1. Field of the Invention
The present invention is related to Integrated Circuit (IC) chip fabrication and more particularly to optimizing IC Input/Output (I/O) cells for improved chip manufacturability.
2. Background Description
A typical integrated circuit (IC) chip includes a stack of several sequentially formed layers of shapes, also known as mask levels. Each layer may be created or printed optically through well known photolithographic masking, developing and level definition, e.g., etching, implanting, deposition and etc. Shapes stacked on or overlaid on shapes on a prior layer define devices (e.g., field effect transistors (FETs)) and connect the devices into circuits. In a typical state of the art complementary insulated gate FET process, such as what is normally referred to as CMOS, device layers are formed on a surface layer of a wafer, e.g., a silicon surface layer of a Silicon On Insulator (SOI) wafer. Islands are defined by removing open or unpopulated areas of the silicon surface layer, for example, using Shallow Trench Isolation (STI). A simple FET is formed by the intersection of two shapes, a gate layer rectangle on a silicon island formed from the silicon surface layer.
A typical gate array, for example, includes a number of identical groups of devices or FETs, in what are known as cells. Logic cells are typically, centrally located in one or more cell arrays. Devices in each cell may be wired together in a simple logic block. Cells may be wired together into more complex logic function. Some larger groups of devices may be clustered together as macros. Ideally, fabrication parameters applied to features on a particular layer to affect all features uniformly on that layer, such that devices form uniformly. Unfortunately, all features do not respond uniformly.
Typically, each gate array has a number of fixed Input/Output (I/O) cells that are independent units with a predefined shape, form and function. In particular, a typical gate array chip footprint has several locations set aside for I/O cells with a fixed space (“one size fits all”) reserved. So, the space reserved for each I/O cell is determined by the predefined I/O cell shape for the largest I/O circuit in the gate array library. Typical I/O cells may have some densely populated levels, while other levels have large areas with nothing. Further, some I/O cells may have simple functions that may be implemented in much less area than others.
Locating a simple (smaller) I/O circuit in a larger I/O cell guarantees unused space with unpopulated or open areas in that I/O location. These large unpopulated or open areas are typically referred to as white areas. So, in a typical state of the art I/O cell the silicon layer is sparsely populated with isolated silicon island shapes surrounded by white space. Consequently, tuning shape formation for denser areas, e.g., in arrays, can cause these isolated shapes to distort, e.g., the shapes wash out. I/O Devices (FETs) formed from these washed out shapes have characteristics that do not match other chip devices and, typically, do not conform to design specifications.
Other levels may include isolated shapes in white areas as well, e.g., deep trenches (relatively narrow trenches that extend well into a silicon substrate below the SOI insulator layer) in the I/O areas. Deep trenches may be included in an I/O cell, for example, for capacitors, guard rings and/or electrostatic discharge (ESD) protect devices. Similarly, if an I/O cell does not include structures with deep trenches, placing the I/O cell adjacent to a memory array with deep trench storage capacitors, guarantees that trenches (at the edge of the array) have white space on at least one side. Because of this white space, these isolated deep trenches can fail to open or at least fail to open sufficiently to fill, e.g., with plate material for a deep trench capacitor. Further, other shape formation parameters, e.g., focus, focus angle and photoresist thickness uniformity may cause feature variations across the chip and wafer, i.e., Across Chip Linewidth Variation (ACLV). White spaces may exacerbate these variations in some locations and minimize them in others, further degrading ACLV. These unintended changes to I/O cell shapes may degrade the chip and
Thus, there is a need for white space compensation in gate array I/O cells.
It is therefore a purpose of the invention to improve Integrated Circuit (IC) chip manufacturability;
It is another purpose of the invention to reduce white space effects in Input/Output cells;
It is another purpose of the invention to reduce white space effects in gate array chips, especially in gate array chip Input/Output cells.
The present invention is related to a method of fabricating an integrated circuit (IC) chip. A standard cell macro (e.g., an Off Chip Interface (OCI) cell) is defined with circuit elements identified as in a macro domain. A variable macro boundary is defined for the standard cell macro. Shapes are selectively added to design layers in the macro boundary to occupy existing white space. Each supplemented layer is checked for technology rules violations in the macro boundary. Each layer is also checked for known sensitivities in the macro boundary.
