You are on page 1of 6

Characterizing the Impact of IQ Imbalance and DC

Bias on Pulse-Agile Radar Processing


Justin G. Metcalf and Shane Flandermeyer Charles A. Mohr ∗† , Andrew Kordik∗‡
School of Electrical and Computer Engineering Patrick McCormick∗† and Cenk Sahin∗
Advanced Radar Research Center ∗ AFRL Sensors Directorate, Dayton, Ohio
University of Oklahoma † University of Kansas, Radar Systems Lab, Lawrence, KS
Norman, Oklahoma ‡ Defense Engineering Corporation, Dayton, OH

Abstract—The advent of commercial-off-the-shelf (COTS) form of noise and/or complementary waveforms (e.g. [5]–
software-defined radars (SDRs) has enabled low-cost, flexible [7]). The second form of pulse agile waveforms are those
experimentation with emerging pulse-agile waveform designs to that introduce dynamic spectral ”gaps” in the transmitted
mitigate spectral congestion, improve radar performance, embed
communications, and other applications. We examine the impact waveform to reduce interference with primary communications
of direct digital downconversion sampling imperfections in the users of the spectrum [8]–[11]. The third class of pulse
form of IQ imbalance and DC bias on pulse-agile waveforms. A agile waveforms are so-called radar-embedded communication
general framework is developed and examples given for a par- waveforms, where a single waveform may be used to both
ticular radar-embedded communication waveform. It is shown perform radar detection and simultaneously transmit data [12]–
that the clutter response due to IQ imbalance is significantly
increased in a pulse agile framework. [15]. The structure of the pulse-to-pulse variation then defines
the mapping that encodes the information within each pulse.
I. I NTRODUCTION A key drawback common across these classes of pulse-
agile waveforms is the impact to the clutter Doppler response.
There are three standard radar receiver architectures: di- Specifically, by changing the waveform from pulse-to-pulse,
rect sampling, digital intermediate frequency (IF), and direct the clutter range sidelobe response likewise changes. As such,
digital downconversion (also known as a complex baseband the clutter experiences a range-sidelobe modulation (RSM)
or quadrature receiver, among other names) [1]. A direct that causes it to spread across Doppler. Considerable work
sampling receiver samples the real-valued received signal has been done in compensating for RSM [16]–[19].
directly at RF, while a digital IF receiver heterodynes down the While the impact of IQ imbalance has long been fully under-
real signal to an IF. In either of those two cases, the sampled stood in the context of a non-pulse agile radar [2], the authors
signal is then digitally downconverted to complex baseband are not aware of a first-principles sensitivity analysis for the
values [1]. In contrast a quadrature receiver uses two channels pulse-agile case. An abnormally high noise floor was noticed
90 degrees out of phase, called an in-phase and a quadrature in previous work that set in motion the present examination
channel, to directly downconvert to complex baseband and [20]. As will be shown, the impact of IQ imbalance on clutter
then sample both the real and imaginary components. in particular must be considered carefully when setting IQ
The imbalance of in-phase and quadrature channels was imbalance requirements. We will show that the IQ imbalance
a well-known problem for radar signal processing in early causes the clutter to spread in Doppler in a manner similar to
quadrature architectures [1]–[3]. DC bias due to local oscillator RSM.
(LO) leakage is a drawback common to both quadrature and
digital IF receivers. The emergence of high speed analog- II. C OMPLEX BASEBAND R ECEIVER
to-digital converters (ADCs) resulted in digital IF sampling In a traditional direct digital downconversion, a real input
becoming the dominant receiver architecture, as it naturally signal A(t) sin[Ωt + θ(t)] at center or intermediate frequency
mitigated the issue of IQ imbalance [1]. Ω is passed to a pair of mixers (denoted in-phase (I) and
However, budgetary and power constraints in the space quadrature (Q), respectively) are driven by an LO at Ω. The I
of commercial off-the-shelf (COTS) software defined radios and Q mixers are driven 90◦ out of phase and their output is
(SDRs) and digital arrays [4] have led to the re-emergence of low-pass filtered (LPF) to remove the upper sideband before
quadrature architectures. We note that COTs SDRs have pro- ADCs sample the signal. Finally, the quadrature component is
vided a new low-cost testing capability for an emerging group digitally multiplied by the square root of -1, j, and the sampled
of pulse-agile radar waveforms, which we generalize into three components are summed together. The structure of the direct
categories. The first category of pulse agile waveforms are digital downconversion is shown in Figure 1, with the ideal
those which utilize pulse-agility to provide diversity to the components shown in black. Note that the signals from the LO
range Doppler response and/or reduce interference between are multiplied by 2 in Figure 1 for mathematical convenience.
the radar and communications users. These typically take the

