WP13620 RCG Upgrade

Jonathan, Ryan C, Fil, EJ, Erik, Dave

This morning we successfully upgraded all the front ends from RCG5.6.5 to RCG5.6.6. This required rebooting all the front end computers, plus the two diskless support machines (h1build, h1ecatmon0).

The main reason for the upgrade was to fix the DAC-Flips-Output-Polarity-At-2-Billion-Counts bug, which has been seen as recently as this weekend on H1 SUS systems driving LIGO-DACs (though the bug applies to all types of DACs).

Other fixes are with mixed DAC type mis-mapping, which required LIGO-DACs to follow Gen Std cards on the bus. Also RCG5.6.6 provides full LIGO-ADC support.

We took the opportunity to verify BIOS settings as part of rebooting the front ends. We also disconnected unused Adnacos from the host adapter cards (i.e. Adnaco backplanes with no cards installed on them).

This all went relatively smoothly, with the only surprise being EDC flapping on its DAQ connections on both legs. Erik remembered that a negative wait time is needed, which fixed the issue.

Before today h1omcpi was running a custom RCG to add an IPC delay for the channels it receives from h1ioplsc0. It no longer needs a custom RCG, the new RCG5.6.6 has provision to set the delay via a file.

At the end of the build/install pre-upgrade work done yesterday (Tue 15sep2026) h1crsproc and h1psliss were falsely reporting modifed INI files. We verified these were false and the ini files were byte identical. These was borne out after the reboots, no DAQ restart was required for any of today's work.

WP13610 Enable PERR/SERR Reporting In BIOS

All front ends now have PERR/SERR PCI bus error reporting where applicable (V4, W22 have these option, W33 do not, we assume it is the default here)

We also verified the BIOS setting to keep the machines powered down following a power distruption when power is returned. This prevents front ends from trying to boot prematurely.

WP13606 Disconnect Unused Adnacos

All done with the exeption of h1susauxex A4.

WP13609 h1oaf0 Adnaco investigation

Erik found the adnaco fibers were dirty and cleaned them. We are waiting to see if this has improved the error rate.