See discussions, stats, and author profiles for this publication at: [Link]

net/publication/321024930

Fundamentals of FPGA Architecture

Chapter · January 2017

CITATIONS READS
2 15,908

2 authors, including:

Sahadev Roy
National Institute of Technology Arunachal Pradesh
195 PUBLICATIONS 11,408 CITATIONS

SEE PROFILE

All content following this page was uploaded by Sahadev Roy on 13 November 2017.

The user has requested enhancement of the downloaded file.

Advanced Engineering

2
Fundamentals of FPGA Architecture

Gopal Krishna and Sahadev Roy

 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

Fundamentals of FPGA Architecture

Gopal Krishna and Sahadev Roy

2.1 Introduction
Field programmable gate array (FPGA) first come in picture in early 1970s [1]. It is developed
version of PLD. In PLD'S programmed after manufacturing in field, it has limited programmability.
Field programmable gate is capable of implementing any digital circuit. This provides developer of
creating wide array of logical structure minimum low cost.
Programmability is high in FPGA with minimum design time. Low power consumption, high speed
input output, large parallelisms are the important features of it. In this chapter we will discuss about
basic building blocks of FPGA's and core technologies of FPGA. To understand basic hardware we
follow bottom to top approach. First discuss building blocks (block diagram) then we will discuss
how they are combined and interconnected.

2.2 Requirement of FPGAs

By the early 1980’s large scale integrated circuits (LSI) formed the back bone [2] of most of the
logic circuits in major systems [3]. Microprocessors, bus/IO controllers, system timers etc were
implemented using integrated circuit fabrication technology. Random glue logic [4] or interconnects
were still required to help connect the large integrated circuits in order to:
1. Generate global control signals (for resets etc.)
2. Data signals from one subsystem to another sub system.
Systems typically consisted of few large scale integrated components and large number of SSI
(small scale integrated circuit) and MSI (medium scale integrated circuit) components. Initial
attempt to solve this problem led to development of Custom ICs which were to replace the large
amount of interconnect. This reduced system complexity and manufacturing cost, and improved
performance. However, custom ICs have their own disadvantages. They are relatively very
expensive to develop, and delay introduced for product to market (time to market) because of
increased design time [5]. There are two kinds of costs involved in development of custom ICs.
13

1. Cost of development and design

Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

2. Cost of manufacture
Therefore the custom ICs method was only efficient for high volume products and which were not
to market sensitive. An alternative to custom ICs FPGA were introduce, FPGA capable of
implementing entire system on chip and it provide flexibility to reprogram for user. Development of
FPGA removes the density relative to discrete SSI/MSI components (aproximately 10× of custom
ICS).
One of the major advantage of FPGA over custom ICs is circuit implementation time is very less
because physical layout, masking etc is absent in this process. Circuit implementation is done with
help of the advanced CAD tools.

2.3 FPGA Evaluation

A digital system is mainly consisting of three components memory, microprocessor and logic
element. Memory is used to store information like data of look-up table and content of database.
Microprocessors are used for execution of software instruction written in word. Logical unit
perform almost all operation like data display, signal processing, device to device interface. Data
communication and time synchronisation is also provided by logical unit.
Programmable Read Only memory (PROM) was the first user programmable chip that could
implement logical circuit. In programmable read only memory data lines are as output whereas
address line as input. For its address input PROM requires a full decoder whereas logic functions
require more than few product term. The PROM is used for implementing logical circuit.
FPGA is developed version of PLA. PLA is used for implementation of any Boolean function using
sum of product form. Input buffer, programmable AND- matrix followed by programmable OR-
matrix are basic part of PLA implementation. Basic function of input buffer is to provide original
and its inverted value as PLA input [6]. Horizontal line is for the input which is implemented on
AND matrix and product term line implemented virtually. So the size of AND matrix is twice than
of product term of inputs.
First PLA was introduced by Philips in early 1970s, but some major drawbacks there in PLA poor
speed performance and high manufacturing cost is very high. The two major disadvantages are due
14

to the two level of configurable logic because programmable logic plane is not easy to manufacture
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

and it has some significant propagation delays. To overcome the drawbacks of PLA programmable
array logic were introduce. PAL has single level programmability it contain wired AND followed
by OR gate in same plane [7].
PAL generally contains flip-flops connected to the OR gate output so that sequential circuit can be
realized. These are called as simple programmable logic devices (SPLDs) Fig. 1 shows simplified
structure of PLA and PAL.