978-1-7281-6813-5/20/$31.00 ©2020 IEEE 129

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:58:04 UTC from IEEE Xplore. Restrictions apply.
Unfortunately, an ideal receiver only exists in theory. There receiver, including non-ideal components, is then [1]
are three primary sources of error within a quadrature receiver. y(t) = A(t) [cos[θ(t)] + γ + j [(1 + ) sin[θ(t) − φ] + κ]]
First, DC bias is typically present where a constant bias is
added to the output of the mixers due to leakage from the = A(t) {α cos[θ(t)] + jβ sin[θ(t)]} + (γ + jκ) (3)
local oscillators. Second, there may be an amplitude mismatch where
between the input paths to the mixers and/or the mixers α = 1 − j(1 + )sinφ (4)
themselves. Third, the mixers may be driven slightly out-
and
of-phase. The amplitude and phase mismatches between the
β = (1 + ) cos φ. (5)
mixer products are collectively termed IQ imbalance. For the
present work the impact of other receiver components, such as It can be shown that (3) can be simplified to [1]
 
anti-aliasing filters, are ignored. A more sophisticated receiver α+β α−β
model will be developed in future work. In addition, it is y(t) = A(t) exp[jθ(t)] + exp[−jθ(t)]
2 2
assumed that the IQ imbalance and DC bias are stable across
+ (γ + jκ). (6)
a coherent processing interval (CPI).
Due to the drawback of IQ imbalance, digital IF sampling is Therefore the IQ imbalance induces two separate compo-
the preferred downconversion technique. However, as the com- nents: the scaled desired component α+β 2 exp[jθ(t)] and an
α−β
plex baseband signal is asymmetric in the frequency domain, image component 2 exp[−jθ(t)]. The image component
the quadrature architecture permits the Nyquist sampling of the is therefore the complex conjugate of the desired signal and
complex baseband signal using two ADCs running at B sam- is scaled by a factor of α−β 2 . By superposition the desired
ples/second, where B is the bandwidth of the baseband signal component, image component, and DC bias may be examined
[1]. In contrast, a digital IF receiver requires a single ADC but separately. Consequently traditional analyses typically center
at a much higher sampling rate (e.g. a factor of 2-4 times B on the ratio of the power of the image component to the desired
[1]). Direct digital-downconverters have enjoyed a resurgence component:
2
in COTs systems due to their lower sampling requirements. Pi |(α − β) /2|
As such, ”what was old is new again”, and we must again Pr = = 2 (7)
Pd |(α + β) /2|
consider the impact of direct digital downconversion on radar
For a pulse agile radar the ratio of the image to the desired
signal processing.
signal component is unchanged. However, as will be shown
Without loss of generality all imbalance errors are defined
the structure of image component will induce a response that
relative to the in-phase channel. Therefore, the output of non-
will impact Doppler processing.
ideal quadrature channel is
III. I DEAL R ECEIVED R ADAR S IGNAL
Q(t) = A(t)(1 + ) sin[θ(t) − φ] + κ. (1) Next, consider the structure of the signal incident at the
radar receiver. In a pulse agile pulse-Doppler radar system,
a series of m = 1, 2, . . . , M pulse compression waveforms
while the output of the ideal in-phase channel is
sm (t) are transmitted at a constant pulse repetition interval
(PRI) TP RI to form a coherent processing interval (CPI). If a
I(t) = A(t) cos[θ(t)] + γ. (2) non-pulse agile waveform is used, then s1 (t) = s2 (t) = · · · =
sm (t). For a ground looking radar waveform there are three
These receiver non-idealities are shown in red components components of the received signal: the target component x(t),
in Figure 1. The output signal from the complex baseband the clutter component c(t), and the noise component u(t).

