Dreaming of bad pixels

I had a dream.last night which I remembered clearly on waking up.

In the dream I was working on the original harry in the dark finding bad frames.

The original harry cames with 2048 PAL frames across 4 Fujitsu Winchester drives two for luma and one each for the two colour difference signals (later the harry lp came with 3024 and the flash harry came with even more before being replaced with the henry which on launch had 5 minutes before the introduction of the 15 min chatterdisk Dylan disk packs). By comparison my current flame has about 5.5 million PAL frames available.

These disks could have a bad pixel similar to what today we’d call a NAN frame but were in factor a bad sector on the drive. In the interests of qc what we would do each time the drives were formatted was to get a black frame from paintbox and check it was ok, then stretch it to the full length of the disk creating original frames on each frame.

Then with the lights out we would sit and jog frame by frame staring into the abys for single bad dots. 2048 times.

Any time we found a frame we would cut it out and attach it to a clip in another reel and then carry on…

We would then delete just the bad frames clip. Get one of the good black frame and stretch again to fill the bad frames and recheck. We’d do this 3 times to be certain that the bad frames were always bad rather than the result of a processing error or signal noise on the smd cabling.

At the end we would have a ‘bad frames do not delete’ clip and save it in the library Effectively quarantining the bad frames away from jobs.

It would generally be about 20 frames long.

If this sounds like insanity it was worthwhile as Sony used us for their camera launch tapes for a while and I had a letter saying that of all the facilities they tested we had the least errors. I loved the qc those guys did. I once did a 17 min d1 tape for the launch of the dcp 360 camera and a month after delivery I got a fax with print outs of two frames with bad pixels circled and ‘please fix’ written below. I was happy to fix things for people with that attention to detail - all we did back then was save a frame into the paintbox buffer then grab the next frame and restore the one pixel…

These days we have too many frames to care and the pixels are so much smaller in 4K we probably can’t see anything other than a bright red NAN error but it made me think about how do we qc bad frames these days. We have Nan fixing plug-ins in nuke but what about flame. How do we find bad pixels without sitting in the dark?

I need better dreams.


Every QC facility that we get reports from has software that finds those buggers.