Capturing Uncompressed 24p from the HD-100

[Alex Bowles]

Hi all,

I'm trying to better understand uncompressed the 720p60 out from the JVC HD-100. Specifically, I want a 720p24 project, and need to know if this is even possible with the HD-100.

I understand that the sensor has a constant rate of 59.94hz, which, I assume, translates into 60 full progressive frames of 720x1280 each. (In other words, if I were shooting a film, I'd be shooting 60fps.)

So how does it deliver 24p when going to HDV? More specifically, how does it get the motion blur that's characteristic of film? If if simply drops enough some frames while repeating others so that 24 distinct frames are delivered, won't the motion take on an awful jerky, almost undercranked quality? And if the motion blur IS there, how does the camera combine frames from the sensor scanning at 60hz so that the 24 frames that are actually recorded have the requisite softness?

More importantly, how do I recreate this processing when tapping the uncompressed output? Or, if I want to use the Uncompressed out, am I forced to accecpt video with the super high dynamic resolution common to video games, and abandon any hope of getting 24p material from this camera in anything but HDV?

[Mark Silva]

I'm not sure you can.

To the best of my knowledge 60p is all you get out of the component connections.

[Alex Bowles]

You're right, but that shouldn't be a problem. After all, the camera itself manages to take that same 720p60 stream coming off the sensor and turn it into a 24p MPEG file that it records to tape.

So if I'm simply bypassing the MPEG encoder by taking the uncompressed out, shouldn't I be able to conduct some operation on this signal that the camera uses when deriving 24p material, but without putting my signal through the brutality that is MPEG-2 compression?

What is this operation? And is it something my capture card could handle (say, Black Magic) or do I need to get another box that would go between the camera and the capture card (assuming I'm recording direct to disk in this scenario.)

A

[John Vincent]

I believe I read that the best possible quality the JVC 100 can produce is when you digitize directly out as you're shooting via the three component outputs (NOT the firewire).

I believe this gives 4:2:2 color space as oppossed to 4:2:0. This is, as you say, because you are bypssing the encoder. So this method would be the best route for doing greenscreen work....

But someone please correct me if this is wrong.

john
evilgeniusentertainment.com

[Bill Ravens]

that is correct. now all you need to do is a 2-3 pulldown if you want 24p cadence.

[Mark Silva]

^^ by that logic one should be able to capture 60P from component live (8 or 10-bit uncompressed)

then just add the pulldown with cinematools?

has anyone done this successfully?

somehow I don't think its that simple, but I could be wrong.

[Matt Setnes]

I understand that component out will give you 4:2:2, but may I ask why not fully uncompressed. If it's bypassing the encoder, what else is reducing the footage?

[Carl Hicks]

Originally Posted by Matt Setnes

I understand that component out will give you 4:2:2, but may I ask why not fully uncompressed. If it's bypassing the encoder, what else is reducing the footage?

The component out IS fully uncompressed, 4:2:2

Regards,

[Matt Setnes]

strike me if I'm wrong but isn't 4:4:4 full uncompressed?

[Stephen L. Noe]

Originally Posted by Matt Setnes

strike me if I'm wrong but isn't 4:4:4 full uncompressed?

It depends on what camera head your getting the uncompressed off of. In the case of JVC HD-100 (ProHD et al) 422 is the sensor A/D converter and it's the uncompressed analog signal's format. It comes directly off the camera head and does not get compressed. The signal is a whopping 869Mb/s.

I've provided you with an uncompress clip off of the HD-100 below.

Click Here for Clip (580MB)

This clip will only be up for a short while to give you an idea of what kind of bandwidth it takes to work in uncompressed. The clip is only 5 seconds long and weighs in at a whopping 580MB. Needless to say you need an ultra fast SATA or SCSI array (Raid 0) to handle it. Now imagine the uncompressed 1080p requirement!

Anyway you can try your 24p pulldown technique on the file.

Regards,

Stephen

[Alex Bowles]

Thanks Stephen

Unfortunately, I don't have the capture path I'm after set up yet, but where I'm going is this:

SDI out from the JVC, into a BlackMagic card (with a Miranda box for an A/D conversion, if necessary) with the card applying the 3:2 pulldown to extract 24p from the 720p60 signal coming out of the camera, compressed at 2:1 (lossless) before going to disk. Capture is managed via FCP.

Has anyone else tried this out?

A

[Steve Oakley]

Originally Posted by Stephen L. Noe

It depends on what camera head your getting the uncompressed off of. In the case of JVC HD-100 (ProHD et al) 422 is the sensor A/D converter and it's the uncompressed analog signal's format. It comes directly off the camera head and does not get compressed. The signal is a whopping 869Mb/s.
Stephen

sorry, but its no where near 869mbytes/sec, not even 1920 is that high

10bit 1920X1080 @ 59.94 = 331Mb/sec
10 bit 1280X720 @ 59.94 = 149mb/sec
8bit 1280X720 @59.94 = 110mb/sec
8 bit 1280X720 @29.97 = 55mb/sec
8bit 1280X720 @23.976 = 44mb/sec


next there seems to be some mass confusion about the basics of this camera.

the CCD can be clocked out at 60,30,25,24 and actually slower with the shutter adjustment, but this is basic defaults. what the CCD gets clocked out at has nothing to do with whats going to tape or out the video outs because its all being buffered in the camera into a suitable output format.

what the camera seems to be doing in 30(29.97) mode is simply putting out 2 frames @ 59.94. visually its 30P, but techincally its 60 with frame doubling. this makes for a "legal" video signal. with some ram for buffering the image in the camera, its not a big deal to do this. likewise you can clock the CCD@24/23.976 and output 59.94 by basically treating each frame more like a field in 29.97 - that is you add 3:2 pulldown into the signal using full frames so that 23.976->3:2->59.94 and you can record that to tape, or send it out as a standard 720 video signal and all is well.

I wish that JVC would publish a white paper on this and really specify what the camera is doing, rather than leaving it up to educated guessing and math on the end user's part. for example, while the camera does 30P, you can't capture it @ 29.97 in FCP.... yet. you have to capture it at 59.97 which doubles the disc space for no good reason. its up to your NLE to detect this through various means - either bits in the FW stream indicating duplicate frames, or in the vertical interval data in analog/SDI in the case of DVCpro. I don't know if HDV supports VI metadata, but certainly in the basic FW data stream.

I've tried to capture from the camera with a kona LH to DVCpro100 just as a test, and found that the audio went wildly out of sync in FCP. I'd like to do a few things in uncompressed, but this test was not helpful. I think that there are still some FCP and Kona driver problems to be fixed before this will work, at least in FCP 5.0.4

Steve Oakley

[Stephen L. Noe]

That's Mb (small b). My system reports 869Mb/s as the datarate for uncompressed 1280x720 for the clips I capture. Do you think the system is misreporting them???

[Thomas Smet]

Mb not MB. you multiple the MB by 8 to get the mb. For example 25 mb DV uses 3.125 MB. There are 8 bits in one byte. 869 mb/s = 108.625 MB/s. or pretty close to the 110 MB/s rate of 59.94 720p.

This is 869 mb compared to 19.7 used for 720p HDV for a 44 to 1 compression ratio.

[Steve Oakley]

Originally Posted by Stephen L. Noe

That's Mb (small b). And yes, it is precisely 869Mb/s

1280X720 8bits @ 59.94 = 110.48Mbytes/sec X 8 = 883Mbits/sec

thats not including any audio...

Steve Oakley