(a) (b)
Fig 1. Illustration of simplified structure (a) PLA and (b) PAL.

In latter on the development of SPLDs with development of technology it is possible to make

device with higher density. In CPLDs chip density increased. In simple language we can say that
CPLDs consist of multiple PLDs with some extra programmable interconnects on a single chip. The
larger size of CPLD allows it to implement more complicated logic design on it. 15
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

Fig 2. Block diagram of Hypothetical CPLD

Figure 2, is block diagram of hypothetical CPLD in this CPLD it is shown that it has four logic
blocks it is simple PLDs. The number of PLD logic blocks may be more than four or less than four.
These logic blocks may be more than four or less than four. These logic blocks also contain macro
cell and interconnection similar as PLD.
In CPLD the switch matrix may be connected or not connected whereas in PLD it is fully
connected. Some of the theoretical connection of logic blocks output and input is not supported
within CPLD. Due to this it is difficult to achieve 100% utilization of macro cells .there is sufficient
number of logic gates and flip-flops available in CPLD despite of that some hardware design is not
allow to implement [8]. CPLD can hold larger design than PLD because of that it is widely used.

2.4 Hardware Design

Majority of electronic circuit is made up of the three basic components.
a. Combinational logic element
b. Memory element
c. Wiring to interconnect

2.4.1 Combinational Logic

A combinational logic is made up of small logic gates and interconnection them [5]. For example
half adder is made up of a XOR gate and a AND gate this give sum output at s port which add input
a AND b. For carry output port c is assigned. If a AND b both are 1 then carry will generate at and
16

gate. Two half adder with one or gate can be combine to build a full adder. It has three input ports
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

(carry-in). Cascading of full adders can be done to make adder with wider word width. Another
example of basic combinational circuit is multiplexer. 2 to 1 multiplexer has two inputs (in0, in1)
and one select line (SEL). SEL line determines which input line appears at output. Wider
multiplexer can be constructed to connect them. No clock is involved in combinational circuit.
Every logic gate has its own propagation delay [9]. As the number of gates increases propagation
delay also increases. The propagation delay of a complex combinational circuit comprises the sum
of propagation delays of its gates along the longest path within the circuit, known as the critical path
[10]. The propagation delay is determined by the critical path of logic circuit.

2.4.2 Sequential Logic

In opposite to combinational circuit the sequential logic has state (memory).in sequential logic
clock is involve. It is combination of combinational logic with memory (Fig 3).

Fig 3. Type of the sequential logic circuit.

State element (Flip-Flop, latch, etc) are basic building block of sequential logic.
 1. Event Driven – asynchronous circuits that change state immediately when enabled.
 2. Clock Driven – synchronous circuits that are synchronized to a specific clock signal.
 3. Pulse Driven – which is a combination of the two that responds to triggering pulses?
17
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

2.4.3 Asynchronous sequential logic

The most basic asynchronous sequential logic is SR latch which mean set reset [11]. It is not
possible to neither construct by using two cross coupled NOR gate. When S and R both at logic low
feedback loop ensure that the output will be in previous state. If 𝑆 = 1 and 𝑅 = 0 set output occur
𝑄 = 1 and 𝑄 = 0. Similarly for reset input S=0 and R=1 reset output occur 𝑄 = 0 and 𝑄 = 0. S and
R are not allowed to give both logic high input because 𝑄 = 0 and 𝑄 = 0not possible. SR latch is
level sensitive (5 volt is consider as one state while zero volt as other). So these circuits can
maintain state but it is purely driven by its input no synchronization is there known as asynchronous
sequential circuit (Fig 4). Time delay is only due to the propagation delay of gates.

(a) (b) (c)

Fig 4. Illustration of (a) Asynchronous sequential circuit with feedback, (b) SR (NOR) latch, (c) D-
flip flop.

