Meeting notes from March 28th, 2023

[glopesdev]

Meeting Notes

Initialization Sequence

• Hyper-nodes: if they are set to same hub index we need this to fail (e.g. two headstage nodes pointing to same index would cause config problems); reserve address at the device level:
  • table on ONI context task
  • pass-through mode flag:
    • no two devices can point to same address;
    • in case of one device setting a port to pass-through mode then all addresses related to that port should be counted as configured;

Pass-through Mode

• hubs are serializer/deserializer
• pass-through mode is actually the purest mode, but there are limitations on how you can use the serializer (write custom transport layer)
  • miniscope v4 there are two nodes (one technically a slave of the other)
    • master configures the entire pass-through mode / serializer / acquisition (get it running): output is video stream
    • a second node contains the BNO data (access is in polling mode / I2C; only one independent exclusive access is allowed); periodically the node polls the I2C to get BNO
• notion of pass-through mode is actually outside the ONI specification:
  • it is an implementation detail (capability) of the specific RIFFA board
  • discussion of whether it makes sense to add it to the spec;
  • or allow extensions to particular hosts (how to integrate with software?)
  • suggested looking at how OpenGL extensions / OpenAL extensions are structured
  • reserved addresses which are hardware-specific (e.g. passthrough mode specific ONIX)