Table of Contents
Programming and verification can pass while the board still fails its first unattended boot. The programming tool and the FPGA’s configuration engine can interact with the same memory through different command sequences and startup conditions.
For 7 Series maintenance, preserve the boot contract between the FPGA, flash and board. Density and package matching are only the opening checks. The FPGA sourcing guide handles the programmable device’s purchasing identity; configuration memory needs its own compatibility record.
What belongs in the original boot configuration record?
Record the exact FPGA and flash, configuration mode, rail voltage, clock settings, image layout and update/recovery process. The replacement must support the operations used by both autonomous startup and field updating.
AMD UG470, the 7 Series configuration guide, describes the configuration modes and startup sequence. Use it with the proposed flash datasheet and the programming-tool support documentation for the version actually deployed. A generic statement that a memory is “SPI compatible” does not establish this combination.
Capture mode straps and pull resistors from the board, not just a schematic that may predate production changes. Preserve the released bitstream properties, including bus width and configuration clock choices. Record whether an external processor changes the flash state before the FPGA accesses it.
Which flash differences can break an otherwise familiar boot path?
Differences in startup state, addressing, commands and reset behavior can prevent boot even when the device has sufficient capacity and a compatible footprint. Compare the actual transactions, including the operations used after a failed update.
| Compatibility field | Evidence to compare | Failure to exercise |
|---|---|---|
| Supply and thresholds | Flash operating range and FPGA bank/configuration requirements | Slow ramp or marginal startup voltage |
| Read command and bus width | Required opcode, dummy cycles and output mode | First read after reset differs from a programmed state |
| Addressing | Image addresses and three-/four-byte address behavior where relevant | Higher-address image cannot be reached |
| Erase/program geometry | Sector boundaries, page handling and write-enable behavior | Update damages adjacent or fallback content |
| Protection and status registers | Reset values and update software assumptions | Field updater cannot write, or unlocks too much |
| Reset/power-down recovery | Device timing and board control connections | Warm restart works but cold boot fails |
The general SPI NOR selection guide explains the broader memory-screening task. The 7 Series decision adds the FPGA configuration engine and released image properties to that screen.
How much configuration storage is actually required?
Size the flash from the generated configuration files and reserved memory map, including recovery images, headers and erase alignment. Do not estimate from the FPGA’s logic count or assume that image compression will always achieve the same ratio.
For a layout with a 12 MiB primary image, a 12 MiB fallback image and 1 MiB of metadata, the minimum is 25 MiB before any additional alignment or reserved area. A nominal 256 Mbit device contains 32 MiB, but that arithmetic alone does not prove address-mode or erase-layout compatibility.
Retargeting an XC7A100T design to XC7A200T should reopen this calculation. Keep the generated image size and its actual placement in the release record, so a later design build cannot silently exceed the assumed boundary.
Test the recovery path before ordering production quantities
Build a trial around states, not a single successful programming event. Start with blank material, run the production programmer, verify the intended regions, then remove power completely and boot using the installed board’s normal sequence.
Repeat with the supported power-ramp extremes, warm reset, field update and recovery-image selection. Where the system claims to tolerate interrupted updates, test controlled interruptions at the stages defined by the update design. Record which image starts, how failures are reported and what action returns the unit to service.
Separate flash electrical qualification from software approval. A changed JEDEC identification response may require an updater change; that change must be distributed to the installed fleet before the new flash reaches it. If the old updater cannot address the replacement safely, purchasing it as a service spare creates a deployment problem even when new production can use it.
Release the new flash ordering code with the validated FPGA, board and software revisions. The Artix-7 series directory can identify the device target, but it cannot supply a missing bitstream or authorize a flash substitution.
Frequently Asked Questions (FAQ)
Does successful JTAG configuration prove that the SPI flash is compatible?
No. JTAG can bypass the autonomous flash boot path. Validate boot from the replacement flash after a complete power removal and with the released mode straps.
Can unused flash capacity hold unrelated application data?
Only if the memory map, erase boundaries, configuration image and update process explicitly reserve that space. An update must not erase or overwrite a recovery image.
Should the new flash be added as an FPGA alternate in the BOM?
No. It is a separate component change with its own ordering code and validation record. Link the approved flash revision to the FPGA configuration and board revision it supports.