A new HVX200 with AVC Intra, possible?

[David Heath]

Originally Posted by Barry Green

Close, but not quite. When you go to 50mbps it's still 10-bit, but it drops to 4:2:0.

Whoops, my apologies. Though I'd happily trade 422 for 420, if I gained 10 bit in exchange. That 50Mbs AVC-Intra also gives a resolution improvement (1440v1280) over DVCProHD AND half the data rate does seem to me like the compelling advantages that TingSern Wong was inquiring about. And that's before the option of the 100Mbs version is even considered.

[TingSern Wong]

That brings forth another question ....

Which one is better? 422 at 8 bits or 420 at 10 bits?

I understand the bit depth in terms of digital camera ... the more bits we have, the greater the dynamic range. But all digital cameras effectively captures at 444 at whatever bit depth the AD converter can cope with (12, 14, or 16 bits).

In terms of video, only few cameras (RED, etc) captures at 444. So, my question will be - in terms of manipulating the digital capture by a video application (After Effects, NLE, etc) - which one is better ... 422 at 8 bits or 420 at 10 bits? And lastly, can AVC-I capture 444 at 10 bits (as a codec) - not the camera?

Thanks,
TS

[David Heath]

Originally Posted by TingSern Wong

That brings forth another question ....

Which one is better? 422 at 8 bits or 420 at 10 bits?

Ha, a good question, and I doubt the answers simple. I suspect a lot depends on what you're thinking of doing - 422 would be most relevant for keying work etc, whereas 10 bit resolution would have more importance if you wanted to post adjust colour, gamma etc etc. If I had to choose between them, my vote would go in favour of a 10bit system, especially for a progressive system.

But all digital cameras effectively captures at 444 at whatever bit depth the AD converter can cope with (12, 14, or 16 bits).

"Capturing at 444" I take to mean capturing with equal resolution of luminance and chrominance, and I don't believe that to be the case with all digital camera front ends. Two exceptions would be such as Bayer sensors and 3 chip designs with pixel shift, both of which inherently have higher luminance resolution than chrominance.

[TingSern Wong]

As I understand digital still cameras (DSLR for example), they all capture RGB at 444 level using a Bayer sensor.

Only video cameras can incorporate 3 sensors (and do away with Bayer) - because they have a prism inside. There is no way a prism can be incorporated into a digital still camera - too bulky.

In theory, video camera can indeed capture all 3 primary colours (RGB) at the sensor level. In digital still camera, the only one (so far) capable of capturing RGB at a given pixel will be the Sigma Foveon sensor.

If I understand video - outputing 444 is a waste of bandwidth because no TV is going to show that. Hence, 422 or even 420 is acceptable.

[David Heath]

Originally Posted by TingSern Wong

As I understand digital still cameras (DSLR for example), they all capture RGB at 444 level using a Bayer sensor.

I don't think that's quite accurate. The chip output goes into a de-moisaicing algorithm, which calculates an R,G,B signal for each pixel site, but the algorithm is having to make approximations, a form of interpolation. ALL the pixel sites are capable of contributing to luminance resolution, but that's not true for chrominance.

In an extreme case, think of a scene lit with a deep red light, such that only the red pixels (1/4 of the total) are giving a meaningful output. The resolution will be only 1/2 (H & V) what it would be for a white light.

If I understand video - outputing 444 is a waste of bandwidth because no TV is going to show that. Hence, 422 or even 420 is acceptable.

For display, the reasoning behind sub sampling is that the human eye is less sensitive to chrominance detail than luminance, so it's more the eye won't perceive it than the TV won't show it. The big advantage to 444 comes when effects work is to be done on the material before display. In such case, chrominance detail on the original can translate in to luminance detail in the finished product.

[TingSern Wong]

Agreed ... using a normal digital SLR with a standard Bayer's pattern - that is.

However, if I use a Sigma Foveon sensor, then there is no Bayer ... and the sensor really captures all the RGB info at a given pixel.

[Sergio Perez]

I sincerely do not like the compromise for 4:2:0.

And a gut feeling inside makes me think this will be what we'll get on the next HVX- no avcintra100, but 50. And, if that's the case ( speculative rant here!), it will still be inferior to capturing to a cineform or a convergent designs box capture, for example...

[Jon Fairhurst]

Originally Posted by Sergio Perez

And a gut feeling inside makes me think this will be what we'll get on the next HVX- no avcintra100, but 50.

50 would be fine for me - as long as they include an HDMI output.

Recording uncompressed to a RAID is really attractive when chroma keying or doing other critical fixed location shots. I can probably live with 50 mbps for remote shots.

HD-SDI is the other option for uncompressed, but it would likely be much more expensive than HDMI - both in the camera as the capture card.

[David Heath]

Originally Posted by Sergio Perez

I sincerely do not like the compromise for 4:2:0.

And a gut feeling inside makes me think this will be what we'll get on the next HVX- no avcintra100, but 50.

Well, compromises are inevitable, and my own feeling is that 420 (for a progressive system) is not too bad, certainly not if it means 10 bit, and coupled with the increase in resolution and recording time, compared to DVCProHD.

And apart from doubling record times per Gigabyte, 50Mbs would also allow SDHC memory to be used instead of P2 - that could allow it to snatch the cost advantage away from Sonys EX.

I agree with Jon - "50 would be fine for me - as long as they include an HDMI output."

[Robert Lane]

Originally Posted by David Heath

And apart from doubling record times per Gigabyte, 50Mbs would also allow SDHC memory to be used instead of P2 - that could allow it to snatch the cost advantage away from Sonys EX.

I'm just curious: What cost advantage? The EX1 costs much more than the 200 does and the SxS cards are the same price as P2 for the same GB size.

[David Heath]

Originally Posted by Robert Lane

I'm just curious: What cost advantage? The EX1 costs much more than the 200 does and the SxS cards are the same price as P2 for the same GB size.

Sorry - I could have been clearer. The reference was intended to refer to memory costs, and whilst P2 and SxS cards are similar in cost per GB, the higher bitrate of the HVX means it's effectively 3x the cost *per minute*. A move from 100 to 50Mbs would cut that to only 1.5x, but the use of SDHC would make memory costs of such a camera far CHEAPER per minute than the EX - possibly only half as much, whilst retaining an I-frame only system.

The latest prices I've got for 16GB cards are £490 for P2 and £440 for SxS, which translates to about £30.60/min for P2 and £9.20/min for SxS. Two SDHC 8GB cards should be about £160, and with 50Mbs AVC-Intra, the cost now works out to something of the order of £5/min.

[TingSern Wong]

Why SDHC? Why not CF - which is more rugged? The present SD slot in HVX202 is not for recording ... it is used to keep configuration data. If you remove that SD card, you will loose the configuration that you have set (F1 -- F6).

[David Heath]

Originally Posted by TingSern Wong

Why SDHC? Why not CF - which is more rugged?

If I was designing the camera, CF would be exactly what I would choose. But Panasonic are currently only backing P2 and SD memory, and if we restrict the choice to one of those, 50Mbs is possible with SD, 100Mbs probably realistically isn't.

But for a long time Sony meant Memory Stick. They seem to have had a big change of heart and adopted both SD and CF for various products - maybe Panasonic will follow suit? My own feeling is that for a next-gen camera, two P2 and two CF slots would be most desirable.