-Doppler classification method on mmwave data, IEEE Radar Conference, 0934 0939 (2012).
[8] Jin Li, Yiming Pi, Xiaobo Yang, Micro-Doppler Signature Feature Analysis in Terahertz Band, Journal of
Infrared, Millimeter, and Terahertz Waves, 31(3), 319 328 (2010).
[9] Middleton, R.J.C, Macfarlane, D.G., Robertson, D.A., High range resolution moving target indication at long
stand-off ranges using minimal sampling rates in linear FMCW radar, 35th Int. Conf. Infrared Millimeter and
Terahertz Waves (IRMMW-THz), doi: 10.1109 / ICIMW.2010.5612329, (2010).
[10] Robertson, D.A., Marsh, P.N., Bolton, D.R., Middleton, R.J.C., Hunter, R.I., Speirs, P.J., Macfarlane, D.G.,
Cassidy, S.L., and Smith, G.M., 340 GHz 3D radar imaging test bed with 10 Hz frame rate., Proc. SPIE 8362,
836206, (2012).
[11] Beasley, P.D.L., Advances in millimetre wave FMCW radar, Microwaves, Radar and Remote Sensing
Symposium, 246 249 (2008).

Proc. of SPIE Vol. 8714 87141C-8

Downloaded From: http://proceedings.spiedigitallibrary.org/ on 04/12/2016 Terms of Use: http://spiedigitallibrary.org/ss/TermsOfUse.aspx