New JVCPRO HDV codec
 

My understanding is that JVC uses a new codec to accomadate the new 60P frame rate on the HD200 / HD250.

Does this new codec improve on the existing 6 GOP (19mb/s) codec?

If so, this would mean that 24P/30P footage COULD have less compression artifacts.

same 19mb/s for 60P - but GOP is more - its 12GOP to tape.

Thanks,
I know this. I'm asking since they did update the codec to offer 12 GOP for 60P, was there any change AT ALL to the existing codec for 24P/30P?

In my opinion ther 12-GOP if anything could potentially worsen it.

Or does it only do 12-GOP on the 60P only?

You really can't get much better than 6-GOP for mpeg2 that actually works well, save for straight I-Frames, but that probably wouldn't work at all.

I have not noticed any kind artifacting at all with my HD100 except for the CA when I zoom in past 75%.

The 12 GOP won't change based on the frame rate of the format you're recording in. My understanding is that, as far as the MPEG codec is concerned, everything is going to be treated as 60p with a 12 GOP. If you're recording in 30p or 24p format, then the proper pulldown will be performed upon playback/capture to sustain the corresponding framerate, but everything will be based off of a 60fps system.

Did you have this idea confirmed by JVC?

It is VERY possible to use 2:1 pulldown to get 25p and 30p carried by 50p and 60p, but has JVC said that's what they'll do?

The codec issue is covered at: "How much better is 12 GOP?"

I though it was not pull down but instead placeholders. When the 24p is captured in Liquid the stream is read as 59.94 but when the clip is in the rack it is 23.98. The placeholders are disregarded when the clip is actually placed in the rack (bin for all other NLE's). When Liquid writes the 24p back to tape it will embed it in 59.94 just the same way it was captured from the tape (markers back into position). So upon capture it always treats the stream as 60p and then disregards placeholders.

Conversely when Liquid captures 30p the stream is aways read as 29.97.

No- I could be way off. Wouldn't be the first time.

I wouldn't be at all surprised if it does straight 60P recording only then adds whatever flagging to the bitstream to get your 30p/24p formats.

I believe thats what the varicam does.

Does the HD100 derive 24fps from 30fps?

Or does it derive 24p/30p from the 60P signal it generates internally
(like you get live on the component outputs)

Not quite how it works. The HD100 does record a 60p stream when shooting 24p, but uses repeat-frame flags to handle the 2:3 pulldown process. That means that those pulldown frames take up no space in the data stream at all, they're just a repeat flag. So the HD100 does record 24p more efficiently than it records 30p, because the same amount of bandwidth only needs to be spread around 24 frames instead of 30. However, the transport stream is flagged as 60. That's why the older JVC products can transport the newer 24p footage; remember the JVC HDV spec doesn't provide for 24p, it provides for 25p, 30p, and 50p and 60p. So they fit 24 in there without breaking the spec or creating a new format. Rather clever actually.

So 30p is a direct encoding of 30fps, because the format allows for it. 60p was part of the HDV spec from day one, they just didn't have an encoder that could handle that much data until now. 24p wasn't part of the spec so they implemented it within a 60p transport stream.

Again, sort of but not exactly. The VariCam does record 24p within a 60p data stream, and that's what the JVC does as well, but the VariCam actually provides the full frames in the stream. JVC doesn't have to do that, MPEG allows them to just say "repeat the last frame". So yes it's similar, but it's more efficient bandwidth-wise in the MPEG implementation because an entire frame can be repeated with a simple flag.

No, it reclocks the CCD and scans it at 48hz instead of 60hz (60hz scanning for 30p, 48hz scanning for 24p). It then discards every other frame, leaving it with 24 frames.

Sort of but not exactly (again). 30p is derived from the 60p signal, yes -- it drops every other frame upon recording. So you get 60 on the live output because that's how the system is running internally, but the MPEG encoder in the HD100 can't handle that much data so it only records every other frame.

Because there are 60 frames being processed by the DSP, that's how they can do their "motion smoothing" filter -- they overlay a ghosted version of the "dropped" frame, so you get the motion from two frames blended into one.

Same thing but slightly different with 24p, they run the CCD at 48hz and scan 48 frames per second, then drop half of them and record the other half. It's a little different because the camera head output is still 60fps, so there's a duplicate frame added every once in a while to round out the 48fps sequence to 60fps. That's only on the analog outputs though; internally it only works with and stores 24 fps.

Wow! Thanks for your very thorough clarifications Barry! :D

It does raise the possibility that the new camcorders could offer two ways to record 25p/30p. Either directly as done now -- and by 2:1 pulldown using Repeat Flags.

Not clear -- if there would be any difference between the current 6 GOP option and a 12 GOP with 2:1 pulldown. Need coffee.