NEX-VG10 + nanoFlash

[Dan Keaton]

Dear Steve,

We are in complete agreement: "The time between frames for 29.97 simply can NOT be be made to be the time between frames for 23.976."

We cannot adjust the time between frames, that is up to the camera.

The NEX-VG10 puts out 1080psf29.97.

The best we can do is record 1080p29.97, or using cranking to achieve 23.976, but we can not change the time between the frames.

[William Hohauser]

Originally Posted by Steve Mullen

If you want 24p, I don't know why folks ignore the real solution: Twixter. It never gets mentioned, yet for several years it keeps doing the job.

Now could 25 be converted to 24 -- I doubt by the nano.

25 can be converted to 24, it's done a lot by video to film facilities. One to one frame transfer with professional audio work to slow down and frequency compensate the audio just as 24f film has been sped up in PAL countries to 25f for decades now. It was the one reason PAL video equipment was in demand here in the US by aspiring film-makers.

The question is.... why do we want to shoot anything destined for film with anything but true 24p equipment anymore now that it's so available? Soon film will be a non-issue anyway as all cinemas will have quality digital playback and will be able to play any frame rate.

[Steve Mullen]

I think the confusion is that the nano can do under and over crank. So yes, 30 can be come 24. But it means the video simply plays faster, like a silent movie plays 18 as 24.

Or maybe it plays slower. Which ever it is, it's a speed change not a frame rate conversion.

Which might mean the nano could speed change 25 to 24 which might be very nice!

[Dan Keaton]

Dear Steve,

If one uses our Crank feature, so that the baseline (first number) is 24, and the Crank Number is 30, then you will have the effect that your described: 24 Frames will play out in 1 second, and the full 30 frames will take longer than 1 second.


But, if one uses our Crank featue, with 24 and 24, one will record 24 frames per second, and one will playback 24 frames per second.

The nanoFlash will take in 30 frames per second, then discard appropriate frames so that only 24 frames are recorded.

[Steve Mullen]

"... then discard appropriate frames so that only 24 frames are recorded."

Unfortunately, there are few or no "appropriate" frames because the TIME each 30fps frame is captured does not match the TIME each 24fps SHOULD HAVE BE captured.

All that can be done to get a 24fps stream is to choose from the 30fps stream 24 frames where each of these frames has the least time ERROR. This creates two problems:

1) Jitter. Since sometimes an EARLIER in time frame is selected from the 30fps stream and sometimes a LATER in time frame is selected from the 30fps stream, the timebase of the 24fps stream is unsteady.

2) Since the same selections are made each second, the pattern of selection creates a cadence that may be visible.

The only way 30 can become 24 is for 30 frames to be broken into 2 million pixels, each of which has a calculated motion vector (obtained from several previous frames).

Now each vector is adjusted to represent 1/24th rather than 1/30th second.

Next, each pixel is moved by the modified motion vector to a new location in what will be the NEXT frame. In this way, 24 new frames are CREATED.

The problem is that unlike MPEG-2/-4 compression there is no frame the CREATED frame can be compared to, hence no ERROR compensating information can be created. Thus, IF an error is made for one or more pixels, these errors can propagate until a computed frame perfectly lines up in time with a 30fps frame. Now, the errors can be found.

Then, the errors must be carefully backed-out of the appropriate previously computed frames.

This is why the conversion process is so slow unless implemented in hardware as they are in 120Hz and 240Hz HDTVs. (Although this is a simpler process because motion vector times are always the same.)

I believe you can download a demo version of Twixtor. When I tried it many years ago it would occasionally drop a frame. I assume it is now much better.

[Dan Keaton]

Dear Steve,

As I said above:

We are in complete agreement: "The time between frames for 29.97 simply can NOT be be made to be the time between frames for 23.976."

Then I responded to your comments that we are creating slow or quick motion, which we are not.

As you properly state, it takes a lot of computing power to do as you suggest.

What we are offering is a solution that may not be for everyone.

If someone wants to record 1080p29.97 out of the nanoFlash, with the proper time per frame, then they can do it.

If they want 1080p24 (without each frame occupying the precise 1/24th of a second time), then they can do it.


And of course one can use Twixtor.

[Steve Mullen]

My only issue with the claim that 24p can be recorded is that very few understand the fine line between getting 24fps with frames each captured at their correct time and simply getting 24fps.

Sony tried this with their Z1.

The problem is that I predict, since I cannot see it, that it's got to look really really bad. There is a huge error every time a the system tries to pick a frames that should be 1/24 th second apart from frames that are actually 1/30 the second apart. No time to calculate the percent error, but it is going to be RELATIVE to the timebase large and not consistent.

I think the average error is 25% in capture time.

Just realized that I didn't consider thar somewhere in every 30 frames 6 have to be dropped.So when a frame goes away there is a gap of 1/60th second which means one or more of the new frames is going to be a long way from where it was captured.

One good thing about the VG10 is that even in the 30fps world the shutter speed can be set to 1/50th.

yes a downloadable ProRes 23.976 file would be perfect!

[Dan Keaton]

Dear Friends,

I think the next step would be to record some video, with motion, to see how it looks.