Testing of memory using franklin method

IJRET: International Journal of Research in Engineering and Technology ISSN: 2319-1163
__________________________________________________________________________________________
Volume: 01 Issue: 04 | Dec-2012, Available @ http://www.ijret.org 593
TESTING OF MEMORY USING FRANKLIN METHOD
Mythri Myneni1
, Madhavi Mallam2
1,2
Assistant professor, DVR&Dr.HimaSekhar,MICcollegeof Technology, kanchikacherla
mythri.ece07@gmail.com and madhavi.418@gmail.com
Abstract
It is natural to be skeptical when someone says it is possible to determine the state-of-health of a complex integrated circuit with a
few simple tests. Justifiable doubt in the unfamiliar can be aided by experience and investment in a traditional approach [5]. So
only the process known as testing is involved after each and every design. Fault may occur even in the memory .So we are
going to test the memory in this paper by the most suitable method [4]. Therefore we are going to test the memory using one of the
pattern sensitivity test method i.e., Franklin method by Verilog approach [6]. The results obtained are in excellent agreement
with theoretical results.
Keywords: Memory, RAM, Pattern sensitivity test, Franklin Method, Verilog approach.
--------------------------------------------------------------------******-------------------------------------------------------------------
1. INTRODUCTION
The use of redundancy to increase the yield of random access
memories (RAM’s) has become very popular. State-of-the-art
memory chips are now invariably designed with spare rows
and spare columns. Once a memory chip is reconfigured,
physically adjacent cells in the spare rows and columns that
were brought in are very unlikely to have consecutive
logical addresses. Test algorithms used at the stages for the
detection of physical neighbourhood faults in these RAM
has to consider that the logical and physical neighbourhoods
of some of the cells are not identical’ [2]. A simple solution to
this problem might be to store the actual logical-to- physical
address mapping of the relevant memory cells for later
recall. In an external tester environment, the tester has to
derive the address mapping either by using post repair
diagnostic techniques such as signature and roll call [SI or
by using Stroboscopic Scanning Electron Microscopes
(SSEM) modify the test algorithm according to the mapping,
and then apply the test patterns[1]. In a Built-In Self-Test
(BIST) environment, the BIST logic either has to be modified
to incorporate the reconfiguration done to the RAM array, or
must have the means to determine the mapping. Either of
these solutions increases the complexity of the BIST logic[2].
To decrease the complexity of the design we are using the
Franklin and Saluja Method.
2. RAM CELL
Before embarking on the testing problem, we shall briefly
review the organization of RAM’s (in particular,
reconfigurable RAM’s), as it plays a major role in
determining the relationship between physical and logical
neighbourhoods. A RAM chip consists of memory cell
arrays, address decoders, address registers, data registers,
and read write logic [2]. The memory cells are physically
organized as a 2 x k 2-D grid of identical m x n 2-D arrays
of cells, as shown in Figure.2.1
Fig: 2.1: RAM cell
This type of partitioning is done to keep the active power,
peak currents, access times, and cycle times within
reasonable limits. By far, the most popular approach to
incorporate redundancy in RAM’s is to add spare rows
(word lines), spare columns (bit lines), or both for each
memory array. In some RAM’s, a spare row or column can
be shared by two or more arrays. In some others, rows or
columns cannot be replaced singly; instead, they must be
replaced in pairs, quads, or larger groups. Interested readers
are referred to [2], for further details of different
redundancy schemes and reconfiguration algorithms. When
a RAM chip is physically organized as a collection of

__________________________________________________________________________________________
smaller arrays, cells and their contents in each of these
physical arrays are independent of the cells in the remaining
arrays. By assuming that no interaction can take place
between cells of different physical arrays, we need consider
only a 2- D physical array for modeling faults, allowing
each array tobe tested independently from the rest of the
chip. If spares are shared among a group of physical arrays,
then each such group of arrays must be tested as a single
unit to detect faults due to interactions between physically
adjacent spare rows (and columns)[2].
Fig: 2.2 : RAM
2.1. RAM Memory Cell:
As RAM density increases, PSF’s become the predominant
faults [9]. Moreover, other memory fault classes such as
shorts, stuck-at faults, and coupling faults can be regarded
as special types of PSF’s. Because testing for unrestricted
PSF’s is impractical, researchers have considered restricted
neighbourhoods, which assume interactions to be limited to
cells within certain physical proximity. These are called
physical neighbourhood PSF’s (PNPSFs). The most
frequently considered neighbourhood is the five-cell
physical neighborhood, in which the four physical neighbors
of a cell are often called N (North), E (East), W (West), and
S (South).
Static PSF (SPSF): In this fault model, a cell is said to be
faulty if its contents change when a certain pattern of 0’s
and -1’s exists in the neighborhood cells.
Dynamic PSF (DPSF): In this fault model, a cell is said to
be faulty if its state changes because of a change in its
neighbourhood pattern [2].
2.2. Partitioning and Testing of 8×8 RAM array
To apply the patterns, namely the patterns in which the N
and S physical neighbors of the cell have distinct values, we
consider all possible pairs of rows in the array as the N-S
neighbors, and initialize the rows in each pair to distinct
values. For identifying all pairs of rows, we use the
following partitioning technique. In a series of steps,
successively partition the set of rows into two; the partitions
at any step can be determined by a specific bit position in
the row address [6]. Thus, at the ith iteration, if the ith bit
position of a row's address is 0, then that row belongs to the
first partition of the RAM array for that step; if the bit is 1,
the row belongs to the second partition. The different
figures in Figure 2.1 show the partitioning steps for an 8
x 8 RAM array. This partitioning technique for
identifying all possible pairs of rows is similar to the one
presented for partitioning memory cells. After each
partitioning step, initialize the rows in the first partition to 0
and the rows in the second partition To apply the remaining
patterns, namely the patterns in which the N and S physical
neighbors of the cell have distinct values, we consider all
possible pairs of rows in the array as the N-S neighbors, and
initialize the rows in each pair to distinct values [2]. For
identifying all pairs of rows, we use the following
partitioning technique [4].
3. FRANKLIN METHOD
3.1 Pattern Sensitivity Faults
Testing is a process which includes test pattern generation,
test pattern application, and output evaluation. The capacity
of random-access memories chips enhances, thus increasing
the test time and cost; on the other hand, the density of
memory circuits grows, therefore more failure modes and
faults need to be taken into account in order to obtain a good
quality product. he quality of a test set depends on its fault
coverage (FC) as well as its size. With the increase in
memory density, neighbourhood pattern sensitive faults
(NPSFs) are not only an important fault model for DRAMs
but will also become so for SRAMs. RAM errors will
become more of a problem as memory sizes grow and
device geometry shrinks With the growing in memory
density and the shrinking into design rule, neighbourhood
pattern sensitive faults (NPSFs) are becoming more and
more important. Though previously proposed patterns for
NPSFs have sufficient fault coverage, the complexity of
those requires long test time. These are common in high
density RAMs. Before designing a test it's wise to consider
exactly what kinds of failures may occur. Contents of
memory cell are affected by contents of neighbouring cells.
Common in memory cells of high density RAMs. Detected
with specific memory test algorithms.
3.2 Franklin and Saluja algorithm:
Franklin and Saluja in presented algorithms that test
reconfigured RAM’s and scrambled address RAM’s for
five-cell and nine-cell physical neighbourhood pattern
sensitive faults (PSF’s)[2]. Those algorithms require O (N
[10g3N14) reads and writes to test an N-bit RAM array.
Furthermore, those algorithms are not simple enough to

