Troubleshooting
Fixing an HMI Stuck on Invalid Database at Boot
Short answer
Send Time on the Link menu starts a Crimson 3.0 panel that boots to INVALID DATABASE long enough to extract the project. Install the build the splash screen shows (647 for a 647.002 panel), extract the .cd3 in that live window, and reload it. If extract reports no uploadable image, Support Upload was off and only the OEM or a backup can supply the file; finish by updating the panel to Crimson 7.11.
On this page
- What does INVALID DATABASE at boot mean?
- Why did my HMI lose its database after a repair?
- Which Crimson build do I need to connect to the panel?
- Why does Send Time revive the panel until the next reboot?
- How do I extract the project while the panel is running?
- Why does Crimson say the device does not contain an uploadable image?
- How do I load a project back onto the panel?
- When is replacement the only option?
- Questions asked
An HMI that halts at INVALID DATABASE has lost the project it needs to run, not its hardware. Send Time on the Link menu starts the runtime, and the window it opens is your one chance to pull the project out. Install the Crimson build the splash screen shows — 647 for a panel reporting Crimson 3.0, build 647.002 — connect, send the time, and run the extract while the screens are up. If extract answers "The device does not contain an uploadable image," the project was downloaded with Support Upload disabled and no copy exists on the panel, so the OEM or another backup is the only source. Load what you recover, finish on Crimson 7.11, and the fault cannot return. Work in that order, because every step after Send Time depends on the window it opens.
What does INVALID DATABASE at boot mean? #
INVALID DATABASE is the Crimson runtime failing to find or verify a valid project database in the panel's flash, so boot stops at the splash screen with no operator screens, no touch response, and no link to the controller.
The runtime mounts the project database from internal flash at every cold start and validates it before any operator screen loads; a missing, wiped, or unverifiable database stops boot at the splash line. The splash line is also your best diagnostic, because it names the runtime and the build — photograph it before touching anything, since that string decides which tool you install next. The field case is a Red Lion G308A terminal fronting a gas-detection system, and the behavior is shared across its G3 series.
Rank the likely causes before swapping parts:
- Flash cleared during repair — a shop rebuild that re-initializes memory takes the project with it, and the database does not come back with the repaired power supply.
- The original download had the upload option cleared, which leaves nothing on the unit for a PC to read back.
- A timestamp the runtime rejects after a power interruption or a date-boundary crossing — the fault the Send Time behavior exposes.
- Card trouble in the CF slot: log files on the card are not the project, and repairing the card's directory structure recovers nothing.
The first two decide what recovery is even possible; the third decides the order of your attempts.
Why did my HMI lose its database after a repair? #
A repair that opens the unit and clears or re-initializes its flash takes the project with it, and the CR2025 clock cell is not the cause: in the field case the old 3 V cell still measured good and a fresh one changed nothing.
A unit that goes out for repair — capacitors replaced in the power-supply section, in this case — comes back with a blank application database when the shop re-initializes flash or the boot block, which is the normal outcome when no backup preceded the shipment. The coin cell is the tempting suspect and the wrong one: the CR2025 is a 3 V lithium cell that keeps the real-time clock alive, not the project storage, and a cell that still reads 3 V is doing its job. Swap it as routine service on hardware of this age, but expect nothing from it against this error. With a good cell and a repaired power supply, empty project storage is the only explanation left standing.
Which Crimson build do I need to connect to the panel? #
Match the tool to the splash-screen build: a panel running Crimson 3.0 build 647.002 wants Crimson 3.0 build 647, because an older build can refuse the link and a mismatched extract can trip a restart that closes your window.
Read the build string 647.002 as Crimson 3.0, build 647, revision 002: the tool that matches is the 647 build, not an older 3.0 build, not Crimson 2.0, and not 7.11 yet, because only the matching build shares the database signature the panel validates. Uninstall newer Crimson releases before installing the legacy build — the installers do not sit cleanly side by side, and a mixed install has gone sideways in the field. Older builds are no longer posted for open download, so request the exact build from vendor support with the unit's serial number in hand; the splash string is the specification you read to them. Confirm the link before attempting anything: the tool should report the panel's model and firmware revision before you send a single command. Tracking down that installer is the same discipline as getting the right firmware file for a legacy HMI panel.
Why does Send Time revive the panel until the next reboot? #
The Send Time command on the Link menu writes the PC's clock to the panel, and the runtime accepts the fresh timestamp, remounts the flash, and comes up to the operator screens, but nothing is rewritten, so the next cold start fails the same way.
Boot-time database validation runs against the panel clock, and a timestamp the check rejects fails the mount; a fresh write from the PC passes it. That is the signature of the date-handling fault carried in older Crimson 3.0 builds: the project itself is intact, the check guarding it is wrong, and the panel looks dead while it is actually holding a runnable application. It also fits the battery evidence — a good cell and a working low-battery circuit rule the clock hardware out, which leaves the timestamp check as the fault.
