HDK2 - video sync change interrupt lost

[rpavlik]

A small number of HDK 2-based setups, so far only noticed with NVIDIA GPUs, do not produce or propagate the video sync change interrupt properly, particularly on signal loss. This can result in a bright bar after direct mode app close or HDMI unplug until the HMD is power-cycled, or in rare cases, trouble running direct mode applications (applications will run but screen will remain dark).

The latter case seems to be that sometimes there is an extra signal gain/loss associated with starting direct mode apps on some NVIDIA GPU and driver combos? This may also be related to AVMUTE - the 0x8520 status register when running in NVIDIA direct mode (when working properly as well as in some failure cases) is 0xbf.

Seems to be very system-specific, unsure if it's tied to a single HMD or if it's associated with the GPU and driver, but if you don't experience the problem with your test hardware, I haven't found a way to help you to reproduce it and trigger the problem.

• Consider adding back low-rate polling? (or checking the interrupt flag, if perhaps the interrupt is getting pre-empted by another?)
  • look into the configuration of the various PMIC modes/levels and confirm that the current setup is the preferred one - should we move both the bno and toshiba chip to low-level (rather than high and medium) and enable round-robin scheduling?
• Is this related to propagating an interrupt through the circuitry (including a bi-directional level shifter) from the combined address-select/interrupt pin on the toshiba chip back to the xmega? Some evidence suggests there may be (in general) lot to lot inconsistencies with the type of level shifter circuitry in use.

[rpavlik]

One device that showed the issue (with the enhanced NVIDIA failure mode, where new direct mode apps won't display) is SN CT1630V00101172 (from a Reddit user)

Mine, which doesn't show this issue, is CT1630V00100812.