For at least 2 Zadorlab sessions all feedback_times are nan for negative feedbackType

[mschart]

All feedback_times are nan for negative feedbackType for
eid = '5522ac4b-0e41-4c53-836a-aaa17e82b9eb'
and
eid ='e349a2e7-50a3-47ca-bc45-20d1899854ec'
Something wrong with that rig? zadorlab/CSH_ZAD_029/2020-09-11(15)

In [118]: eid = '5522ac4b-0e41-4c53-836a-aaa17e82b9eb'

In [119]: trials = one.load_object(eid,'trials')

In [120]: trs = np.where(trials['feedbackType']==-1)[0]

In [121]: sum(np.isnan([trials['feedback_times'][tr] for tr in trs]))
Out[121]: 138

In [122]: len(trs)
Out[122]: 138

None of them are nan for a different session, say eid = 'd0ea3148-948d-4817-94f8-dcaf2342bbbe'.

[juhuntenburg]

This affects all sessions between 2019-05-06 and 2019-09-17 and multiple (not all) sessions between 2020-09-11 and 2020-09-20. Has not happened since, does that ring a bell to anyone?

[nbonacchi]

So both sessions are EphysCW 5522ac4b-0e41-4c53-836a-aaa17e82b9eb on September 15th 2020 and e349a2e7-50a3-47ca-bc45-20d1899854ec on September 11th Same rig: _iblrig_zadorlab_ephys_0

FeedbackTimes for negative trials should be the onset of the white noise extracted from the TTL of the CFSoundCard.
The QC plots show that errodCue, goCue, and iti delays are failing all these metrics rely on the sound card TTL so it could be a bad HW connection. If this is the case i.e. the raw data of the sync channels reflect this, it's not an extractor problem and the data would be unrecoverable.

I tried looking at the raw sync pulses but it's not working for me. @k1o0 the task_qc_viewer had something that showed the raw sync pulses right?