Frame corruption

[Suren A. Chilingaryan]

The hardware running on ipecamera4 on April 22, 2015 corrupts frame (at least 12bit, 4Mp default configuration). Whenever I request a frame there is extra 64 bytes before frame magic. You will see it even if you request a single frame. This is no problem per se, but the following frame is 32 byte short. I.e. full size of data between two frame magics is 33800096 bytes while according to the format I expect 33800128. The sample data is attached.

[Suren A. Chilingaryan]

Reproduction: You will get extra data in the beginning even with: /usr/local/bin/grab.sh --single

The shorter frames I get in the streaming with AUTO triggering (but not with software triggering). To see the problem, you can use 'pci -g' with AUTO triggering enabled. You will see a lot of lost frames. To see actual data, I have to enable raw data logging in the pcitool. Then, all data comming from DMA will be written to the files in /mnt/frames and one can see that happening. You can enable/disable this mode by uncommenting #define IPECAMERA_DEBUG_RAW_PACKETS "/mnt/frames" in the /root/software/pcitool/ipecamera/private.h

[Suren A. Chilingaryan]

Ok. According to the new format, we should now have 33800096 bytes long packets. But we still, though more rare, have shorter frames. For instance, take a look on 33800032 byte frame in the attachment.

[Matthias Vogelgesang]

What is the estimated time for fixing this issue? As of now, the camera is barely usable with the GUI and benchmarking it brings up really strange results.