Fig. 1. Complex Baseband Receiver with IQ Imbalance and DC Bias

130

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:58:04 UTC from IEEE Xplore. Restrictions apply.
A. Ideal Received Signal for 0 ≤ t ≤ Tp , where Tp is the pulse length and k is the
The received signal is simply the superposition of these chirp rate.
three signals, sampled at intervals of TP RI for pulses m = There are numerous emerging pulse-agile radar waveform
1, .., M . designs. Expressions are derived for a general pulse agile
waveform, but for simulation results we examine one particular
y(t − mTP RI ) = x(t − mTP RI ) + c(t − mTP RI ) + u(t). (8) waveform, the continuous phase modulation-polyphase coded
The noise component is assumed to be white additive complex frequency modulated (CPM-PCFM) implementation of the
Gaussian distributed thermal noise, and is therefore indepen- phase-attached radar-communications (PARC) framework. The
dent from sample to sample. However, the target and clutter CPM-PCFM waveform is given as [13], [21]
components are not independent from pulse-to-pulse, hence  

the dependence on pulse number. sPARC (t; β̄, α, h, Ts ) = exp j ψ(t; β̄) + ξ(t; α, h, Ts )

The target signal x(t) is the superposition of the delayed = exp j(ψ(t; β̄) exp {ξ(t; α, h, Ts )}
echoes from K targets (assumed range and Doppler unam- (12)
biguous), where the target signal on the mth transmitted pulse
sm (t) is defined as where a pulse-to-pulse varying communication waveform
  exp {ξ(t; α, h, Ts )} is phase-attached

(i.e. embedded on) a
K
vr,i
x(t − mTP RI ) = ξi sm (t − τi )exp −j4π t . (9) base radar waveform exp j(ψ(t; β̄) . In order to maintain
i=1
λc similarity across pulses, the base radar waveform is unchanged
from pulse-to-pulse. For the CPM-PCFM implementation of
The two-way time delay to the ith target is τi , the target
PARC, the communication waveform is a CPM waveform
reflection coefficient is ξi and the imparted Doppler shift
with parameters α the communication sequence (in itself
is dependent on the relative radial velocity vr,i and the
dependent on a modulation order Mc ), h the modulation
wavelength of the center frequency λc .
index, and Ts the symbol time. The modulation index h in
The clutter signal c(t) is composed of the convolution of
particular controls the degree of variation
from

pulse-to-pulse.
the transmitted waveform s(t) and the clutter process z(t),
Further, the radar waveform exp j(ψ(t; β̄) is a continuous
assumed to be a zero-mean, stationary white complex Gaussian
FM waveform generated from phase code β̄. As with the LFM
process with variance σz2 . For the present examination it
pulse defined in (11), the PARC waveform of (12) is non-
is assumed that the clutter has no internal clutter motion,
zero for 0 ≤ t ≤ Tp . For further details of the CPM-PCFM
meaning the samples of z(t) are unchanged from pulse-to-
waveform implementation, see [13], [21].
pulse. Therefore, the clutter component for each pulse is given
For the purposes of simulation, all CPM-PCFM PARC
as
waveforms generated have a time-bandwidth product of 224
c(t − mTP RI ) = sm (t) ∗ z(t − mTP RI ), (10)
(i.e. sampling the 3 dB bandwidth B of the radar waveform
where ∗ denotes convolution. with sample time Tr requires Tr B = 224 complex baseband
After reception the sampled signal of (8) from each pulse samples). These baseband waveforms are oversampled by a
is matched filtered (i.e. convolved with hm (t) = s†m (−t), factor of NOS = 4, have a modulation index of h = 2 and
where (•)† denotes the complex conjugation operation) to a modulation order of Mc = 2 bits/symbol. The LFM wave-
pulse compress the range response and maximize the signal- forms are generated using a PCFM modulator with identical
to-noise ratio. To match to the target Doppler shifts of (9) a values of Tr B = 224 and NOS = 4. Unless otherwise noted,
discrete Fourier transform (almost always implemented as a all range Doppler simulations hereafter use a CPI of length
fast Fourier transform with a Doppler oversampling factor) is 64 pulses and a Doppler oversampling factor of 4. A Taylor
applied across the M pulses to match filter the range response Doppler window is applied with 8 equal magnitude sidelobes
to sampled Doppler frequency shifts [1]. and a maximum sidelobe level of -55 dB.
Doppler filtering is critical to the performance of ground
moving target indication (GMTI) radars, as the clutter response IV. I MPACT OF DC B IAS
is much stronger (e.g. 20-35 dB [1]) than the target responses,
but the clutter is confined to the zero-Doppler bin (or within While the primary drawback of direct digital receivers is
several low-Doppler bins if there is internal clutter motion). the IQ imbalance, it is instructive to examine the impact of
Suppression of the Doppler sidelobes of the clutter response DC bias. This analysis will help frame the characterization of
is commonly achieved via a Doppler window (or taper) [1]. the IQ imbalance. Note that the DC bias is a constant value
B. Radar Waveform Definition at all samples and is unaffected by the waveform transmitted.
Therefore, the DC bias is the vector (γ + jκ)1, where 1 is
For the simulations presented two waveform types are used
the column vector of all ones with length NRB equal to the
to provide context. For the non-pulse agile case, a linear fre-
number of samples in the unambiguous range. However, the
quency modulated (LFM) upchirp is used, where the complex
match filtered (i.e. pulse compressed) response of the DC bias
baseband waveform is defined as
   is dependent on the waveform transmitted. Therefore, for any
sLFM (t) = exp jπ kt2 − Bt (11) non-eclipsed range bin the pulse compressed response of the

