Military FPGA Die Banking for Long Duration Programs
Table of Contents
- Obsolescence Risks in Long Duration Military FPGA Programs
- What Is FPGA Die Banking and How It Works
- Military-Specific Requirements for Die Banking
- Building a Die Banking Strategy with a Trusted Distributor
- Securing Your Military FPGA Supply Through a Die Banking Partnership
- Common Questions About Military FPGA Die Banking
- What is the minimum quantity of dies required for a die banking agreement?
- How long can FPGA dies be stored without reliability loss?
- Does die banking work for radiation-hardened FPGAs?
- Who owns the die — the distributor or the program?
- How does die banking interact with configuration memory and other companion devices?
When a defense program spans two or three decades, the FPGA devices designed into it at the start are almost guaranteed to go obsolete long before the last system is delivered. Military FPGA die banking is a proactive strategy that can prevent this obsolescence from turning into a production crisis. In my twelve years managing military component supply chains, I have seen die banking secure the long-term availability of critical FPGAs when no other source remained. Yet many program teams misunderstand the process or underestimate the military-specific requirements that make die banking more than a simple purchase of surplus die. This article explains how defense programs can implement effective die banking strategies that align with MIL-SPEC traceability, trusted supply chains, and the practical realities of long-duration program support.
Obsolescence Risks in Long Duration Military FPGA Programs
FPGA manufacturers regularly discontinue older process nodes as they migrate to newer technologies. For commercial applications, a last-time buy and a five to seven-year support window is often sufficient. Military programs, however, may need 20 to 30 years of active production. When Xilinx ended fabrication of the Virtex-5 family, programs that had designed it in faced a hard stop unless they had already secured a die bank. Rad-hard and rad-tolerant devices, produced in low volumes, are even more vulnerable because the original wafer fabs may not maintain the necessary process certifications indefinitely. I have worked with contractors who discovered that their sole FPGA source was unavailable just 18 months after the last-time buy window closed, because the die lot had been exhausted. The only remaining option was a costly and schedule-disrupting redesign. These situations make die banking a logical early investment, not a late-stage contingency.
What Is FPGA Die Banking and How It Works
Die banking involves a contractual arrangement with an authorized semiconductor distributor or supplier to purchase a specific quantity of known-good die from the original wafer lot, store them under controlled conditions, and then make them available for packaging and testing over the life of the program. This differs from wafer banking, where entire wafers are stored, or last-time buys, which deliver finished packaged parts that may have limited shelf life.
| Strategy | How It Works | Best For |
|---|---|---|
| Die banking | Purchase bare die from fab; store; package/test on demand | Long-duration programs needing gradual consumption |
| Last-time buy | Buy finished parts before EOL; inventory management needed | Short-to-medium term after discontinuation |
| Wafer banking | Store entire wafers; dice and package later | Programs with very high volume over many years |
The advantage of die banking is flexibility: dies can be packaged in the required configuration — commercial, industrial, or military temperature range — and tested to the applicable standard (MIL-STD-883, QML) when the need arises. This avoids the degradation that can affect pre-packaged parts stored for extended periods. A typical die bank arrangement specifies the die quantity, storage duration (often 10 to 15 years), packaging capacity, and price structure. The dies themselves are stored in nitrogen-purged, temperature-controlled environments to meet JEDEC storage standards, ensuring long-term reliability.
Military-Specific Requirements for Die Banking
Die banking for a commercial product is straightforward compared to the compliance-driven world of military electronics. MIL-PRF-38535, which defines Qualified Manufacturer List devices, requires full traceability from the die back to the wafer lot, mask set, and fabrication facility. Any die banking arrangement must preserve this chain of custody without gaps. This means the die bank provider must maintain auditable records of storage conditions, lot identity, and handling history.
Additionally, many military FPGAs are produced under the DMEA Trusted Foundry program, which restricts the facilities permitted to fabricate and handle the die. A distributor involved in die banking must either be an accredited supplier or work through an accredited facility for all storage and processing steps. I have supported programs where the die bank was nearly rejected during a government audit because the storage logs were missing humidity data for a two-month period. Even minor documentation gaps can jeopardize the entire die bank investment.
Another consideration is configuration memory. FPGAs are SRAM-based and require an external configuration source, often a PROM. Die banking for the FPGA alone does not solve the problem if the companion configuration PROMs are also obsolete. A complete die banking strategy should include the full signal chain, not just the FPGA.
