Cineform issues with realtime capture on 7d - NeoHD v3

[Anmol Mishra]

OK. Cineform records fine with my HV20 and does the pulldown in realtime. The settings used are Progressive AND pulldown. 29.97 PsF comes up in the BM window.
I have been trying to get it to work with the 7D over the past week - but have had problems.
It seems that it loses the pulldown cadence on 24P in the middle of a capture. I used the .AVI option for capture.
I am pretty sure that its the cadence thats lost, since the first few frames have a similar issue while CF tries to detect the right cadence.
Has anyone else seen a problem recording direct from the 7D ?
Here are the screen grabs.

[David Newman]

What class of PC are you capturing to? What CPU load do you get while capturing? The compute load will likely be higher for the 7D vs HV20, so if your PC is burdened it can drop a frame, throwing the pulldown off.

[Anmol Mishra]

Hi David. I am using a 2.4GHz C2D. I am using a version of XP that operates from RAM - so no writes to boot disk. Unnecessary processes disabled.
Now I was using Film Scan. If your premise is correct, a High or a medium would work.

Also, what happens if I record without removing pulldown ? Is there a smaller chance of lost frames ?
Also, is there any loss if I record without removing pulldown and then remove pulldown later with HDLink ?

[Anmol Mishra]

Also, you stated that the compute load is higher for 7D vs HV20. Why is that ? Both offer a 1080p signal within a 60i stream. In fact, the 7D is letterboxed and pillarboxed, so the effective compression area is smaller. Should'nt it be easier to compress the 7D signal ?

[David Newman]

While it is true reduce pixel count of active data, the data is a lot sharper, to the point of aliasing in the 7D, whereas the HV20 is softer through an optical low pass filter designed for 1440 (less than the 7D's 1620 with OLPF for 5K.) 2.4Ghz C2D is on the edge for Real-time capture, with or without pulldown removal (generally pulldown helps a little.) Dropping to High will help, maybe enough.

[Anmol Mishra]

Hi David. This part I do not understand - so please humour me a but more. Pulldown requires extra processing due to the act of cadence detection. However, it probably saves some time spent doing compression.
Are you suggesting that the act of pulldown takes less time than compression for the extra 6 frames ? Hence the total time spent compressing 24P is faster than compressing 60i ?

[David Newman]

It is basically a wash, the savings of 20% reduction in frames to encode, is almost consumed by the pulldown detection filter.