Singaporean Journal Scientific Research (SJSR)

ISSN: 2231 - 0061Vol.3, No.1pp.133 - 137

©Singaporean Publishing Inc. 2010
available at: : http://www.iaaet.org/sjsr
Area Efficient High Speed FPGAImplementation of DUC
forUltrasonic NDE SignalProcessing Techniques
P.K.Magendran1, Dr. S.Sumathi2
1
Rayalaseema University,Research Scholar, ECE department, Kurnool, A.P, India
Email: [email protected]
2
Adhiyamman College of Engineering, Professor, ECE department, Hosur, India
Email: [email protected]

Abstract-In many Ultrasound Non Destructive Evaluation identify the location of defects. The operation process of
(NDE) [1] applications, the transducers are located far ultrasonic NDE includes transmission of pulse, receiving of
away from the evaluation site. Traditionally, low level echoed signal, signal processing, and post-processing, among
analog signals are sent over long lines for which signal processing is undoubtedly the most important
instrumentation, which is prone to noise problems. In this stage, because it’s in this stage that the received signals are
project a different approach is followed by integrating denoised and analysed. Most ultrasonic inspection is done at
digital signal processing functions in a single FPGA, to frequencies between 0.1M and 25MHz.
improve the information content of the output data The Digital up Converter is a digital circuit which
stream.Digital Up Converter (DUC) is a digital circuit implements the conversion of a complex digital baseband
that implements the conversion of a complex digital signal to a real passband signal. The input complex baseband
baseband signal into a real passband signal and vice- signal is sampled at a relatively low sampling rate, typically
versa. A DUC consists of a series of cascaded at the digital modulation symbol rate. The baseband signal is
interpolation finite impulse response (FIR) filters, a filtered and converted to a higher sampling rate before being
mixer and a direct digital synthesizer (DDS). Typical modulated onto a direct digitally synthesized carrier
design requirements shall include:(i) A relatively low frequency. The DUC typically performs pulse shaping and
sampling rate for the input complex base band signal, modulation of an intermediate carrier frequency.The input to
typically at digital modulation symbol rate. (ii) The the Digital Up Converter is an Ultrasonic signal ranging in
baseband signal is filtered and converted to a higher frequency from 50k-1000 kHz.
sampling rate by modulating it to a direct digitally The DUC consists of the following blocks
synthesized (DDS) carrier frequency. In this paper, these • Pulse shaping FIR filter
filters are designed using MATLAB [2] and Verilog HDL • Cascaded Integrator Comb (CIC) filter [4]
[3]. Functional verification is carried out using Xilinx ISE • Compensation FIR filter
and the hardware implementation is done on Spartan 3E • Direct Digital Synthesizer (DDS)
FPGA. The design issues explored in this paper include • Multiplier [5]
the optimization of DUC with respect to different • Channel FIR filter [4]
parameters such as (i) The filters should be generic and
programmable for a wide range of frequencies with good 2.ARCHITECTURE AND
scalability (ii) The number of FIR interpolation Filter DESIGNIMPLEMENTATION
sections to be placed in cascade (iii) The required order
of the filter (iv) The required sections of Cascaded A. Description Of DUC
Integrator Comb (CIC) Interpolation Filter[4] (v) Area A block diagram of the Digital up Converter is shown in
efficient architecture and high speed implementation of Figure 1.
the multiplier algorithm for the mixer [5], etc. The
implemented DUC typically perform pulse shaping and
modulation using an intermediate carrier frequency and
make it extensively usable in NDE Signal Processing
Techniques.
Keywords: Digital Up Converter, FPGA, CIC Filter,
Interpolation filters, NDE