131

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:58:04 UTC from IEEE Xplore. Restrictions apply.
DC bias is simply the convolution of the matched filter with 
the constant DC bias, which reduces to


Tr BN OS −1

0DJQLWXGH5HVSRQVH G%
dDC,m = (γ + jκ) s†m (i), m = 1, . . . , M. (13)
i=0


It is important to note that while the magnitude of each sample



of a constant-modulus waveform is equal (e.g. for a unit-
2
normed matched fitter each sample has magnitude s† (i) = 
1
Tr BNOS ), the complex sample is different. Therefore, the
summation of all complex samples of each unique waveform  /)0
will likewise be unique. Subsequently, the range-slow-time /)0:LQGRZHG
response of the DC bias for the NRB non-eclipsed range bins 3$5&
 3$5&:LQGRZHG
is the (NRB × M ) matrix 0HDQ3$5&

DDC = 1dDC T . (14)   


1RUPDOL]HG'RSSOHU6KLIW
where (•)T is the transpose operation and dDC =
T
[dDC,1 , dDC,1 , . . . , dDC,M ] . If the matched filter does not Fig. 2. Doppler Response of DC Bias
change from pulse-to-pulse, the DC bias results in all entries
of D having a constant value dDC,m = dDC , ∀m. However,
if the waveform does change from pulse to pulse, then each ξt is

column is unique. x(nm ) = 
Pi ξt rm (nm − nt )exp (jπfd n) , (15)
Assuming the DC bias does not change from pulse-to-pulse,
the net effect of the filtered DC bias is to add a constant where the power of the image Pi was given in (7), the
v
increase to the clutter response. Consequently the Doppler normalized Doppler frequency is defined as fd = 4 λr,i
c
and
response of the DC bias in this case is a sinc function centered the smeared cross-correlation between the matched filter for
at zero Doppler. In practice, even non-pulse agile systems the the mth pulse and the image of the waveform transmitted on
DC bias must be estimated and subtracted from the received the same pulse is
signal. Provided the residue from DC bias removal is no 
Tr BN OS −1