The foregoing and other objects, aspects and advantages will be better understood from the following detailed description of a preferred embodiment of the invention with reference to the drawings, in which:
Turning now to the drawings, and more particularly,
As shown in the example of
The I/O circuit 104 connects off chip through one or more pad 106, e.g., for solder balls or wire bonds. The protect diode or electrostatic discharge (ESD) protect device 108 protects the chip 120 and, especially, the I/O circuit 104 from static discharges that might otherwise permanently damage the chip 120. Capacitors 110 provide local supply decoupling, improve ESD protect device 108 protection and may be provided for inclusion in the I/O circuit 104. Normally, the guard ring 112 surrounds the I/O circuit 104 and other OCI structures 106, 108 and 110. The guard ring 112 isolates the OCI 100 itself from other circuits and the rest of the chip structures from the OCI 100, e.g., from parasitic charge/current from electrostatic discharges that might be coupled into the OCI 100 from off chip. The OCI domain (OCID) includes any element necessary for forming the OCI 100, in this example the I/O circuit 104, chip pad(s) 106, the ESD protect device 108, capacitors 110 and OCI guard ring 112. It is understood that although represented in this example as a single off-chip circuit being included in each OCI cell 100, this is for example only. Each OCI cell 100 may include multiple off-chip circuits, each including an I/O circuit 104 and other OCI structures 106, 108 and 110, 112 and all encompassed by the OCIB 102.
Also, according to a preferred embodiment of the present invention, each OCI cell 100 may include densification or background optimizer (OCIO) shapes (e.g., 114) on one or more or all levels within its OCIB 102. These OCIO shapes 114 occupy white space on each layer in the OCIB 102. Typically, the densification shapes on each layer are between shapes in OCI 100 structures including the I/O circuit 104, the chip pad(s) 106, the ESD protect device 108, capacitors 110 and the guard ring 112.
Normally, other than the guard ring 112, each of the OCI 100 structures including the I/O circuit 104, the chip pad(s) 106, the ESD protect device 108, capacitors 110 do not have shapes on every chip layer. Instead, each OCI structure 104, 106, 108, 110 usually has one or more shapes occupying space on only a few levels. For a CMOS IC chip 100, for example, the I/O circuit 104 and the ESD protect device 108 may have shapes confined, primarily, to lower levels (e.g., to the lowest wiring level) with upper levels relatively free. Pads 106 and, possibly, capacitors 110 may be primarily in upper levels, i.e., above the lowest wiring level.
Thus, it is likely that at each level includes some white space within the OCIB 102, white space that is at least partially occupied by OCIO shapes 114. How that white space would otherwise normally affect each particular OCI cell 100 on any particular layer depends upon relatively close topological features. These close topological features may not be part of the particular OCI cell 100 itself, but in adjacent logic/arrays 122. As noted hereinabove, on the lowest (silicon) levels, white space may washout shapes that are isolated by Shallow Trench Isolation (STI) in otherwise isolated devices in the OCIB 102. However, with OCIO shapes 114 in the OCIB 102, those shapes are not so isolated. Similarly, because of white space in prior OCI cells, deep trenches (e.g., for forming Dynamic Random Access Memory (DRAM) storage capacitors) may not open at the edge of a DRAM array bordered by the OCI cells. Similarly, with OCIO shapes 114 (i.e., densification trench shapes) in the OCIB 102, those edge trenches are no longer edge trenches in a preferred chip.
Without the OCIO shapes 114, even if deep trenches form or devices do not wash out, shape variations may be such as to exceed Across Chip Line Variation (ACLV) targets. Excessive ACLV causes yield loss with circuits frequently failing in out of tolerance sites. Previous steps taken to counteract ACLV in logic/arrays 122 have degraded ESD protection and increased OCI 100 susceptibility to latch-up.
