Processing Scheme

TWD-UWB radar operation:

1. A narrow RF pulse is transmitted and the echoes are samples from different receiver terminals
sequentially one after another. This radar does not corroborate received signal from different
antenna elements and thus it is NOT and array antenna with larger aperture.
2. Effective sampling rate of 4 GSPS is achieved by sampling echoes from multiple pulses (20, in this
case) and by delaying the sampling clock in the steps of 250ps. See fig 4.
Pulse width:
A pulse width of 700ps is planned for this radar.
The narrow pulse radar works best with effective sampling once per pulse-width duration.
The pulse shape not rectangular but is similar to a wavelet.
Effective sampling rate
In this view, it is planned to have an effective sampling rate of 250ps (4 GSPS), followed by a
3-point summation. This method ensures that the echo is not missed due to sampling inappropriate
instance.
However, the sampling rate of 4GSPS is impractical. This is achieved by
(a) Using lower sampling lower sampling (complex sampling at 200MSPS)
(b) Sampling at different PRFs by clock pulse delay (20 delay stages to achieve effective
4GSPS)
(Hence 1 channel sampling requires 20 PRFs taking total time of 100 X 20 =2μs)
(c) 8 microseconds will be required to sample 4 Channels (maximum provision)
Accurate phase measurement:
Cross range target positioning accuracy is dependent on the capability of accurate phase
measurement. Higher SNR at the ADC input directly improves the phase estimation accuracy in the
receiver signal.
Example: Phase measurement with an accuracy of 1⁰
The required SNR at the sampling point is (1/Sin (0.5)) =114.59 (41.18dB)
Typical estimated value of single pulse SNR is 10 dB.
Coherent integration of 2000 echoes shall lead to SNR improvement of 33 dB.
Multi-channel processing shall offer additional SNR improvement by a factor of 4/8 (6/9 dB)
Time required for data capture:
With maximum provision of 4 channels the data capture time is 8 μs X 2000= 16 mS.
Tx
Pulse
700pS t1 t2 t3 t4 tk tk+1
Rx1
c1 c2 c3 c4 3 point summation
Rx2

Rx3
c1 c2 c3 c4 3 point summation
Rx4

Rx5

Rx6

Rx7

time Rx8
Fig 3 Ideal requirement of receiver channel sampling (Δt=tk+1-tk =250 ps )
Achieving high sampling rate by clock pulse delay:
The timing diagram of the sampling scheme is shown in Fig. 4.

IPP=100ns, 4 GSPS
400 sample per IPP

IPP1→ Sampling ck IPP2→ Sampling ck IPP3→ Sampling ck IPP4→ Sampling ck

@ at 1st, 21st, 41st @ at 2nd, 22st, 42st @ at 3rd , 23rd, 43rd @ at 4th, 24th , 44th
Delay the sampling Delay the sampling Delay the sampling
clock by 250 ps clock by 250 ps clock by 250 ps

The target movements are much slower that the PRF of 10 MHz. Therefore it
is assumed to be stationary offering identical echoes for 20 IPPs
Fig 4 Realization of 4GSPS by delayed clock pulse and sampling from 20 IPPs for a
single receiver channel. The process is repeated for all (4/8) receiver channels

Pulse shape may be defined in

terms of 3 points

Bearing Estimation algorithm

After the above sampling process the data is segregated in appropriate range bins. It may be
assumed that that data is collected after a delay of 2nS say.
This corresponds to about 30 cm. A blind period of 30 cm is reasonable.
In a single IPP we have (100-2)/ 0.25= 392 samples, one every 250ps.
(a) If the processor allows one may perform correlation and get the data at every 250ps.
(b) After correlation, the data may be subjected to consecutive sample addition➔ 130 points
Three consecutive samples are added (x1 +x2+x3), (x1 +x2+x3), (x1 +x2+x3),… and so on. Thus a sequence
of @ 1/700ps (≈1.4285GSPS) is created.

For each range bin (echo of a particular time slot of 700ps), every receiver shall have ONE complex
sample. This samples of individual ‘range-bin’ shall correspond to (≈10 cm. Let the vector representing
the samples for nth Range bin be
➔ XRBn= (XR1RBn, XR2RBn, XR3RBn, XR4RBn)T (it will be 8 element vector for 8 receiver case)
Target bearing determination (or the angular position from the antenna ‘boresight’) can be
computed using following matrix equation.

