|
|
|
| United States Patent | 6395566 |
| Link to this page | http://www.wikipatents.com/6395566.html |
| Inventor(s) | Farnworth; Warren M. (Nampa, ID) |
| Abstract | A multi-chip module (MCM) that fails testing after its assembly is repaired
by generating a wire-bonding solution for a repair die during testing,
storing the repair solution in a computer system in association with a
unique ID code read from the MCM, and then using the repair solution at a
wire-bonding station to correctly bond out a repair die for the MCM. The
use of a stored repair solution at the wire-bonding station eliminates the
need for a human operator to manually select the repair solution, and thus
reduces the opportunity for error while shortening the length of time it
takes to complete the repair process. |
|
|
|
Title Information  |
|
|
|
|
|
|
| Publication Date |
May 28, 2002 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| Parent Case |
CROSS-REFERENCE TO RELATED APPLICATION
This application is a continuation of application Ser. No. 09/056,408,
filed Apr. 7, 1998, now U.S. Pat. No. 6,238,942, issued May 29, 2001. |
|
|
|
|
|
|
|
|
|
|
|
|
|
Title Information  |
|
|
Description  |
|
|
BACKGROUND OF THE INVENTION
1. Technical Field
This invention relates in general to multi-chip modules and, more specifically, to methods for repairing such modules by bonding out a module repair site using a repair solution generated during module testing.
2. State of the Art
As shown in FIG. 1, multi-chip modules (MCMs), comprising multiple dice carried on a printed circuit board (PCB) or other carrier substrate bearing circuit traces, are generally tested after being assembled, and those MCMs that pass testing are
then shipped to customers. Those MCMs that fail testing are marked as such, typically by printing an "X" on the top of the "bad" (i.e., at least partially defective) dice that caused each MCM to fail testing. After being marked, any bad dice on a given
MCM are disabled through various means including, for example, cutting conductive traces on the MCM to isolate the bad dice, or shorting conductive traces together to force a high-impedance condition on all inputs and outputs of the bad dice (e.g., by
pulling the RAS* input of a Dynamic Random Access Memory (DRAM) permanently high).
Once the bad dice are disabled, known-good die (KGD) repair dice are attached to each failing MCM at an empty repair site. The failing MCMs with attached repair dice then proceed to a wire-bonding station, where an operator manually selects the
wire-bonding pattern (also referred to as a wire-bonding "solution" herein) necessary to connect the repair dice to traces on the carrier substrate to replace the disabled bad dice. The repair dice are wire-bonded using the manually selected
wire-bonding solution, and the repaired MCMs are then retested, with those repaired MCMs that pass the retesting proceeding on to shipping, and those repaired MCMs that fail retesting proceeding once again to repair or, alternatively, to scrap.
Because a wire-bonding machine operator manually selects a particular wire-bonding pattern for each MCM to be repaired, the process of repairing MCMs is subject to human error. Specifically, the operator generally selects the particular
wire-bonding pattern to be used based on which dice are marked as bad (e.g., with the "X" described above), so the operator may make a mistake in observing which dice are marked as bad, or in selecting the appropriate wire-bonding pattern, or both.
Therefore, there is a need in the art for a method of selecting a wire-bonding solution for a MCM to be repaired that avoids the chance for human error associated with the conventional method described above.
SUMMARY OF THE INVENTION
In one embodiment of this invention, a multi-chip module (MCM) proceeds to testing after it is assembled. During testing, an ID code (e.g., a bar code) of the MCM is read. Then, if the MCM fails testing, a wire-bonding repair solution is
generated and stored in a computer system in association with the ID code. The wire-bonding repair solution may identify, for example, which of several sets of bonding pads on the surface of a substrate of the MCM a repair integrated circuit (IC) die or
dice should be wire-bonded to. A bad IC die or dice of the MCM that caused the MCM to fail testing may then be disabled by, for example, disconnecting the bad IC die or dice from the rest of the MCM, or by permanently connecting the bad IC die or dice
so all of its inputs and outputs are in a high-impedance state. The repair IC die or dice is then attached to the substrate of the MCM, the stored repair solution is accessed, and the repair IC die or dice is wire-bonded to the substrate in accordance
with the accessed repair solution. As a result, the opportunity for human error is avoided, and the length of time it takes to repair the MCM is shortened.