So, according to a preferred embodiment of the present invention, the aspect ratio is not fixed for all OCI cells (i.e., the OCIB may be different for each OCI cell) and each OCI cell 100 may include OCIO shapes 114 within its OCIB 102 on different layers. Thus, the present invention has application to customizing the physical layout of placed OCI cells 100 to self-compensate for effects of local geography. In particular, the OCIO shapes 114 self-compensate for local STI density as well as ACLV and STI topography issues, white space optimization, ESD, and latch-up. So, while each placed OCI cell 100 may have a different physical structure within its OCIB 102, all of the placed OCI cells 100 have substantially the same electrical characteristic.
In step 204, the initial OCI cell 100 is optimized for placement in standard chip floor plan and for routing, e.g., using a typical place and route design flow step. The optimization is based on appropriate circuit guidelines and technology specific files 206, e.g., process ground rules, Design For Manufacturability (DFM) guidelines and circuit timing specifications. Typically, this optimization 204 identifies rule violations and removes shapes to address those violations.
After optimizing for placement in step 204, design manufacturability analysis is applied to the cell design in step 208, e.g., in a layer by layer analysis. The design manufacturability analysis checks shapes on a particular layer for known design sensitivities for that layer. For the surface silicon layer (RX) in a Silicon On Oxide (SOI) wafer, for example, RX/STI is checked for white areas that are known to cause problems with both shallow trench isolation and with deep trench formation. So appropriate rules are provided in step 210, for example, for checking within the OCIB 102 the surface silicon layer (RX/STI), e.g., for white areas, for densification, for ACLV violations, for potential sources of ESD/latchup violations and guidelines. Shapes are added or removed to the OCIB 102 in any layer to address identified design sensitivities. Where shapes are added in checking step 208, preferably gate array cells are added to offset any identified problems, e.g., filling white areas. Optionally, if the added shapes are gate array cells, those added cells may be used for remaining logic cell placement.
In step 212 placement and wiring information is added (e.g., symbols identifying open channels that pass through the cell and connection points to cell I/O) to the modified OCI cell 100 with the OCIO shapes 114 and expanded OCIB 102. In step 214 the OCI cell 100 design is checked whether more layers remain for optimization mode and, if any remain, returning to step 202, the expanded OCIB 102 is taken as the boundary for checking the next layer of the modified OCI cell 100. Once all layers have been considered in step 214 or at least all layers of interest, then in step 216 modified OCI cell 100 has all features contained within the OCIB 102 and meets design specifications.
Once the OCIB 102 is customized to meet design specifications, customized OCIs 100 can be placed in a chip design, e.g., chip 120 in
As noted hereinabove,
Similarly,
Continuing this example, OCIB 102 definition begins with in step 222 defining a “super I/O” cell around the initial OCI cell 100. In step 224 the initial OCI cells 100 are placed in the standard chip floor plan (e.g., clustered in peripheral off chip banks 124, 126 or surrounded by logic in logic/array 128) and wiring is routed based on the placement. Place and wiring step 224 is guided by appropriate wafer level guidelines and technology specific files, typically similar to chip level guidelines and technology specific files 206. After optimization for placement in step 224, wafer design manufacturability analysis is applied in step 226, checking chip shapes on all affected layers for known design sensitivities 210. In step 228 placement and wiring information is added to the wafer design, which is made available for review in step 252.
Again, it should be noted that cell customization may be done serially or in a flat approach as describe for
Advantageously, a boundary layer (OCIB) is defined for customizing each I/O cell in both cell definition (size and aspect ratio) and content, with preferred embodiment I/O cells having white space in the boundary layer filled with prototype gate array shapes. Further, the boundary layer width may be adjusted based on density analysis both inside and outside of the boundary layer region. The boundary layer may be compressed (changing the aspect ratio of a particular OCI) to optimize shapes contained in the boundary layer to address process and device issues. Moreover, each OCI may have multiple individual virtual boundary layers (OCIBs), each defined based on physical phenomenon expected to be encountered for the particular design layer.
While the invention has been described in terms of preferred embodiments, those skilled in the art will recognize that the invention can be practiced with modification within the spirit and scope of the appended claims. It is intended that all such variations and modifications fall within the scope of the appended claims. Examples and drawings are, accordingly, to be regarded as illustrative rather than restrictive.