Walking 1's

In more detail: use a pattern of address or data that has 0 in every bit position but one; and as you run your test, progressively shift the 1 bit along the data so that you put it on each data line in turn. Easy to implement and check, and very useful to test whether you've got all your data and address lines wired up correctly. Be careful about *where* you do your checking, to avoid missing errors in one part of the data path that are corrected later in another part because you have the same mis-wiring error in two places.

Walking-1s are also good for locating "stuck-at" bits, but it's usually quicker to find them by using just two test patterns: all-1 and all-0.

--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL * Verilog * SystemC * e * Perl * Tcl/Tk * Project Services

Doulos Ltd., 22 Market Place, Ringwood, BH24 1AW, UK
jonathan.bromley@MYCOMPANY.com
http://www.MYCOMPANY.com

The contents of this message may contain personal views which 
are not the views of Doulos Ltd., unless specifically stated.

"Walking ones" refers to a set of data patterns where, in each data word, a single bit is set to one and the rest to zero.

Similarly, "walking zeroes" refers to a set of data patterns where, in each data word, a single bit is set to zero and the rest to one.

Such sets of data patterns have been used to test (or diagnose) memory faults. For example, where the data width is eight bits, we may try to write a set of patterns:

00000001 00000010 00000100 00001000 (The "one" bit "walks" along the data word) 00010000 00100000 01000000 10000000

and, in each case, read the same memory location to check that the same data is recovered.

The choice of such patterns is based on their simplicity and on the ease of diagnosing (from the recovered data) any faults which they detect. They are most useful for the detection and diagnosis of interconnection faults where two data signals are shorted together (or show "crosstalk").

Other data patterns may be chosen according to the type of fault thought to be likely with a particular memory technology. For example, in the case of magnetic "core" memory, certain patterns are chosen on the basis that they are considered to be the most likely to expose faults in that type of memory. Such data patterns may be referred to as "disturb" patterns and may extend over several data words.

Note that a simple "walking ones" or "walking zeroes" pattern is usually of limited use in detection of memory addressing faults (e.g. where a memory address signal is shorted to a fixed logic level or two memory address signals are shorted together).

In practice, almost all memory faults are "hard" failures, which can be identified simply by writing all ones or all zeroes to each address. "Soft" failures can be identified and corrected in working systems by use of error-correction (e.g. Hamming) codes, without disturbance of the running program. Ideally, where this is done, the occurrence of a corrected or uncorrected fault is notified (by interrupt) to the system error handler, so that appropriate action can be taken. For example, the system error handler will usually begin the procedure of logging the fault.

For these reasons, walking ones and walking zeroes are of limited use, except to detect shorted data signals in manufacturers' tests.

Mike

Walking 1s is a term used (generally) for memory testing, although it has other applications.

Cheers

PeteS