Send Time is your door, not your fix. Nothing is rewritten to flash, so the next power cycle repeats the same failed validation. The moment the screens return, the window is open, and what you do in it decides whether the project survives the next reboot.
How do I extract the project while the panel is running? #
Send Time first, then run the extract command in the same session over USB or Ethernet, because the live window closes at the next power cycle and only a matching build keeps the extract from triggering a restart.
Sequence matters more than speed, because one wrong move closes the window:
- Connect the PC running the matching build and confirm the tool reports the panel and its firmware revision before anything else.
- Choose Send Time on the Link menu and wait for the operator screens to return.
- Run the extract in that same session, over USB or Ethernet, and save the result as a .cd3 file named for the machine.
- Power-cycle the panel deliberately once the file is on the PC. If it boots to the operator screens again, the database is durable and the job is done; if it falls back to INVALID DATABASE, you still hold the file, and the reload is covered below.
Treat the panel as live controls the moment the screens return: operators regain a start and reset surface that has been dead, so warn the floor before you send the time.
If the touchscreen itself is also failing, backing up an HMI with a scratched touchscreen before replacement covers the same race from the hardware side — get the file out while the panel still runs.
Why does Crimson say the device does not contain an uploadable image? #
That message means the project was downloaded with Support Upload cleared, so Crimson stripped the readable image when it wrote the panel and there is nothing on the device for any tool to extract — the OEM or another backup is now the only source.
Support Upload is a download-time option, set to protect the integrator's project. When it is cleared, Crimson strips the readable image as it writes the unit: the panel runs its screens perfectly but holds nothing another PC can read back. No command, battery, or build changes that after the fact — the message describes what was written, not the tool reading it.
That narrows the work to sourcing the file. Ask the OEM for the .cd3 first; a panel of this vintage dates to around 2015, so the project may sit in an integrator's archive even after the contact has gone quiet. If the request fails, work through recovering an HMI program when the source file is lost, which ends at recreating the database rather than extracting it.
How do I load a project back onto the panel? #
Download the recovered .cd3 from the matching build to bring the panel back, then make the final download from Crimson 7.11 so the panel's operating system and project move to the current build together and the timestamp fault is gone for good.
That order matters: recovery first, upgrade second. A database authored in a newer Crimson release than the panel's runtime is refused and boots straight back to INVALID DATABASE, so keep the authoring version and the panel build in step on every future download.
Close the job with records, not just a working screen: power-cycle twice to confirm the database survives cold starts, then store the final .cd3 with the machine's drawings and note the build that wrote it. The next person who meets this fault should find the file in minutes, not weeks.
When is replacement the only option? #
If extract reports no uploadable image, the OEM cannot produce the project, and no archive copy exists, there is nothing left to load and the panel cannot run its application again — that is where replacement becomes the engineering decision.
Three findings together end the repair: extract reports no uploadable image, the OEM cannot produce the .cd3, and no archive copy exists anywhere. At that point the panel is sound hardware running an application nobody can restore, and rebuilding the project screen by screen is slower and riskier than starting on current hardware.
For a legacy terminal of this class, that decision points at a replacement such as current industrial panel PCs, which take over the same operator role on a supported runtime — and, unlike the original install, keep the project on storage you can back up. Whatever hardware lands in the cut-out, download with the upload option enabled and archive the project file every time, because that single choice is what made this fault unrecoverable.
Questions asked
Does the CR2025 clock battery cause the INVALID DATABASE error?
No. In the field case the old 3 V cell still measured good and a new CR2025 changed nothing, and pulling the cell only proved the low-battery alarm works. The coin cell keeps the real-time clock alive; it does not store the .cd3 project. Replace it as routine service on an older panel, but treat it as a red herring in this fault.
Can I extract the project with Crimson 2.0 instead of 3.0?
No. The splash line names the runtime the panel holds, Crimson 3.0 build 647.002 in the field case, and Crimson 2.0 predates that database format, so it either refuses the link or returns the same uploadable-image message. A mismatched attempt can also trigger the restart that closes the live window Send Time opened.
Can I rebuild the project by hand if no backup exists?
Only by recreating it. A .cd3 file compiles the whole application, including tags, screens, alarm text, and driver setup, and a panel downloaded with Support Upload cleared holds no readable copy of any of it. Recreating means rebuilding the database in Crimson against the machine's actual behavior, then proving every screen and setpoint against the controller before release.
Will updating the panel to Crimson 7.11 lose the project again?
No, if the recovered .cd3 is the source. Do the Crimson 7.11 download with that file, so the panel takes the current build and the working project in the same step, then power-cycle twice to confirm. Archive the final file with the machine records; a later download made from a mismatched Crimson version is what sends a panel back to INVALID DATABASE.