Error detection/correction and fault detection/recovery – Pulse or data error handling – Memory testing
Reexamination Certificate
2003-02-10
2004-09-28
Moise, Emmanuel L. (Department: 2136)
Error detection/correction and fault detection/recovery
Pulse or data error handling
Memory testing
C714S724000, C714S733000, C714S030000, C324S1540PB
Reexamination Certificate
active
06799289
ABSTRACT:
TECHNICAL FIELD
The present invention relates generally to testing of integrated circuits, and more specifically to a method and apparatus that reduces the time and testing resources needed for testing of memory integrated circuits.
BACKGROUND OF THE INVENTION
Integrated circuits are extensively tested both during and after production, and, in some cases, routinely during use after they have been installed in products. For example, memory devices, such as dynamic random access memories (“DRAMs”), are tested during production at the wafer level and after packaging, and they are also routinely tested each time a computer system using the DRAMs executes a power-up or “boot” routine when power is initially applied to the computer system. As the capacity of DRAMs and other memory devices continues to increase, the time require to test the DRAMs continues to increase, even though memory access times continue to decrease.
A typical RAM integrated circuit includes at least one array of memory cells arranged in rows and columns. Each memory cell must be tested to ensure that it is operating properly. In a typical prior art test method, data having a first binary value (e.g., a “1”) are written to and read from all memory cells in the arrays, and thereafter data having a second binary value (e.g., a “0”) are typically written to and read from the memory cells. A memory cell is determined to be defective when the date that is read from the memory cell does not equal the data that was written to the memory cell. As understood by one skilled in the art, other test data patterns may be utilized in testing the memory cells, such as an alternating bit pattern, e.g., 101010. . . , written to the memory cells in each row of the arrays.
One situation requiring testing of memory integrated circuits occurs during fabrication of memory integrated circuits. Fabrication yields are reduced when fabrication errors occur. Testing of memory integrated circuits during fabrication allows the sources of some fabrication errors to be promptly identified and corrected.
Another situation requiring testing of integrated circuits also occurs in fabrication of memory integrated circuits. Defective memory cells are identified by testing and are replaced with non-defective memory cells from a set of spare or redundant memory cells. In one conventional method for replacing defective memory cells, fuses on the integrated circuit are blown in a pattern corresponding to the addresses of defective memory cells. The pattern is then read to select redundant memory cells to replace the defective memory cells.
FIG. 1
is a simplified block diagram of several integrated circuit memory devices
10
and an automated tester
12
according to the prior art. Separate buses
14
couple each of the memory devices
10
on a circuit board
16
to the automated tester
12
through a connector
18
. The buses
14
convey stimuli, such as write data, from the automated tester
12
to the memory devices
10
that are being tested. Transmission of the write data to the memory devices
10
does not require separate buses
14
because the same data are typically written to all of the memory devices
10
.
Each memory device
10
generates a response, such as read data, from the data that are written to that memory device
10
. The buses
14
convey the read data from each memory device
10
back to the automated tester
12
. The automated tester
12
compares the read data from each memory device
10
to expect data, which correspond to the write data. The expect data thus correspond to read data that would be provided by the memory device
10
if it was operating properly. When the read data and the corresponding expect data match, the memory device
10
is considered to be functioning normally. When the read data do not match the corresponding expect data, the memory device
10
is considered to be malfunctioning.
If the memory devices
10
are read at the same time, transmission of the read data from the memory devices
10
requires separate buses
14
because read data resulting from failures in one or more of the memory devices will differ from each other. When these differing read data are transmitted over a common bus, contention between the differing data results in ambiguity as to which memory device
10
provided the read data corresponding to the defective memory cell and may also result in ambiguity in determining the data that are present on the bus
14
. The automated tester
12
must therefore read data from each of the memory devices
10
individually. As a result, reading data from the memory device
10
requires more time than does writing to the memory devices
10
. The requirement that data be read from each memory device
10
individually results in relatively long test times. Yet test time is a significant cost factor for manufacturers of memory devices
10
. Additionally, since longer tests increase the number of automatic testers required to test a given number of memory devices
10
, the cost of testing in further increased, particularly since automated testers
12
may cost several million dollars apiece.
Testing times may be minimized by testing multiple memory devices
10
at the same time. However, each automated tester
12
can only accommodate a finite number of buses
14
, thereby limiting the number of memory devices
10
that may be simultaneously tested. The number of memory devices
10
that may be simultaneously coupled to the automated tester
12
is known as the “fanout” for the automated tester
12
. One factor limiting fanout for each automated tester
12
involves the connector
18
that couples the circuit board
16
holding the memory devices
10
. A practical upper limit for the number of pins on each connector
18
is about 300. Larger numbers of pins tend to result in connectors
18
that are not sufficiently reliable. As a result, the number M of memory devices
10
that can be simultaneously tested is limited to about 300/N, where N is the number of connections that an be made to each memory device
10
. For example, an automated tester
12
having a capacity of about 300 data lines may be employed to simultaneously test, without data compression, two memory devices
10
having 128 bit wide data buses, or four memory devices
10
having 64 bit wide data buses. With data compression, the same automated tester
12
may test, for example, sixteen (or possibly even up to eighteen) memory devices
10
, but will only be able to receive 16 bits of read data from each memory device
10
.
In any of these cases, the time required to analyze the read data obtained by testing the memory devices
10
often exceeds the time required to perform the tests. The time needed to analyze the test results may decrease when data compression is used, but the results of testing done using data compression may be ambiguous or may not be useful for some purposes. For example, testing using data compression may be incapable of identifying a specific faulty memory cell but instead may be capable of identifying only a group, e.g., a row or column, of memory cells containing the defective memory cell.
As a result, data compression tests tend to be “go
o-go” tests for a given memory device
10
as a whole, rather than diagnostic tests providing detailed information (e.g., addresses) regarding specific defective memory cells. However, data compression tests may be used to replace groups of rows or columns when the compressed data show that at least one of the rows or columns in the memory array includes one or more defects.
While compressed data do not always support the repair operations described above, they are extremely useful for other test purposes because they greatly speed testing. These other test purposes include speed grading (i.e., determining the maximum clock frequency permitting reliable operation) of memory device
10
.
In speed grading, the number of failures is relevant, and some inaccuracy in the measured number of failures may be quite acceptable. The number of failures measured in a test using read data
Snodgrass Charles K.
Totorica Robert L.
Dorsey & Whitney LLP
Micro)n Technology, Inc.
Moise Emmanuel L.
LandOfFree
On-board testing circuit and method for improving testing of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with On-board testing circuit and method for improving testing of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and On-board testing circuit and method for improving testing of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3242634