A recent report from IC Insights described 72 wafer fabs that have closed in the past five years. Eight more plants have gone in 2014, showing the trend is continuing.
This leaves their customers with a problem: what can they do when the fab shuts down? Some may recognise that their own technology has reached the end of its life and work on generating a replacement while others will place a ‘last time buy’ to stockpile as many chips as they can. However, most will be left with a headache as they go looking for a new supplier.
Moving any existing circuit from one foundry to another is difficult and it’s much worse when dealing with legacy chips. Detailed databases may be difficult to locate and the original design team is likely to be long gone, leaving little circuit knowledge for the product. Companies may be left with a few files and a pressing need to find a new supply of silicon.
As process migration specialists, IN2FAB often works with companies who have to find a new foundry. A redesign will be time consuming and too costly when no new functionality is required so a migration based path to a new supplier is very attractive.
When facing this problem, the chip’s owner should gather as many design files as possible and a GDSII file and a netlist is usually the minimum. A schematic database is preferred although the tools used to create them may be gone so translators can be used to pull them in to a Cadence system to match the new foundry’s PDK. A layout database made with parameterised cells is useful but polygon based layout will suffice.
Analog and mixed signal circuits are typically defined by the transistors’ voltage thresholds and matching of the passives. If the ohms per square in the target process are much lower, resistors will have to grow which can lead to spacing problems and a similar match must be made for capacitors. IN2FAB usually conducts a detailed feasibility study to identify similarities and differences between components which is essential when choosing the new foundry and process.
Delay files for digital circuits may not be to hand which makes regeneration through place and route almost impossible. Instead, the design can be migrated as a custom circuit to exactly match the original and maintain the placement and routing as before. This retains balance and prevents the introduction of new timing offsets. Digital circuits can often move to a smaller node and gate sizes and routing adjusted to meet the new rules without losing the integrity of the original circuit.
Other elements like bond pads and ESD or difficult components like inductors must also be considered but the key to the migration is to match the design to the new process and use automation to modify the design as needed. While some dedicated engineering input may be needed to address fine details, the automation in our own EDA technology means that the chip can usually be transferred to the new process in weeks.
Losing a foundry is a problem but it needn’t be a disaster. Migration technology can move circuits to a new process from basic design files or old CAD systems and bring them up to date, giving them a new lease of life. Redesign is expensive and putting designers to work on old products is poor allocation of resources. Migration presents an effective alternative and is usually the fastest way to move circuits to a new foundry.
President and CTO