If your program involves FPGAs from a DMEA trusted foundry or requires QML-V certification, it is worth confirming that your die banking partner has the accredited storage and processing capability before committing to an agreement. Send your program details and FPGA part numbers to [email protected], and I can help you assess the die banking feasibility for your specific devices.
Building a Die Banking Strategy with a Trusted Distributor
A successful die banking strategy starts with identifying which FPGAs are truly critical and likely to become obsolete within the program’s support horizon. Not every component on the BOM needs a die bank; focus on those with long lead times, limited second sources, or unique qualification status. For military FPGAs from Xilinx, Microchip, or other major suppliers, the production timeline and planned technology node migration should be assessed through the authorized distribution channel.
Once the target FPGAs are identified, the program team should forecast the total number of dies needed over the remaining production and sustainment phase, including spares. With that forecast, a request for proposal can be issued to qualified distributors. I recommend looking for distributors with experience in military die banking, ideally holding AS9120 certification or equivalent quality management credentials. At Sparkle Electronics, we have structured die bank agreements for several long-duration defense programs, managing everything from wafer procurement through final packaging and test. The agreement typically includes regular lot testing to verify die integrity, staggered packaging to align with production batches, and full documentation to support government audits.
Securing Your Military FPGA Supply Through a Die Banking Partnership
Die banking represents a substantial upfront investment, but the cost of not securing critical FPGA supply is almost always higher. I have seen programs where a single FPGA redesign took 18 months and cost over two million dollars in engineering and requalification, not counting the schedule impact. A die bank negotiated early in the program lifecycle prevents that disruption.
The key to success is working with a distributor that understands the military supply chain, can provide the required traceability and storage documentation, and has the flexibility to package devices on a schedule that matches your production ramp. If your program is entering a long-term sustainment phase or you have just received a last-time buy notice for a critical FPGA, reach out at [email protected] with your BOM and timeline. I will help you evaluate whether die banking is the right solution and outline a path that protects your program’s supply chain for the long haul.
Common Questions About Military FPGA Die Banking
What is the minimum quantity of dies required for a die banking agreement?
Most die banking agreements require a minimum purchase of several thousand die, depending on the wafer size and die area. For low-volume military programs, this can be a challenge, but working through a distributor that aggregates multiple program commitments can sometimes reduce the minimum. In my experience, even two thousand to five thousand die can be negotiated if the program is a long-term commitment. Share your specific die quantity requirement and I can check current program options.
How long can FPGA dies be stored without reliability loss?
With proper storage in nitrogen-purged, temperature-controlled environments following JEDEC guidelines, bare dies can be stored for 10 to 15 years with minimal degradation. The key is periodic inspection and, when possible, wire bonding of sample dice to confirm that the metallization and passivation layers remain intact. Major die banks I have managed include a 12-year storage plan with annual lot testing, and the die continue to meet original specifications.
Does die banking work for radiation-hardened FPGAs?
Yes, but radiation-hardened devices often have specific process flows that require the original fab to maintain process certifications. If the original fab is decommissioned, rad-hard die banking becomes impossible because alternative fabs cannot replicate the process. Therefore, for rad-hard FPGAs, die banking should be initiated while the original process is still active. I always recommend starting the die bank discussion at least two years before the expected foundry process end-of-life.
Who owns the die — the distributor or the program?
Ownership is defined in the agreement. Typically, the program purchases the die and the distributor stores and manages them on a fee basis. This ensures that the die remain an asset of the program and are not sold to another customer. Make sure the contract specifies ownership, storage obligations, and liability for loss or damage.
How does die banking interact with configuration memory and other companion devices?
Die banking for the FPGA alone is insufficient if the configuration PROM or other supporting devices are also obsolete. I have seen programs where a die-banked FPGA was available but no compatible PROM existed. The solution is to plan die banking for the entire configuration chain, which may include antifuse or flash-based configuration devices. If your program’s FPGA uses a custom configuration scheme, it is worth confirming the availability of all companion devices before finalizing the die bank agreement. Send your complete BOM to [email protected] and we can assess the full die banking requirement for your program.
If you’re interested, check out these related articles:
A1020B-PG84B ACT2 FPGA: Specs, Sourcing, and Availability
XCKU115 UltraScale FPGA: Powering Critical Defense Systems