BRIEF DESCRIPTION OF THE FIGURES
FIG. 1 is a flow diagram illustrating a conventional method for wire-bonding a repair die in a multi-chip module (MCM);
FIG. 2 is a flow diagram illustrating a method in accordance with this invention for wire-bonding a repair die in a MCM;
FIG. 3 is a top view of a MCM illustrating the arrangement of repair sites on the MCM; and
FIG. 4 is an isometric view showing the MCM of FIG. 3 in more detail.
DETAILED DESCRIPTION OF THE ILLUSTRATED EMBODIMENTS
As shown in FIGS. 2 and 3, the manufacture of a multi-chip module (MCM) 10 using the principles of this invention begins in assembly, where integrated circuit (IC) dice 12 and 14 are attached to a printed circuit board (PCB) 16 or other carrier
substrate and are then wire-bonded to conductive traces 32 (FIG. 4) on the surface of the PCB 16.
It will be understood by those having skill in the technical field of this invention that the invention is applicable to any MCM including, for example, Dynamic Random Access Memory (DRAM), Single In-Line Memory Modules (SIMMs) and Dual In-Line
Memory Modules (DIMMs). Also, it will be understood that this invention is applicable to any IC dice including, for example, DRAM dice, SRAM dice, Synchronous DRAM (SDRAM) dice, Sync-Link DRAM (SLDRAM) dice, processor dice (e.g., Intel Corp.'s
Pentium.RTM. and Pentium II.RTM. chips), Extended Data-Out DRAM (EDO DRAM) dice, and flash Erasable Electrically Programmable Read Only Memory (EEPROM) dice.
After assembly, the MCM 10 proceeds to testing. If the MCM 10 passes testing, it is generally ready for shipping to customers. If, instead, it fails because of a bad die 14, a wire-bonding repair solution is generated during testing that will
later allow a repair die 18 positioned at one of two repair sites 20 to be wire-bonded such that it replaces the bad die 14. Specifically, this repair solution may identify, for example, which one of multiple sets of pads 30 (see FIG. 4) on the PCB 16
the repair die 18 should be wire-bonded to in order for it to replace the identified bad die 14 (i.e., one set of pads connects the repair die 18 to replace the bad die 14, while the other sets of pads can connect the repair die 18 to replace other dice
12 if they are bad). The repair solution is stored in a computer system 22 in association with a unique identification (ID) code of the MCM 10 read during testing. This ID code may be, for example, a bar code, an Optical Character Recognition (OCR)
code, an optically readable punch hole code, or an electronically readable laser fuse code embedded in the PCB 16.
Once the repair solution is generated, the MCM 10 moves on to a rework station, where the bad die 14 is isolated or disabled by cutting conductive traces 32 (FIG. 4) on the PCB 16 connected to the bad die 14, or by shorting such conductive traces
together in order to bring all inputs and outputs of the bad die 14 to a high-impedance state (e.g., by pulling the RAS* input of a DRAM permanently high).
After the rework station, the MCM 10 moves to a die attach station, where the repair die 18 is attached at the repair site 20 shown in FIG. 3. The particular repair site 20 used, as shown in FIG. 3, is dedicated to repairing any bad dice among
the four dice 12 (including the bad die 14) shown at the top of the MCM 10 in FIG. 3, while the other repair site 20 is dedicated to repairing any bad dice among the four dice 12 shown at the bottom of the MCM 10 in FIG. 3. Of course, it will be
understood that this invention is not restricted to the configuration of repair sites and dice shown in FIG. 3. Rather, this invention includes within its scope MCMs having more or less than the eight dice shown in FIG. 3, and having more or less than
the two repair sites shown in FIG. 3.
Once the repair die 18 is attached, the MCM 10 moves to a wire-bonding station. There, a wire-bonding machine reads the unique ID code of the MCM 10, accesses the repair solution stored in the computer system 22 in association with this ID code,
and wire bonds the repair die 18 accordingly. As a result, the opportunity for human error in selecting the appropriate repair solution is minimized or eliminated.
As shown in FIG. 4, the die 12 is wire-bonded to bond pads 30 on the surface of the PCB 16, while the repair die 18 is wire-bonded to one of multiple sets of bond pads and conductive traces 32 on the same surface.
Although the present invention has been described with reference to a particular embodiment, the invention is not limited to this embodiment. For example, while the various steps of this embodiment have been described as occurring in a
particular order, it will be understood that these steps need not necessarily occur in the described order to fall within the scope of the present invention. Thus, the invention is limited only by the appended claims, which include within their scope
all equivalent methods that operate according to the principles of the invention as described.
* * * * *
|
|
|
|
|
Description  |
|