1. INTRODUCTION
A. Ultrasonic NDE
Non-Destructive Evaluation (NDE) [1] has found broad
applications in the evaluation of product qualities, detection
of flaws, etc., and ultrasonic NDE technique is the most
widely used for its distinct features of broad spectrum of test
objects. Ultrasonic NDE makes use of reflection and Figure 1. Digital Up Converter Block Diagram
transmission properties of ultrasonic wave in materials to
Area Efficient High Speed FPGA Implementation of DUC for Ultrasonic NDE Signal Processing Techniques
134
The input to the Digital Up Converter is an Ultrasonic D. Characteristics of CIC Filter
signal ranging in frequency from 50 KHz-1000 KHz. A 12 Typical characteristics are
bit Analog-to-Digital (ADC) converter has the capacity of • Linear phase response;
sampling signals up to the rate of 500 Ksps. The input • Utilize only delay and sum block (no multipliers);
Ultrasonic signal consists of frequency components that lie • The integrator and comb structure need to be independent
between 50 K- 1000KHz.An optimum value of 256KHz of rate changes (there is no need toreproject the filter on
sampling frequency is chosen for the ADC. This 12 bit interpolation rate change).
sampled composite digitized signal is fed to the multiplier in
the DUC block and is configured in Spartan 3EFPGA. E. CIC Interpolator
The input signal is given to the multiplier wherein the A CIC Interpolation filter has two major sections: a
signal is multiplied with 200 KHz fixed IF carrier generated comb section, (acascade of N combs) and an integrator
by the DDS. The output generated from this DDS is 8-bit section, (a cascade of N integrators). There is aninterpolator
with a Spurious Free Dynamic Range (SFDR) of 60 dB. The or rate expansion switch (change by a factor R) between the
12-bit output from the multiplier is then filtered using a two filter sections. The ratechange switch is also known as a
High-Pass Finite Impulse Response (FIR) filter to eliminate zero-stuffer as it pads zeros. The interpolator up samples the
the lower sideband produced at the multiplier stage. A outputof the last comb stage increasing the sample rate from
Cascaded Integrator-Comb filter up-samples the output of the fs/R to fs. One of the distinguishing factors ofCIC filters is,
FIR filter by a factor of 20 to attain a sampling rate of 5.12 the sampling rate of Comb filters is different from sampling
MHz from 256 KHz. The output from the CIC filter is passed rate of integrator, and the comb runs at a lower sampling
through a Channel Filter in order to take out the desired frequency, which makes it easily programmable. Figure 2
output signal (Upper Sideband). The resultant 5.12 MHz gives adetailed structure of a CIC interpolator for 8 stages.
upsampled data is passed on to the DAC for converting the
digital signal to analog form. The Digital-to-Analog (DAC)
converter processes the 12-bit output from the channel filter.
The DAC can handle maximum sampling rate of 125 Msps.
Finally the modulated output is passed to a current to voltage
converter (amplifier) for final transmission.

B. Fir Filters
Spectral shaping of the complex input signal is
performed by the Pulse shaping FIR filter (PFIR) filter. The Figure 2.CIC Filter Structure Block Diagram
compensation finite impulse response (CFIR) filter and PFIR
are polyphase multirate filter structures that interpolate by a F. Comb
factor of 2, 4 or 8. The filter length for each filter is A comb filter running at a low sampling rate fs/R, for a
configurable from 4 to 1024 taps. The coefficient precision rate change of R is an odd symmetric filter described by
can also be customized and ranges from 4 to 32 bits. In the Y[n] = x[n]-x[n-RM] ……………………………... (1)
typical interpolation filtering applications a reasonably flat In the equation, M is a design parameter and is known as
pass band and narrow transition region filter performance is differential delay. M is usually limited to 1 or 2. The
required. These desirable properties are not provided by the corresponding transfer function at fs is
CIC filters alone, with their drooping pass band gains. Hc (z) = 1-z-RM…………………………………… (2)
Hence, the CFIR Filter is used to compensate for the droop The comb sections are combined with a rate changer.
in the passband of the CIC filter. Using a technique for multirate analysis of LTI systems the
comb sections can be pushed through the rate changer and
C. CIC Filter then become
Cascaded integrator-comb, also called Hogenauer filters Y[n] = x[n]-x[n-M] ……………………………….. (3)
[8], are multi-rate filters that are used for realizing large From the above equations the following are achieved:
sample rate conversions in digital systems. The main 1. Half of the filter has been slowed down and therefore
advantage of this filter is that, it does not use multipliers, and efficiency is increased.
consists of only adders, subtractors and registers [4]. They 2. The number of delay elements required in the comb
are typically employed in applications that have a large section had been reduced.
excess sample rate (the system sample rate is much larger 3. The integrator and comb stages are independent of rate
than the bandwidth occupied by the signal). changer.
CIC filters find application in The basic structure of a comb is as shown in figure 3.
• Digital upconverters and digital down converters.
• Channelization functions in a digital radio or MODEM
• For ultra-tight integration of GPS/INS/PL sensors
• Any filter structure that is required to efficiently effect a
large sample rate change.