2.4.4 Synchronous Sequential Logic

In synchronous sequential circuit all memory element is synchronized with fixed clock signal. It is
sequential circuit can be positive edge triggered (0 to 1) or negative edge triggered (1 to 0). The
basic sequential logic is D flip-flop. It is used to store the value of input at occurring of clock edges.
D flip-flop is implemented using to latches with additional logic gates. The basic need of sequential
logic is due to its simplicity and reliability. The whole operation of logic circuit should be getting
completed in a single clock time. The clock time is defined by critical path of combination path in
18

flip-flop.
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

2.5 The Key to Reconfigurable Hardware

A combinational logic is made up of basic logic gate connected through wire. In FPGA these gates
are simulated using look-up table (LUT).

2.5.1 Internal Architecture of LUT

A, n-input LUT requires 2^n bits of SRAM to store lookup table and 2^n: 1 multiplexer to read
individual bit. In general 4-input look-up table is used but nowadays 6 input look-up table is used
(Fig 5).

(a) (b)
Fig 5. Four input LUT (a) circuitry for read (b) circuitry for write.

The value in0 to in3 are used to determine which SRAM bit is given at output. Boolean function is
stored in SRAM cell. SRAM cell is simply a shift register with one-bit width and 2^n bit depth. The
Bit is shifted bit-by-bit into LUT when FPGA is programmed.
LUT can also be used as memory element on FPGA. When FPGA is programmed LUT is used as
19

distributed RAM. The multiple LUTs are combined to make wider or deeper memories.
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

2.6 FPGA Architecture

After discussing LUT now we will discuss how these LUT are combined and distributed over entire
FPGA architecture. Some fixed number of LUT are grouped and embedded on programmable logic
device which is called as elementary logic unit (Xilinx refer it as slices).
Four main basic structural element of elementary logic unit are
1. Number of LUT
2. Proportional number of one bit register.
3. Several multiplexer.
4. Arithmetic/carry logic
An example of elementary logic unit is given in Fig 6. Which contain 4-input LUT and two Flip-
flop register (Fig 6). Modern FPGA has more LUTs per elementary logic unit.

(a) (b)
Fig 6. Functionality within an (a) elementary logic unit and (b) a full adder constructed by
combining a LUT with elements of the carry logic (right).
20
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

Each LUT is paired with a memory element which is flip-flop to store result. Flip-Flops are the
second type memory in FPGA.

2.7 Routing Architecture

Direct wires to neighbouring elementary logic units allow combining multiple units to build more
sophisticated circuits such as adders and multipliers. Modern FPGAs provide enough configurable
resources to host an entire system on chip (SoC) [12]. To build that much complex network proper
routing is required this is known as interconnect. The interconnections of the Logic Island are
arranged as a two-dimensional array on FPGA. In this pattern any logic island can communicate to
any logic island even at other end at matrix (Fig 7).

Fig 7. Internal structure of Logic Island (left) and two dimensional arrangement of Logic Island
with input output blocks.

A small number of elementary logic units are grouped together into a coarser grained unit that we
refer to as Logic Island. In this example logic island consist of two elementary logic units. Each
logic element in Logic Island is separated set of wire which is adjacently connected to different
21

logic island [13]. For general communication interconnection is done by switch matrix.
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

2.7.1 Interconnect
Interconnect is used for communicate between different logic island (LIs) these interconnect are
configurable. It consists of horizontal and vertical channels (bundle of wire) and vertical channels
forming grid contain a logic island in every grid cell (Fig 8).

Fig 8. Routing architecture with switch matrix, programmable links at intersection points, and
programmable switches.

At interconnect of routing channel there is a programmable links which determine how wire is
connected, how input output is routed in particular logic island. All wires are connected to
22
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

additional three wires at interconnection point but which connection is active is determine by
programmable switch.

2.8 High Speed I/O

For communicating with large number of logic islands on FPGA large number of I/O blocks are
there at peripheral of FPGA. The IoBs are programmed to serve for different need and allow FPGA
to communicate with multitude of other devices. Two min I/O standard are single ended or for high
performance differential (used in SATA, 10G Ethernet etc.). IoBs can be programmed to implement
the physical layer of many common communication schemes. High speed input output can be
achieved with the help of extremely fast transreceivers. Few fast transreceivers are available in
markets are VIRTEX-7 HT FPGA from Xilinx providing 28 Gb/s serial bandwidth each. Aggregate
bandwidth is more than terabit can be achieved because it has sixteen 28Gb/s and seventy two
13Gb/s transreceiver.

