Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Delay Configurable Metal Only Cells for Hold Fixing with Flexi Structural Design - A Novel Approach to Resolve Hold Violations During Metal ECOs

IP.com Disclosure Number: IPCOM000234559D
Publication Date: 2014-Jan-17
Document File: 5 page(s) / 429K

Publishing Venue

The IP.com Prior Art Database

Abstract

In the era of lower technology nodes, more pessimism is required for timing closure due to greater variation across corners. In a typical scenario, the number of hold buffers required for timing signoff may be of the order of 100 thousand. The buffers added to fix hold are small having less drive strength. These buffers are more prone to cross corner variations. Also, this contributes significantly to leakage power. So, there arises a need for a hold fixing solution that contributes less to leakage and provides an area efficient solution for hold fixing. In this paper, we present a solution to fix hold timing violations with the help of dangling nets. We discuss the pros and cons and what advantage it provides in terms of area, power and noise immunity to the design. Also, in the end, we discuss the algorithm for automatic metal-only cell creation to fix hold violations.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Delay Configurable Metal Only Cells for Hold Fixing with Flexi Structural Design - A Novel Approach to Resolve Hold Violations During Metal ECOs

In the era of lower technology nodes, more pessimism is required for timing closure due to greater variation across corners. In a typical scenario, the number of hold buffers required for timing signoff may be of the order of 100 thousand. The buffers added to fix hold are small having less drive strength. These buffers are more prone to cross corner variations. Also, this contributes significantly to leakage power. So, there arises a need for a hold fixing solution that contributes less to leakage and provides an area efficient solution for hold fixing. In this paper, we present a solution to fix hold timing violations with the help of dangling nets. We discuss the pros and cons and what advantage it provides in terms of area, power and noise immunity to the design. Also, in the end, we discuss the algorithm for automatic metal-only cell creation to fix hold violations.

Proposition: We propose the use of “delay configurable metal-only cells” to fix hold timing violations. These metal-only structures have no active element involvement so they can be put anywhere along a net dangling. With this, we achieve a kind of dummy load for an existing cell without any slew degradation for the load of the existing net.

 What problem does it solve: By this proposition, we improve hold timing bearing in mind the disadvantages of existing techniques and methodologies. Our solution is:

      1. Less susceptible for cross corner variation.
      2. Less leaky compared to existing solution
      3. Enable Metal only closure for 1.X Tapeout

Thus, it is superior in comparison to the existing methodologies for hold fixing.

Existing methodologies: The most common way to fix hold timing violations is by inserting delays in data path with the help of buffers. Figure 1 shows a typical such scenario. The insertion of buffer in data path will introduce delay and provide more susceptibility to hold failure. The insertion of buffer adds both to area, leakage power and dynamic power. Also, during metal-only ECOs, this method can only be adopted if there are spare cells available in the vicinity of the violating path. There are a number of buffers inserted in the designs just to provide robustness from hold violations perspective. According to an observation, as many as 100 thousand buffers may be required to provide robustness to designs for hold failures. This much is the penalty to be paid in terms of leakage power, dynamic power and area. As technologies are being scaled down to lower nodes, variation across corners requires more data path delays to be inserted for robustness. The main reason behind such large number of hold buffer insertions is that cell delay/net delay variation is very large across different net lengths. So, there arises need for a solution that is robust in terms of area, power and variation across corners.

Figure 1: Fixing ho...