__________________________________________________________________________________________
allow easy BIST implementations. We can easily understand
the Franklin Method by the following procedure.
Assume the RAM memory cell as portioned rows and
columns as follows:
Columns 0 1 2 3
Rows
0 1 1 1 1
1 1 1 1 1
2 1 1 1 1
3 1 1 1 1
First change the 3rd column
Columns 0 1 2 3
Rows
0 1 1 1 0
1 1 1 1 0
2 1 1 1 0
3 1 1 1 0
Then change the 0th row
Columns 0 1 2 3
Rows
0 0 0 0 1
1 1 1 1 0
2 1 1 1 0
3 1 1 1 0
Then change the 2nd column
Columns 0 1 2 3
Rows
0 0 0 1 1
1 1 1 0 0
2 1 1 0 0
3 1 1 0 0
Then change the 1st row
Columns 0 1 2 3
Rows
0 0 0 1 1
1 0 0 1 1
2 1 1 0 0
3 1 1 0 0
Then change the 1st column
Columns 0 1 2 3
Rows
0 0 0 1 1
1 1 1 0 0
2 1 1 0 0
3 1 1 0 0
Continue this till 0th column so after 3rd
row
We will get the answer as follows:
Columns 0 1 2 3
Rows
0 0 0 0 1
1 0 0 0 1
2 0 0 0 1
3 0 0 0 1
Like this we need to continue till 1st column and
changing all the rows [4].
And finally if we change the 0th column we will get
the same answer.
i.e.
Columns 0 1 2 3
Rows
0 1 1 1 1
1 1 1 1 1
2 1 1 1 1
3 1 1 1 1
Suppose if there is any fault in any of the row or column it
may be 0.So by observing the rows and coloumns we can
say where the fault occurs.This is one of the way to identify
the fault by using Franklin and Saluja method.
4. RESULTS
4.1.RAM(FIFO) Without Fault:

__________________________________________________________________________________________
4.2. RAM (FIFO) With Fault:
CONCLUSIONS
In this work, we have succeeded in generating Structural
test data from HDL functional descriptions of a memory cell
(RAM). The proposed technique is based on a software
testing technique: Franklin and Saluja Method [4].High low-
level fault coverage can be achieved with short high-level
test sequences but further investigations should improve the
current results. Even if our study was initially limited to a
given type of memory cell, the first results obtained,
whatever the number of fault sequences or the fault
coverage, encourage us to apply our approach to other types
of memory cells. We can eliminate the maximum number of
faults being occurred in the device as the memory is tested.
REFERENCES
[1] Richard A. DeMillo, Aditya P. Mathur, and W.Eric
Wong. “Some Critical Remarks on a Hierarchy of Fault-
Detecting Abilities of Test Methods” IEEE
TRANSACTIONS ON SOFTWARE ENGINEERING,VOL.
21, NO. 10, OCTOBER 1995.
[2] Manoj Franklin and Kewal K.Saluja.“Reconfigured
RAM’s and Scrambled RAM’s for Pattern Sensitive Faults.”
[3] Michael J. Liebelt “Detecting Exitory Stuck-At Faults in
Semi modular asynchronous Circuits”.
IEEE TRANSACTIONS ON COMPUTERS, VOL.48,
NO.4,APRIL 1999.
[4] Parag K. Lala. “An Introduction to Logic Circuit
Testing”.Editor: Mitchell Thornton,Southern Methodist
University.Morgan & Claypool Publishers,2009.
[5] M. Abramovici, M.A. Breuer, and A.D.
Friedman,Digital Systems Testing and Testable Design,
IEEE,1990.
[6] R.Leveugle. “Fault Injection in VHDL Descriptions and
Emulation”. TIMA Laboratory.

Testing of memory using franklin method

More Related Content

What's hot (15)

Viewers also liked (20)

Similar to Testing of memory using franklin method (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

Testing of memory using franklin method