Harp sound card fails to output TTLs

[k1o0]

This card emits sound, and provides a TTL output marking the onset / offset of such sound.

We train mice on a task that requires auditory signals (pure tones) to be played. Typically, mice do ~500 trials on this task.

We noticed that on ~1% of the trials, and this on all setup we deployed (10), the Harp TTL for the audio signal does not transition from High to Low state after the sound has been emitted. We observed this in more than 100 sessions.

As a result, the next sound generation is not visible on the TTL trace, and we cannot know precisely when the sound was emitted, which is an issue when it comes to synchronization.

[bimac]

Using the following state machine and the harp soundcard I can't observe any issues with TTL output whatsoever:

sma = StateMachine(self.bpod)
sma.add_state('tone', 0, {'BNC2High': 'high'}, [self.bpod.actions.play_noise])
sma.add_state('high', 0, {'BNC2Low': 'low'}, [])
sma.add_state('low', 0.1, {'Tup': 'tone'}, [])  

[bimac]

here's a graphical representation of the above ... I let it run for >1000 periods, no issues at all. Pretty sure that this is not a hardware issue.

[mayofaulkner]

All sessions where goCue_delay qc check fails (Miles to send Florian query)

[bimac]

Quering datasets

• from 2023-01-01
• recorded with harp
• with goCue_delay failing in up to 25% of trials:

- 4b02080d-3a7f-4185-a747-2206fdf3dc1b / 2024-10-03 / _iblrig_cortexlab_mesoscope / samuel_cuedBiasedChoiceWorld8.23.1+dirty/_iblrig_tasks_passiveChoiceWorld8.23.1+dirty/_iblrig_tasks_passiveChoiceWorld8.23.1+dirty
- 8f2acc4a-7977-4dcb-8ed8-f1f5a61e66cf / 2024-09-19 / _iblrig_cortexlab_mesoscope / samuel_cuedBiasedChoiceWorld8.23.1+dirty
- d975cd8a-860a-458e-804f-3c1729c1746b / 2024-09-19 / _iblrig_cortexlab_mesoscope / samuel_cuedBiasedChoiceWorld8.23.1+dirty
- 1ce1d217-f1e6-461b-b184-322310a4d893 / 2024-06-18 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_trainingChoiceWorld8.19.7/_iblrig_tasks_passiveChoiceWorld8.19.7
- 38751b1c-2b64-40c9-b0a7-95a922a4de98 / 2024-05-28 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_trainingChoiceWorld8.19.2
- 115412e5-0d00-4c86-b7b7-1409067ce89b / 2024-05-24 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_trainingChoiceWorld8.19.2/_iblrig_tasks_passiveChoiceWorld8.19.2
- 36242389-967d-4db6-b00f-fcece8c5b43f / 2024-05-14 / _iblrig_cortexlab_mesoscope / samuel_cuedBiasedChoiceWorld8.19.0/_iblrig_tasks_passiveChoiceWorld8.19.0
- d085b101-dc41-4268-9a51-4dc6d0a415dd / 2024-05-13 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_advancedChoiceWorld8.19.0
- e2a4702d-54f4-4e60-ac07-5eed06fa6b65 / 2024-05-10 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_advancedChoiceWorld8.19.0/_iblrig_tasks_passiveChoiceWorld8.19.0
- b3501ee8-0829-4399-a500-0dab3e62c14d / 2024-05-09 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_advancedChoiceWorld8.19.0
- 0dd8a981-3db4-47c8-8357-4aaf6b90c9c0 / 2024-05-08 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_advancedChoiceWorld8.19.0/_iblrig_tasks_passiveChoiceWorld8.19.0
- 908ac556-b741-4ee2-8665-7897ba5f2683 / 2024-02-19 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_trainingChoiceWorld8.15.5
- 3dfc4da1-a432-470a-800d-5abbd296eb61 / 2024-01-29 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_trainingChoiceWorld8.14.2
- 600ce551-a3ab-4f41-9c79-a2e781abe2a7 / 2023-08-03 / _iblrig_cortexlab_mesoscope / _iblrig_tasks_biasedChoiceWorld8.8.1

I can't see the phenomenon of the TTL not going low again though

[bimac]

Could not reproduce - closing.

Both Bpod data and DAQ data show correct pulse lengths and no mayor outliers whatsoever. This may have been an issue with extraction or plotting that has been resolved in the meantime?

[bimac]

For reference, this is the original observation (unknown EID)