Knowledge Center
Knowledge Center

Double Patterning Methodologies

Colored and colorless flows for double patterning


The way that a designer will interact with the constraints that double patterning (DP) brings is very dependent on the design methodology used. One key DP methodology decision is if the designers will see colors at all. This is called a “colorless” design flow. The alternative is a two-color flow, in which the designer tapes out two masks, choosing one of several decomposition options. There are trade-offs with any design flow, so ultimately you must decide based on what is best for your organization and which flows your foundry will support.
The diagram illustrates four possible DP design flows. The first three flows are two-color flows, in which the designer tapes out two separate “color” layers for any layer that will be double-patterned. The fourth flow is the “colorless” flow, in which the designer only tapes out a single layer, and the foundry performs the decomposition into two layers. The difference between the first three flows is the level of automation involved in creating the two layers.

In a manual decomposition flow, the designer draws both mask layers by hand. The accuracy of the decomposition is verified with traditional design rule checking (DRC) tools as spacing constraints between masks. The designer manually decides how to implement the design in two masks. Decomposing layouts by hand can be very time-consuming and difficult, and on a large block, impractical. For this reason, the other flows are more common.

In an automated decomposition flow, the designer draws a single layer. An EDA tool then automatically decomposes the layer into two masks, which are checked with traditional DRC operations. Specialized odd cycle DP checks can also be run, to help the designer better understand the complete set of polygon interactions causing any decomposition errors.

A mixed decomposition flow is a hybrid of the first two. The designer can perform some decomposition manually, and then have an EDA tool automatically decompose and check the remainder of the design. This flow is commonly used when the designer is incorporating cells that were previously decomposed into a new layout that is still in one layer.

In a colorless flow, the designer never generates the two mask layers, relying instead on the specialized odd cycle checking within an EDA tool to determine locations that cannot be correctly decomposed. The single layer must be modified until these checks are clean. The foundry then decomposes the layout into two layers after tapeout.

Although a colorless flow does not require the designer to decompose the layer, there may be special conditions in the layout for which the designer wants to control polygon coloring. For instance, it may be important to place power and ground rails on a specific mask, or, in an analog circuit, “matched” devices may need to be decomposed identically, to ensure their performance will match. In these cases, “anchor” markers are used to communicate these special requests to the decomposition tool used by the foundry. These anchors are typically designated layer markers placed over the “fabric” polygons to assign them to certain masks. The EDA tool will then decompose the unassigned polygons influenced by the specified polygons.

The diagram below shows how the drawn design and decomposed layout would look for these flows. In the colorless option, the drawn layout contains only a single layer of polygons. After decomposition by the foundry, two layers are generated for this layout. In the anchors example, although there is still only one drawn layer, there are anchor markers forcing the power and ground rails to be on one mask, and constraining one of the inner shapes to the other mask. As you can see in the decomposition, the polygon colors (mask assignments) do not match the first example because the tool honored the anchor markers.

Original contents provided by Mentor Graphics