Error detection/correction and fault detection/recovery – Pulse or data error handling – Memory testing
Reexamination Certificate
2000-11-27
2003-04-15
Decady, Albert (Department: 2133)
Error detection/correction and fault detection/recovery
Pulse or data error handling
Memory testing
C365S189040, C365S201000
Reexamination Certificate
active
06550026
ABSTRACT:
TECHNICAL FIELD
This invention relates generally to computers, and more particularly to memory devices and methods of testing such devices.
BACKGROUND OF THE INVENTION
Computer designers desire fast and reliable memory devices that will allow them to design fast and reliable computers. Manufacturers of memory devices, such as random access semiconductor memories, must test a full range of functionality and timing characteristics of the memory devices in order to provide a reliable product to their customers. Because each memory cell of the device must be tested, the time and equipment necessary for testing memory devices having increasing density represents a significant portion of the overall manufacturing cost of such devices. Any reduction in the time to test each unit will reduce manufacturing costs.
Semiconductor manufacturers have developed fast testing routines to allow a greater number of chips to be tested simultaneously using a given testing device. One known testing routine, Jedec, simply compares the data written to a memory device with the data read from that memory device, and assigns a 1 value to one or more memory cell addresses if the data matches (passes), or a 0 if the data does not match (fails). While the Jedec routine is fast, it does not output the actual data written to the memory device. As a result, if the tester outputs a continuous string of 1 s, indicating that the memory device passes, a technician is unsure whether the device actually passes, or if an error has occurred in the device, or at some point along the path from the device to the tester, to cause such an output.
To compensate for this shortcoming of the Jedec routine, a Micron Test Mode Routine provides three outputs. The Micron Routine outputs the actual data, as a 0 or a 1, and a mid-level tri-state value therebetween. If the tri-state value is output, rather than a 1 or a 0, the technician recognizes that an error has occurred. Unfortunately, while the Micron Routine provides superior testing of most semiconductor devices, the routine typically cannot bias the output back to the tri-state value before the beginning of the next read/write cycle rapidly enough to allow current high-speed memory devices to be tested at their normal operating speed. As a result, such high-speed memory devices must be tested at speeds slower than their typical operating speed.
To save testing time and cost, manufacturers of memory devices increasingly automate the testing procedure so that a tester applies the testing routine simultaneously to several chips. Automated testing is most easily accomplished after the memory device has been packaged as a semiconductor chip, because the chip can be automatically inserted into a test socket using pick and place machinery. Automated testing circuitry then performs the testing routine by applying predetermined voltages and signals to the chip, writing test data patterns to the memory, reading data, and analyzing the results to detect memory speed, timing, failures, etc. The more chips that can be tested simultaneously, the greater testing time savings per chip.
Most testers used in testing semiconductor chips are expensive. For example, a current tester manufactured by Teradyne has 128 input/output (“I/O”) lines. To maximize the number of chips that this tester can test simultaneously, the on-chip data input/output lines, or “DQ lines,” are multiplexed so that fewer I/O lines from the tester are required to be coupled to each chip. For example, the tester writes a predetermined data pattern simultaneously to multiple locations in each memory device and then accesses the written data during a read cycle. Comparator circuits fabricated on-chip compare the data read from the multiple locations and indicate whether all the data read matches the data written. If the chip has 32 DQ lines (DQ
0
-DQ
31
), on-chip 4:1 multiplexers and testing circuitry compress data onto only 8 of the 32 DQ lines. As a result, only 8 of the 128 lines of the tester are required for each chip. Consequently, the tester's 128 I/O lines can simultaneously test 16 chips.
In another solution, certain semiconductor memory devices, manufactured by Micron Technology, Inc., provide on-chip test mode circuitry that helps compensate for such delays during testing of devices. Under such test mode circuitry, the external testing device writes data to the chip during a first interval, and then writes the same data again to the DQ lines during a second interval. During the second interval, while the data is written again to the DQ lines, the data previously written to the memory device is read therefrom and latched. On-chip comparators then compare the latched data to the data written during the second interval. If the latched data equals the data written during the second interval, then the chip passes. Such a device can rapidly analyze the read data written to the device.
While the above solutions can detect for typical cell-to-cell defects and functionality of the chip, they cannot accurately test the speed of the chips. As semiconductor memory chips provide increasingly faster data I/O rates, particularly with synchronous DRAMs, data is required to be transferred to and from the chips in as little as 9 nanoseconds or less, based on a 10-nanosecond or faster clock cycle. As a result, such chips provide only a 1-nanosecond margin of error. Today's increasingly fast memory devices require highly precise generation of timing signals and precise measurement of the memory device's response thereto. Gate delays caused by the multiplexing circuitry required during testing cause the data to be read from the chips in greater than 10 nanoseconds. As a result, the tester cannot determine if the chip accurately output data within the required 9 nanoseconds. In other words, the on-chip testing circuitry prohibits the tester from testing the speed of such chips.
Obviously, it is desirable to determine the performance, and thus the speed of, semiconductor memory chips, especially high-speed chips. Additionally, because of manufacturing process tolerance and variations, one memory device of a particular design may be faster than another memory device of the very same design. Manufacturers therefore typically also desire to test the speed of such chips so that such chips can be sorted based on speed grades. To provide such speed testing, typical address compression mode testing, and on-chip multiplexing of DQ lines, must be abandoned. As a result, where 16 or more chips could previously be simultaneously tested using multiplexing, only 4 of such chips can be simultaneously speed tested because all 32 DQ lines of each chip must be coupled to the tester's I/O lines. As a result, there is a need to simultaneously speed test an increasing number of chips using a given tester.
One solution has been to purchase a larger number of testers, or more expensive testers having a greater number of I/O lines. However, as noted above, such testers are quite expensive.
SUMMARY OF THE INVENTION
The present invention provides a semiconductor device that can be efficiently speed tested, and which overcomes at least the shortcomings of the prior art discussed above. The memory device requires, at a minimum, only two I/O lines from an external testing device to be coupled thereto. A first DQ line from the memory device provides a direct data path from the array so that the external tester can read data from the array at the maximum speed of the memory device. Test mode circuitry for multiplexing and comparing multiple DQ lines in address compression mode is coupled to two or more DQ lines, including the first DQ line. The compression mode testing circuitry can include on-chip comparators that compare the data simultaneously written to, and read from, the memory device. The comparison circuitry outputs a data test flag indicating whether the data read from the memory device matches or whether the data does not match. The test flag is output through a multiplexer to a second DQ line. As a result, the speed of the device ca
Fuller Paul M.
Wright Jeffrey P.
Zheng Hua
De'cady Albert
Lamarre Guy
LandOfFree
High speed test system for a memory device does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with High speed test system for a memory device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and High speed test system for a memory device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3087315