Published: Singaporean Publishing

Area Efficient High Speed FPGA Implementation of DUC for Ultrasonic NDE Signal Processing Techniques
135
FIR filters, each having a rectangular impulse response with
unit coefficients. The obtained frequency response of a CIC
interpolator for an interpolation factor of 3 is as shown in
figure 5.

x[n] y[n]

Figure 3.Basic Comb

G. Integrator
An integrator is a single pole IIR filter with a unity
Figure 5.CIC Frequency Response
feedback coefficient given by
Y[n] = y[n-1]+x[n] ………………………………(4)
I. Direct Digital Synthesizer
A Direct Digital Synthesizer also known as Numerically
The transfer function for an integrator on the z-plane is
Controlled oscillator (NCO) synthesizes a discrete-time,
HI (z) = 1/ (1-z-1) ……………………………….....(5)
discrete-valued representation of a sinusoidal waveform. It is
The basic structure of an integrator is as shown in figure
an established method for generating periodic sinusoid
4.
signals, whenever high frequency resolution, fast changes in
frequency, phase and high spectral purity of the output signal
is required. A major advantage of the DDS is that its output
frequency, phase and amplitude can be precisely and rapidly
manipulated under digital processor control. Other DDS
y[n] attributes include the ability to tune with extremely fine
x[n]
frequency and phase resolution, and to rapidly hop between
frequencies.
Figure 4.Basic Integrator J. Theory of Operation
A direct digital synthesizer operates by storing the points
Table 1 shows the difference between Comb and of a waveform in digital format, and then recalling them to
Integrator stages in CIC filter derived in the above equations. generate the waveform. The basic block diagram of a DDS is
TABLE 1. Structure and Characteristics of Comb and as shown in Figure 6.
Integrator
Items Integrator Comb

Structure

Sampling
Rate
y[n] y[n-1] + x[n] x[n] – x[n-RM]

H I (z)

Figure 6.DDS Block Diagram with DAC and Filter

The implementation of the DDS can be divided into two

distinct parts namely the phase accumulator (phase
H. Frequency Characteristics generator) and the phase to amplitude converter.The DDS
The transfer function for a CIC filter at fs is can generate frequencies ranging from 50-1000 KHz with a
H(z) = HIN(z) HCN(z) ……………………………….(6) resolution of 0.5 KHz and spurious free dynamic range
H(z) = (1 – z-RM)N / (1 – z-1)N……………………….(7) (SFDR) of 60 dB. The 8-bit output data represents the
This equation shows that even though a CIC has generated sine wave.
integrators in it, which by themselves have an infinite
impulse response, a CIC filter is equal to cascade of an N
Published: Singaporean Publishing
Area Efficient High Speed FPGA Implementation of DUC for Ultrasonic NDE Signal Processing Techniques
136
K. Multiply Accumulator
MAC [5] is a parallel multiply-accumulator module. FIR
filters typically are implemented in Xilinx FPGAs using
either the MAC implementation or the DA (Distributed
Arithmetic) technique. In DUC FIR filter
implementationMAC based method is used. Figure 7 shows a
MAC FIR filter simplified block diagram.

Figure 9. Simulation Results of DDS in MODELSIM

Figure 7.Simplified Block diagram of MAC

The MAC based architecture uses a multiplier to
perform the tap product calculations, followed by an
accumulator to perform the filter addition operations.