2.9 FPGA Programming

To illustrate a typical FPGA design flow, we will examine the tools of the Xilinx tool chain, as well
as their intermediate circuit representations. The most important steps and tools of the design flow
to produce an FPGA-circuit are depicted in Fig 9.

Fig 9. FPGA design flow: Xilinx tool chain and intermediate circuit specification formats.
23
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

Programming on FPGA is same as connecting wires in circuit this is done by using a hardware
description language (HDL) such as VHDL or Verilog. Synthesizer converts HDL into gate level
netlist, native generic circuit (NGC) format mapped to technology library provided by Xilinx. At
this level third party synthesizer can also be used.
2.9.1 Translate
Tool ngbuild merge and translate all input netlists and constrain into a single netlist and save it as
native generic database (NGD) file. User constrain file (UCF) is specify by the FPGA designer at
time of manufacturing constrained are used for assigning special physical element of FPGA (e.g.:
I/O pin, clock etc) in the design as well as timing require for design. NGC netlist is based on
UNISIM library whereas NGD netlist based on SIMPRIM library. NGC allow behavioural
simulation and NGD allow timing simulation [14].
2.9.2 Map
The map tool maps the SIMPRIM primitives in an NGD netlist to specific device resources such as
logic islands, I/O blocks, etc. The map tool then generates a native circuit description (NCD) file
that describes the circuit, now mapped to physical FPGA components.
2.9.3 Place and Route
The par tool performs placement and routing. The physical element specified in NCD file is specify
at particular location on FPGA and interconnected. Place and route is most time consuming process
in design flow it is based on simulated annealing algorithms. The par tool takes the mapped NCD
file and generates routed NCD files which also contain the routing information.
2.9.4 Bitstream Generation
The implemented design has to dump on FPGA readable format. This is done by the bitgen tool, it
encode design in binary known as Bitstream. Then Bitstream is loaded on FPGA using JTAG cable.
Inside FPGA a finite state machine control by Bitstream which extract configuration data from
Bitstream.

2.10 Auxiliary On-Chip Components

The logic resources of FPGAs discussed so far are in principle sufficient to implement a wide range
24

of circuits. However, to address high-performance and usability needs of some applications, FPGA
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

vendors additionally intersperse FPGAs with special silicon components, such as dedicated RAM
blocks (BRAM), multipliers and adders (DSP units), and in some cases even full-edged CPU cores.
Hence, [HGV+08] observed that the model for FPGAs has evolved from a \bag of gates to a \bag of
computer parts.

2.10.1 Block RAM (BRAM)

The BRAM is a configurable module, which is attached with variety of BRAM interface
controllers. Figure 10 shows the BRAM blocks.
Main features
1. Fully automated generation and configuration of HDL through EDK Platgen/Simgen tools.
2. Number of BRAM primitives utilized is a function of the configuration parameters for:
memory address range, number of byte-write enables, the data width, and the targeted
architecture.
3. Both Port A and Port B of the memory block can be connected to independent BRAM
Interface Controllers: LMB (Local Memory Bus), OPB (On-chip Peripheral Bus), PLB
(Processor Local Bus), and OCM (On-Chip Memory).
4. Supports byte, half-word, word, and double word transfers provided the correct number of
byte-write enables have been configured.

2.10.2 Digital Signal Processing (DSP) Units

FPGA is very widely used for digital signal processing (filter design, Fourier transform, convolution
etc.), these all are mathematical based expression. Large number of adder and multiplier is
incorporated in FPGA (hundred too few thousand). Adder and multiplier can be realising using
LUTs and input-output in FPGA [15]. These all are power efficient high performance and minimum
space circuit. Similar to other components of FPGA, the digital signal processing unit also needed
and combine with the adjacent DSP section. For example we consider Xilinx DSP$*E slice which
has three input ports of different width (25bits,18 bits and 48 bits) and it provide 25*18 bit
multiplier with a pipelined second stage that can perform as subtractions or adder (48 bit) also with
optimal accumulation feedback. The DSP unit of FPGA is used for various application like
25

multiply, multiply and accumulate, multiply and add/subtract, three input addition , barrel shifting,
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

wide bus multiplexing etc are various operation which can perform on FPGA . It can also be used in
different modes. These all function can be implemented in one or two clock cycle one over a one
wide input. For implementing hash function fast multiplier is very useful.