greater than the mean clutter power, the sidelobes of the rm (k) = hm [k − i]s†m [i]. (16)
DC bias Doppler response will be below the average clutter i=−Tr BNOS +1
sidelobes. Further, as the energy of the DC bias response is
The range-domain response of the target is then the
contained within the zero Doppler bin, a Doppler window will
2Tr BNOS − 1 × M matrix
further suppress the sidelobes of the DC bias response.
In contrast, for a pulse-agile radar the vector d will not be R = [r1 , exp (jπfd ) r2 , . . . , exp (jπfd (M − 1)TP RI ) rM ]
constant values. Therefore, the DFT of d will contain arbitrary (17)
spectral content across the Doppler spectrum. Further, as the where r m is the length (2Tr BNOS − 1) vector denoting the
spectral content is largely outside of the zero Doppler bin full cross-correlation response of (16). The DFT of (17) is the
windowing will not reduce the spectral response in the high range-Doppler response of the image of the target response.
Doppler regions. If the transmitted pulse is constant, then r1 = r2 = · · · = rM
An example of the spectral response of the DC bias for and the image response for a target manifests as a scaled copy
LFM and PARC waveforms is shown in Figure 2 where the of r with a progressive phase shift equal to samples of the
waveform parameters were given in Section III-B. The sinc true target’s Doppler frequency mirrored across zero Doppler.
sidelobes of the LFM are clearly seen in the blue trace, and However, if the cross-correlation response rm varies from
the red trace demonstrates the reduction in sidelobe response pulse-to-pulse, then the response to the target will likewise
due to the application of the Taylor window. However, the vary across both range and Doppler. Figure 3 demonstrates
PARC waveform is spread across the Doppler space. Note the nature of the Doppler response of the image as compared
that in the higher Doppler regions > 10 dB of increased DC to the true target response, where  = 1.0116 and φ = 0.25◦ ,
bias suppression is required relative to the non-pulse agile for a real-to-image power ratio of Pr = −44 dB. While the
case to maintain a similar level of DC bias distortion. This pulse-agile PARC image response is spread in Doppler, it is
requirement increases to a > 35 dB reduction in residue if the ∼ 20 dB below the LFM image response, which itself is 70
windowed response is used as the baseline. dB below the target response. Therefore, the target image due
to IQ imbalance is likely to be well below the noise floor
V. TARGET R ESPONSE TO IQ I MBALANCE of typical range-Doppler processing. The full range-Doppler
From (6) the image response due to IQ imbalance of a single spread nature of the PARC image response is shown in Figure
target located at sample nt with complex reflection coefficient 4. It is important to note that the Doppler shift due to the

132

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:58:04 UTC from IEEE Xplore. Restrictions apply.
 random variable does not change its distribution. However, the
7UXH7DUJHW variance of the random variable qm (k) is
/)0,PDJH
 3$5&,PDJH 2Tr BN
OS −2
σz2 2
V ar [qm (k)] = Pi |rm (i)|
0DJQLWXGH5HVSRQVH G%

(19)
Nr i=0

where σz2 is the variance of the original clutter random process
 (scaled by the waveform oversampling factor Nr ). Therefore,
it can be seen that for pulse agile waveforms there will be

a Doppler-varying response dictated by the smeared, pulse-
varying cross-correlation function rm . This response is similar
to the RSM commonly encountered by pulse-agile ground-

looking radars, but the underlying response characteristics are
unique and will not be mitigated by traditional techniques.
 As an example, a clutter process with 30 dB sample variance
   (to simulate a 30 dB clutter-to-noise ratio, but for clarity
1RUPDOL]HG'RSSOHU is shown without the impact of noise) was generated and
convolved with LFM and PARC waveforms using the IQ
Fig. 3. Doppler Response of Target and Target Image
imbalance parameters given in Section V. A comparison of the
clutter image responses is shown in Figure 5. Note that as with
target velocity has no meaningful impact on the range-Doppler RSM the clutter image response sets a dynamic range floor
response of the PARC image. that is coherent with the transmitted waveform. Therefore, the
floor will not reduce as more pulses are averaged together. The
 


 



0DJQLWXGH5HVSRQVH G%


/DJ VDPSOHV


 

 

 

 

 

  /)0LPDJH:LQGRZHG
 3$5&LPDJH:LQGRZHG
   
1RUPDOL]HG'RSSOHU   
1RUPDOL]HG'RSSOHU6KLIW
Fig. 4. range-Doppler Response of Target and Target Image for PARC
Waveforms Fig. 5. Doppler response of clutter for LFM and PARC waveforms

full range-Doppler response for the CPI of PARC waveforms is


VI. C LUTTER R ESPONSE TO IQ I MBALANCE shown in Figure 6. Note that the peaks of the Doppler response
In contrast to the target response analyzed in Section V the of the pulse-to-pulse varying cross-correlation rm causes faint
clutter response is not modeled as a simple point target. Rather, constant-Doppler ”streaks” across the entire range response.
in the non-eclipsed region of the range profile the clutter Limited work has been done in the literature to characterize the
response is the convolution of the smeared autocorrelation expected value of RSM [22]. This characterization is highly
rm and the image of the clutter process z(t). The collective dependent on waveform design, and is an important topic of
response of the clutter image at range bin k is then future work.
Due to page constraints the derivation of the noise image
OS −2
2Tr BN response is not shown here. However, following a similar