3.RESULTS AND DISCUSSIONS

DUC was developed in Verilog code [3] using Xilinx
ISE tool. The code was simulated in ModelSim tool.

CIC
Response CFIR
. Response

Figure 10. Simulated output of DDS, CIC filter, CFR FIR

Cascaded CIC-CFIR
Response
filter

Figure 8.Working of the CIC-CFIR Filter and its

Combined Response in MATLAB

Figure 11.Simulation of Digital Up Converter

The device utilization summary is listed in Table 2.

Published: Singaporean Publishing

Area Efficient High Speed FPGA Implementation of DUC for Ultrasonic NDE Signal Processing Techniques
137
Table 2.Device utilization summary of Spartan 3E FPGA
Device Utilization Summary: REFERENCES
-------------------------------------
Selected Device : 3s500epq208-5 [1] Charles J. Hellier, “Handbook ofNon-destructive
Evaluation”, McGRAW-HILL
Number of Slices: 1104 out of 4656 [2]
23% http://www.mathworks.com/products/filterhdl/demo
Number of Slice Flip Flops: 850 out of 9312 s.html?file=/products/demos/shipping/hdlfilter/hdlduc.ht
9% m
Number of 4 input LUTs: 3428 out of 9312 [3] Samir Palnitkar, “Verilog HDL, A Guide to Digital
36% Design and Synthesis", Second Edition, Pearson
Number used as logic: 1812 Education, ISBN: 81-297-0092-1
Number used as RAMs: 1616 [4] Uwe Meyer-Baese, “Digital Signal Processing with
Number of IOs: 111 Field Programmable Gate Arrays”, Third Edition,
Number of bonded IOBs: 83 out of 158 Springer
52% [5] Xilinx LogiCORE, "Multiply Accumulator v4.0”,
IOB Flip Flops: 13 DS336 April 28, 2005 www.xilinx.com
Number of MULT18X18SIOs: 6 out of 20 30% [6] Lattice Semiconductor Corporation, “The FPGA as a
Number of GCLKs: 7 out of 24 29% Flexible and Low-Cost Digital Solutionfor Wireless
Number of DCMs: 1 out of 4 25% Base Stations”, A Lattice Semiconductor White Paper,
March 2007
Timing Summary: [7] Justin Davis and Robert Reese, “Finite State Machine
------------------------ Datapath Design, Optimization, and Implementation”,
Speed Grade: -5 Morgan & Claypool Publishers, ISSN 1932-3166.
Minimum period: 31.402ns (Max Freq: [8] E. B. Hogenauer, “An economical class of digital filters
31.845MHz) for decimation and interpolation,” IEEE Trans. on
Minimum input arrival time before clock: 7.497ns Acoustics, Speech. and Signal Processing, Vol. 29, pp.
Maximum output required time after clock: 155-162, April 1981.
15.523ns [9] Berkeley Design Technology, Inc. “FPGAs for
Maximum combinational path delay: 17.713ns DSP,”Focus Report, July 2002,
http://www.bdti.com/products/reports_focus.html
From the above results, the verilog code for DUC is
optimized to achieve less area and high speed of operation.

4. CONCLUSIONS
This paper deals with the development of a DUC for
NDE Signal processing applications. The base band signal
that is received by the equipment is in the range of 50KHz –
1000 KHz. The carrier signal frequency that is used to
modulate the signal lies between 100 KHz – 1000KHz. DUC
block was developed to modulate the message signal. In this
design the interpolation filters were cascaded along with a
DDS and a mixer to get the required output. Coding of the
sub-blocks of DUC has been implemented using Verilog.
Simulations were carried out for the coded blocks and the
results have been verified using ModelSim.The Verilog code
is synthesized using Xilinx ISE and the coefficients for the
interpolation filters were verified using MATLAB.
Synthesized Verilog code is implemented using Spartan 3E
XC3S500E-5PQ208 that has 158 user IO’s.

Published: Singaporean Publishing