Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

JOURNAL OF ELECTRONIC TESTING: Theory and Applications 19, 233–244, 2003

c 2003 Kluwer Academic Publishers. Manufactured in The Netherlands.

LFSR Characteristic Polynomials for Pseudo-Exhaustive TPG with Low


Number of Seeds∗

DIMITRI KAGARIS AND SPYROS TRAGOUDAS


Department of Electrical and Computer Engineering, Southern Illinois University, Carbondale, IL 62901

Received March 14, 2001; Revised March 25, 2002

Editor: S. Kajihara

Abstract. A fully scanned digital circuit can be tested pseudo-exhaustively by first introducing a number of extra
bypass storage cells to limit the test-phase input dependency of each test-phase output and then using a Linear
Feedback Shift Register (LFSR) to feed the chain of the original scan cells and the extra cells. For the design of
the LFSR, the goal is to minimize the pseudo-exhaustive test length with low hardware overhead. If the LFSR uses
a primitive characteristic polynomial then it requires only one seed, but the candidate primitive polynomials may
all fail to satisfy the target test length. In this paper, we present a methodology that enlarges the list of candidate
polynomials, if the prescribed number of seeds is more than one. Experimental results show that the new candidate
polynomials are often instrumental in satisfying the given test length and seed restriction.

Keywords: built-in self-test (BIST), test pattern generation (TPG), Linear Feedback Shift Registers (LFSR)

1. Introduction test phase) in to segments satisfying the condition. For


such partition methods see, for instance [3, 4]. We as-
Linear Feedback Shift Registers (LFSRs) are widely sume here a partitioned circuit for some specified value
used for built-in self-test (see, e.g., [1]) of digital w, which in practice, is close to 25.
combinational or fully scanned circuits, for pseudo- Consider a fully scanned circuit, where the set of
exhaustive or pseudo-random test pattern generation original flip-flops, along with the test cells for the pri-
(TPG). In pseudo-exhaustive TPG, the goal is to ap- mary inputs and any bypass storage cells that may have
ply to each test-phase output of the circuit all possible been inserted are linked into a single scan chain. The
patterns that it needs to be tested. A necessary condi- w first cells of that scan chain are configured into an
tion for a pseudo-exhaustive test length of 2w is that LFSR with an appropriate characteristic polynomial.
each test-phase output depend on at most w test-phase (Alternatively, the w LFSR cells may be extra cells,
inputs. The set of test-phase inputs that affect the func- distinct from the scan chain cells, but this is not usually
tion of a specific test-phase output will be referred to as the case.) The scheme will be referred to as LFSR/SR,
Aset (Affector set). If the size of an Aset in the original where the SR (Shift Register) part signifies the chain of
circuit is greater than the target value off w, a number the scan cells and/or any extra bypass cells. An example
of additional cells (referred to as bypass storage cells in is shown in Fig. 1. Assume we want to have a pseudo-
[2]) have to be inserted so as to partition the circuit (in exhaustive test set of size no more than 16, i.e., w = 4.
In the original circuit of Fig. 1(a), the Aset of output E
∗ Thiswork was partially supported by NSF grant CCR-0096119. A is A E = {I1 , I2 , I3 , I4 , I5 }, which has size larger than
preliminary version appeared in the Proceedings of the IEEE Inter- w = 4. A bypass storage cell (cell C6 in Fig. 1(b)) can
national Conference on Computer Design, Sept. 2000, pp. 42–47. be inserted to satisfy the requirement for w. Then the
234 Kagaris and Tragoudas

C_0 C_1 C_2 C_3 C_4

I_1 I_2
I_3 I_4 I_5 I_1 I_2 I_3 I_4 I_5

A B A B

C D C D

FF C_6 C_5

E F E F

(a) (b)

Fig. 1. (a) Original circuit. (b) Overall scan chain and LFSR.

primary inputs (cells C0 –C4 ), the original flip-flop (cell of the LFSR are shifted through the rest of the chain,
C5 ) and the bypass storage cell (cell C6 ) are connected they provide test patterns for the test-phase outputs.
to a scan chain (local conversions for the flip-flops and The efficacy of this scheme depends on the choice
the multiplexers for the primary inputs are not shown of an appropriate characteristic polynomial. The prob-
in the figure). Cells C0 –C3 are configured into an LFSR lem is that linear dependencies (see., e.g., [6]) may
with characteristic polynomial P(x) = x 4 + x 3 + 1. occur between Asets. In order to guarantee a pseudo-
The test-phase inputs are C0 –C6 , the test-phase out- exhaustive test set, the characteristic polynomial must
puts are C, D, E, F, and the Asets of the test–phase satisfy certain criteria, one of which was described in
outputs are AC = {C0 , C1 , C2 , C3 }, A D = {C2 , C3 }, [7–9] and has been widely used in practice. According
A E = {C0 , C1 , C4 , C6 }, and A F = {C4 , C5 , C6 }. (In to this criterion,
the following we assume that the Asets consist actu-
ally of the numerical indices of the cells in the chain.) Lemma 1.1 (Criterion of [7–9]). An LFSR with char-
When the LFSR part is initialized by a non-zero vector acteristic polynomial P(x) of degree d ≥ w yields
(seed), it cycles through p distinct states, where p is a pseudo-exhaustive test set of size (2d − 1), if and
the period of the characteristic polynomial. The period only if, in every Aset Ai = {a1 , a2 , . . . , awi }, wi ≤ w,
of a polynomial is the least integer p for which the all polynomials x a1 mod P(x), x a2 mod P(x), . . . ,
polynomial divides x p + 1 [5]. As the successive states x awi mod P(x) are linearly independent.

You might also like