Background

Let a particular “range sector (±70⁰) be divided into 14 sections of 10 deg each, 7 on each side.
The center bearings of each sections are ➔ ±(5+10 k)⁰; k = 0,1,2,3,4,5,6,.
In nth range-bin, Let a target be in the cross-range bin of 30⁰-40⁰, center at 35⁰ (say). The signals
received by 4 receivers shall have approximately magnitude (B -35RBn, Say) and phase gradient
depending on the angular position of the target. As an example, the phases of the receiver signals will
be (phase references to Rx1)
XR1= B-35RBn exp (j 0), XR2= B-35RBn= exp (jφ35), XR3= B-35RBn exp (j 2 φ35), XR1= B-35RBn exp (j 3φ35).
If The distance between two receiver is ‘d’, we have, φ35 (In degrees)=(d Sin35⁰/λ)x 360.
The generalized expression for this will be given as follows:
  B−65 (RBn )
T
1 e − j65 e − 2 j65 e −3 j65 
 −3 j55   B (RBn )
1 e − j55 e − 2 j55 e   −55 
1 e − j45 e − 2 j45 e −3 j45 
 B− 45 (RBn )
   
1 e − j35 e − 235 e −335   B−35 (RBn )
1 e −25 e − 225 e −325   B− 25 (RBn )
   
1 e −15 e − 215 e −315   B−15 (RBn )  XR1RBn 
1 −35   B (RBn )   XR2 
e −5 e − 25 e
   −5
 = RBn 

1 e 5 e 25 e 35
  B5 (RBn )   XR3 RBn 
     
1 e 15 e 215 e 315  B (RBn )  XR4 RBn  ( 41)
 15 
1 e 25 e 225 e 325   B25 (RBn ) 
   
1 e 35 e 235 e 335   B35 (RBn ) 
1 e 45 e 245 e 345   B45 (RBn ) 
   
1 e 55 e 255 e 355   B55 (RBn ) 
1
 e 65 e 265 e 365  ( 414)  B65 (RBn )  (141) (1)
Let the matrices be named as follows:
Path matrix P(4X14) , Reflectivity matrix B(14X1) and received X(4x1) is the received signal Matrix .
Eq.(1) can be written in matrix form as PB=X (2)
The problem: Given X it is required to find B. The matrix P is known by the geometry.

Solution to (1) can be obtained by Singular Value Decomposition (SVD) technique. This will identify 3
targets corresponding to the prominent Eigenvalues; as presented by Lipa et.al. [1].

However, it is possible to perform more elaborate mathematics and extract maximum possible
information. The method is as follows:

Pre multiply by PH. Where, PH is a Hermitian (or conjugate transpose) of P.

(PH P)B= PH X (3)

Now, (PH P) is a complex symmetrical matrix. Let it be indicated by ₱

₱ B = PH X. (4)
Now we factorize ₱ using singular value decomposition using method given [2]. Let the
symmetrical SVD or SSVD be ₱= UΣVH. (There may be a direct function in Matlab). Substituting,

UΣVH.B = PH X (5)
Now we pre-multiply, by (VΣUH.) to both sides,

(VΣUH.) UΣVH.B = (VΣUH.) PH X ➔ B = (VΣUH.) PH X (6)

All the terms on the right are known. Hence the value of B can be computed.

Where, Pn is the spatial spectral vector for the n th range bin.,indicating the echo strength in
every 10⁰ sector. The above equations assume following sign convention.
(a) System configuration: 1 Transceiver (Tx and Rx1 co-located) at the origin (or reference)
and 3 receivers placed linearly on -ve Xaxis, 0.7λ apart ( d=0.7 λ)
(b) Bearing angles measured anticlockwise from boresight are positive.
(c) Øϴ is the value of compensatory angle for the target at the bearing ϴ.
 The value is computed using Øϴ=(d Sin ϴ)/λ) X360 = 252 Sinϴ.
After acquiring the data, it is above equations, it is possible to plot the information graphically as given
in Fig.5.

Constant Range
lines

Positioning targets in
TWD-
appropriate range and cross-
UWB
range bins after processing
Radar

Constant Range
lines (combination
of 2 ellips
sections)

Fig 5 Two-Dimensional position plotting of the target location after Range

bearing analysis Dimentional p