qm (k) = Pi rm (i)zm (k − i), m = 1, . . . , M.
i=0
derivation of (18) and (19) the distribution of each noise
(18) sample may be shown Trto be zero mean, complex Gaussian
BNOS −1 2
The distribution of the clutter image is still complex Gaussian, with variance Pi σu2 i=0 |sm (i)| . As Pi << Pr the
as the complex scaling of the complex conjugate of a circular noise image will be well below the noise floor and incoherent.

133

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:58:04 UTC from IEEE Xplore. Restrictions apply.
 [3] K. Gerlach, “The effect of i,q mismatch errors on adaptive cancellation,”
IEEE Transactions on Aerospace and Electronic Systems, vol. 28, no. 3,
 pp. 729–740, Jul. 1992.

[4] J. E. Hodkin, O. F. Somerlock, M. D. Sharp, C. L. Farthing, S. H. Talisa,
and K. W. O’Haver, “IQ imbalance decorrelation in digital array radars,”
  in Proc. IEEE Aerospace Conf, Mar. 2016, pp. 1–8.
[5] J. Jakabosky, S. D. Blunt, and B. Himed, “Spectral-shape optimized FM
5DQJH%LQ

  noise radar for pulse agility,” in Proc. IEEE Radar Conf. (RadarConf),
May 2016, pp. 1–6.
 [6] C. A. Mohr, P. M. McCormick, and S. D. Blunt, “Optimized comple-
 mentary waveform subsets within an FM noise radar cpi,” in Proc. IEEE
Radar Conf. (RadarConf18), Apr. 2018, pp. 0687–0692.

[7] C. A. Mohr and S. D. Blunt, “FM noise waveforms optimized according
 to a temporal template error (tte) metric,” in Proc. IEEE Radar Conf.
 (RadarConf), Apr. 2019, pp. 1–6.
 [8] B. Ravenscroft, J. W. Owen, J. Jakabosky, S. D. Blunt, A. F. Martone,
 and K. D. Sherbondy, “Experimental demonstration and analysis of
cognitive spectrum sensing and notching for radar,” Sonar Navigation
 IET Radar, vol. 12, no. 12, pp. 1466–1475, 2018.