Fig 10. FPGA layout with interspersed (a) BRAM blocks and DSP units, and (b) the simplified
interface of a dual-ported BRAM block.

2.11 Commercial FPGA Devices

Some commercial FPGA devices are available in market; Altra Stratix is one of them. It consist
major five components all these are connected through programmable routing network.

These five components are:

1. Logic array blocks (LABs):- to implement logic array function.

2. Memory block:- for storing data
3. DSP unit:- for speed up mathematical operation like multiply and accumulate
4. Phased locked loop modules:- to alter the phase and frequency of input clock
5. I/O pads: - to connect to outside world.

Each logic array block contains ten logic elements (LE) and these can operate in normal mode
or dynamic arithmetic mode. In normal mode each logic element is configure of four input
26

lookup table (LUT) and a register. The output of Logic Element is output of lookup table or the
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

output of register whose data is also part of Look-up table. This type of mode is very useful in
implementing the logic arbitrary logic functions. For implementing arithmetic operation logic
element can be configure in dynamic arithmetic mode one for sum output and two is carry
output signal. The carry output signal is connected to adjacent logic element through dedicated
routing. The addition result is generated by one of the two input LUTs and it also depends on
the carry input value, where each two LUT compute the sum for a possible carry-in of either 0
or 1 [16]. The both carry-out is generated similarly which depends on the carry input value
either it is 0 or 1.

The Altra Stratix 2 is second generation of Altra Stratix. The difference to the earlier version is
that it has modified logic array block used for implementation arbitrary logic function. The
Stratix 2 logic array block has 8 adaptive logic modes (ALMs). Figure 11 shows high level
schematic of adaptive logic mode.

Fig 11. High level diagram of the Stratix II ALM [Altera05b.]

There is four mode of operation in ALM arithmetic, share arithmetic, extended LUT and normal
mode. There is four output in normal mode of operation, two is generated by register of ALM and
other two are generated by combination logic circuit. these are configure in several different ways :
a pair of four input LUTs, a pair of five input LUTs, a pair of six input LUTs, a five and a three
27

input LUTs , a five and a four input LUTs and many more. In latter three cases, some input signal is
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

share between different LUTs. It can also implement 6 input and one output function. Also using
extended LUT mode, 7-input function can also be implemented in single ALM.
In the arithmetic mode ALM is suitable for circuit such as accumulator and adder, in addition to
four outputs in normal mode one more carry output also produces. The carry out of one ALM is
carry-in signal for adjacent ALM and connected through routing for fast propagation.
In similar to the arithmetic mode share arithmetic is also same except it has additional carry chain.
In shared arithmetic mode the second carry chain is fed into next adder which is in same or in
adjacent ALM. For second carry chain carry-in signal to share_airth_in and carry-out signal is the
shared_airh_out signal. Due to the presence of two carry chains in ALM it is possible to implement
a circuit that can add three two bit numbers. This design is suitable of designing adder trees.

Summery

In this chapter, we discussed requirement of FPGA and evolution of FPGA, design architecture,
basics of combinational and sequential circuit implementation using FPGA. We also further
discussed regarding the implementation of asynchronous and synchronous sequential logic circuit.
The internal architecture of LUT base circuit also discussed here. Stratix ALM also briefly
discussed in this chapter.