   [9] J. W. Owen, B. Ravenscroft, B. H. Kirk, S. D. Blunt, C. T. Allen,
1RUPDOL]HG'RSSOHU6KLIW A. F. Martone, K. D. Sherbondy, and R. M. Narayanan, “Experimental
demonstration of cognitive spectrum sensing notching for radar,” in Proc.
IEEE Radar Conf. (RadarConf18), Apr. 2018, pp. 0957–0962.
Fig. 6. Range-Doppler response of clutter image for PARC waveforms [10] B. H. Kirk, K. A. Gallagher, J. W. Owen, R. M. Narayanan, A. F.
Martone, and K. D. Sherbondy, “Cognitive software defined radar:
A reactive approach to rfi avoidance,” in Proc. IEEE Radar Conf.
Therefore there is no significant difference in noise response (RadarConf18), Apr. 2018, pp. 0630–0635.
[11] B. H. Kirk, R. M. Narayanan, K. A. Gallagher, A. F. Martone, and K. D.
to IQ imbalance for pulse agile systems as compared to non- Sherbondy, “Avoidance of time-varying radio frequency interference
pulse agile systems. with software-defined cognitive radar,” IEEE Transactions on Aerospace
Note that all results have been for non-eclipsed regions of and Electronic Systems, vol. 55, no. 3, pp. 1090–1107, Jun. 2019.
[12] S. D. Blunt, M. R. Cook, and J. Stiles, “Embedding information into
the range response. However, this region is quite large for the radar emissions via waveform implementation,” in Proc. Int. Waveform
smeared cross-correlation function. For space, the derivation Diversity and Design Conf, Aug. 2010, pp. 000 195–000 199.
of the eclipsed regions is omitted, but is straightforward to [13] C. Sahin, J. Jakabosky, P. M. McCormick, J. G. Metcalf, and S. D. Blunt,
“A novel approach for embedding communication symbols into physical
derive from the expressions given here. radar waveforms,” in Proc. IEEE Radar Conf. (RadarConf), May 2017,
pp. 1498–1503.
VII. C ONCLUSION [14] B. Ravenscroft, P. M. McCormick, S. D. Blunt, J. Jakabosky, and J. G.
Metcalf, “Tandem-hopped OFDM communications in spectral gaps of
The impact of IQ imbalance and DC bias on pulse agile FM noise radar,” in Proc. IEEE Radar Conf. (RadarConf), May 2017,
waveforms was compared to non-pulse agile waveforms. It was pp. 1262–1267.
shown that DC bias caused an increase in Doppler sidelobes [15] B. Ravenscroft, P. M. McCormick, S. D. Blunt, E. Perrins, and J. G.
Metcalf, “A power-efficient formulation of tandem-hopped radar amp;
for pulse-agile systems that was not mitigated by windowing. communications,” in Proc. IEEE Radar Conf. (RadarConf18), Apr. 2018,
Further, it was shown that IQ imbalance can significantly pp. 1061–1066.
reduce the dynamic range of a ground-looking system due [16] M. R. Cook, S. D. Blunt, and J. Jakabosky, “Optimization of wave-
form diversity and performance for pulse-agile radar,” in Proc. IEEE
to an increased clutter response caused by changing the RadarCon (RADAR), May 2011, pp. 812–817.
waveform from pulse-to-pulse. [17] C. Sahin, J. G. Metcalf, and S. D. Blunt, “Filter design to address range
sidelobe modulation in transmit-encoded radar-embedded communica-
ACKNOWLEDGMENT tions,” in Proc. IEEE Radar Conf. (RadarConf), May 2017, pp. 1509–
1514.
The authors would like to thank Mr. Charles Mott and Dr. [18] B. Ravenscroft, J. W. Owen, S. D. Blunt, A. F. Martone, and K. D.
Sherbondy, “Optimal mismatched filtering to address clutter spread from
Braham Himed of the AFRL Sensors Directorate, as well intra-cpi variation of spectral notches,” in Proc. IEEE Radar Conf.
as Dr. Carl Rossler of Matrix Research Inc. for their many (RadarConf), Apr. 2019, pp. 1–6.
insightful discussions on this topic. [19] K. D. S. Benjamin H. Kirk, Anthony F. Martone and R. M. Narayanan,
“Mitigation of target distortion in pulse-agile sensors via richardson-lucy
This work was supported by a subcontract with Matrix Re- deconvolution,” IET Radar, Sonar & Navigation Letters, 2019.
search, Inc. for research sponsored by the Air Force Research [20] C. Sahin, J. G. Metcalf, A. Kordik, T. Kendo, and T. Corigliano,
Laboratory under Prime Contract #FA86 50-14-D-1722. This “Experimental validation of phase-attached radar/communication (parc)
waveforms: Radar performance,” in Proc. Int. Conf. Radar (RADAR),
work was performed while Mr. Mohr and Dr. McCormick were Aug. 2018, pp. 1–6.
at the University of Kansas, Radar Systems Laboratory, and [21] C. Sahin, P. M. McCormick, J. G. Metcalf, and S. D. Blunt, “Power-
Mr. Kordik was at Defense Engineering Corporation. efficient multi-beam phase-attached radar/communications,” in Proc.
IEEE Radar Conf. (RadarConf), Apr. 2019, pp. 1–6.
[22] C. Sahin, J. G. Metcalf, and S. D. Blunt, “Characterization of range
R EFERENCES sidelobe modulation arising from radar-embedded communications,” in
[1] M. A. Richards, Fundamentals of radar signal processing. Tata Proc. Int. Conf. Radar Systems (Radar 2017), Oct. 2017, pp. 1–6.
McGraw-Hill Education, 2005.
[2] A. I. Sinsky and P. C. p. Wang, “Error analysis of a quadrature coherent
detector processor,” IEEE Transactions on Aerospace and Electronic
Systems, vol. AES-10, no. 6, pp. 880–883, Nov. 1974.

134

Authorized licensed use limited to: AMITY University. Downloaded on November 30,2022 at 04:58:04 UTC from IEEE Xplore. Restrictions apply.

You might also like