References
[1] Meyer-Baese, U., & Meyer-Baese, U. (2007). Digital signal processing with field
programmable gate arrays (Vol. 65). Berlin: springer.
[2] Monmasson, E., & Cirstea, M. N. (2007). FPGA design methodology for industrial control
systems—A review. IEEE transactions on industrial electronics, 54(4), 1824-1842.
[3] Weste, N., Harris, D., & Banerjee, A. (2005). Cmos vlsi design. A circuits and systems
perspective, 11, 739.
[4] Sulaiman, N., Obaid, Z. A., Marhaban, M. H., & Hamidon, M. N. (2009). Design and
implementation of FPGA-based systems-a review. Australian Journal of Basic and Applied
Sciences, 3(4), 3575-3596.
[5] Roy, S., & Bhunia, C. T. (2014, January). Minimization algorithm for multiple input to two
input variables. In Control, Instrumentation, Energy and Communication (CIEC), 2014
International Conference on (pp. 555-557). IEEE.
28
Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

[6] Ghosh, S., & Ray, S. S. (2007). 4th Generation Programmable Logic Computing: A Road Map.
IETE Technical Review, 24(6), 439-452.
[7] Xia, Q., Robinett, W., Cumbie, M. W., Banerjee, N., Cardinali, T. J., Yang, J. J.,. & Snider, G.
S. (2009). Memristor− CMOS hybrid integrated circuits for reconfigurable logic. Nano letters,
9(10), 3640-3645.
[8] Brown, S., & Rose, J. (1996). Architecture of FPGAs and CPLDs: A tutorial. IEEE Design and
Test of Computers, 13(2), 42-57.
[9] Kumar, R., Roy, S., & Bhunia, C. T. (2016). Study of Threshold Gate And CMOS Logic Style
Based Full Adder Circuits. In Proc. IEEE, 3rd Int. Conference on Electronics and
Communication Systems (ICECS), IEEE (pp. 173-179).
[10] Roy, S., & Bhunia, C. T. (2015). On synthesis of combinational logic circuits. International J of
Computer Applications, 127(1), 21-6.
[11] McAuley, A. J. (1992). Four state asynchronous architectures. IEEE transactions on computers,
41(2), 129-142.
[12] Wang, X., & Ziavras, S. G. (2004). Parallel LU factorization of sparse matrices on FPGA‐based
configurable computing engines. Concurrency and Computation: Practice and Experience,
16(4), 319-343.
[13] Smith, S. C. (2007). Design of an FPGA logic element for implementing asynchronous NULL
convention logic circuits. IEEE Transactions on Very Large Scale Integration (VLSI) Systems,
15(6), 672-683.
[14] Tee, H. H. (2010). FPGA unsolicited commercial email inline filter design using Levenshtein
distance algorithm and longest common subsequence algorithm (Doctoral dissertation,
University of Malaya).
[15] Sasao, T., Nagayama, S., & Butler, J. T. (2007). Numerical function generators using LUT
cascades. IEEE Transactions on Computers, 56(6), 826-838.
[16] Cherepacha, D., & Lewis, D. (1996). DP-FPGA: An FPGA architecture optimized for
datapaths. VLSI Design, 4(4), 329-343.

Authors

Gopal Krishna: [Link] in VLSI & Embedded System Design (Department of

Electronics & Communication Engineering) from National Institute of Technology,
Arunachal Pradesh, India. Completed [Link] from Poornima College of
29

Engineering, Jaipur, under Rajasthan Technical University.

Page

Technical and Scientific Publisher

[Link]
 Fundamentals of FPGA Architecture
Gopal Krishna and Sahadev Roy
Advanced Engineering
Technical and Scientific Publisher
Ch. 2, pp. 12-30, 2017

krishna100491@[Link]

Sahadev Roy: Author’s is currently working as an Assistant Professor in the

Department of Electronics & Communication Engineering at National Institute of
Technology, Arunachal Pradesh, India. He has more than 13 years’ experience in
research and teaching field. He received his [Link]. degree in Chemistry from
Burdwan University, [Link] degree in Electronics and Communication
Engineering from Kalyani University and [Link] in Mechatronics from Bengal
Engineering and Science University, Shibpur India. He received PhD in Electronics
and Communication Engineering from National Institute of Technology, Arunachal
Pradesh. He has published more than 50 research articles in International referred
journals. He served as the Editor-in-Chief of International Journal of Advanced
Engineering and Management (ISSN 2456-8066) monthly blind peer-reviewed
journal published by the Technical and Scientific Publisher. His research interest
includes Robotics, Microwave and VLSI design.
sahadevroy@[Link]

30
Page

Technical and Scientific Publisher

[Link]

View publication stats