4:4:4 12-bit Uncompressed DVX100

[Wayne Morellini]

DV uses something like mjpeg, HDV uses mpeg2 varient (and it seems more variability than the JVC HDV mod, though I'm still really unclear on what I was reeading late last night). So you can regard mjpeg as more sloppy than mpeg2 at compression, while mpeg4 is more tight again, and wavelet as possibly more tight again. A company like Cineform has a very tight visually lossless codec yielding upto 6-10:1.

So with the newer techniques the latter codecs store they can store more detail in less space. But still I think 50Mb's would have been better, until I saw this variability thing (if I'm reading it right). This might also explain it's adoption, that Sony has Bluray as there pro version, so that canbe really variable.

I think cineform is making it's codec available to variouse licencies, so a hardware codec chip might become avaialble for it.

Any thoughts?

Wayne.

[Régine Weinberg]

quote:
-----------------------------------------------------------

The image can be demosaiced in a similar manner to obtain the HD image, which is 4 times the size of an individual sensor, 1546x990.

_____________________________________

that is near HD
so if it is true how to do this anyway
maybe a stupid question
but I had a lousy night
ronald

[Régine Weinberg]

WOW
than we take a static adapter
and we ha do have the Indie Cam
DO WE ??
that sounds so cool dammed cool anyway
to cool to be true

OR ??

[Marin Tchergarov]

<<<-- Originally posted by Ronald Biese : WOW
than we take a static adapter
and we ha do have the Indie Cam
DO WE ??
-->>>
Yes Ronald,WE DO!
As I understand the SD RAW mode for DVX100 is done and that "HD mode" is a BONUS !!!:)
Canon XL2 is the next one in the Andromeda's "Roadmap" - more resolution plus changeble lenses! ...I know nothing about what lenses can be used...may be lenses from Canon still cameras?

[Wayne Morellini]

I know the programing development kit may make it easier with xl2 but if it doesn't do pixel shifting too get at least 720p, then you would be better to go to the Sony instead, cheaper, HD is understood natively in the controls etc. Just an opinion.

Actaully with the programing developement kit, can you just program it to take the raw data and output it out the firewire?

[Régine Weinberg]

Dear Juan & Wayne

Yes, found it in the German XL2 documentation, written like this, exact like this:"CCD Pixel Shift, horizintal"

A lousy translation but as the XL2 has changables lenses
a mini 35 Adapter, the Juan Mod and the Indiecam is born

What do you think Juan ?? will it work I like to know

Ronald

[Wayne Morellini]

I forgot, I don't know where ever that pixel shift is usefull for HD or not. But if so, great. But still, does the programmers kit allow you to output raw to firewire? This would solve many problems (and be cheaper) now if the camera was half the price ;).

[Marin Tchergarov]

I'm waiting for input from Juan too...but while waiting here is my thoughts:


Wayne,
Green pixel shift on DVX100 produses some kind of a mosaic since the greens are in
between other pixels,so with the help of a Juan's custom "deBayer" application the
result is REAL HD image!
If we assume that the REDs end BLUEs are shifted too,then the RAW output should be equal
to the output from CMOS sensor with the 3xHorizontal and 3xVertical number of pixels...
However the REDs and BLUEs are aligned,but still I like the images posted from DVX100 ( http://forum.reel-stream.com/viewtop...568415fb2eebc4) more than the ones on SI webpage...

In current state the Andromeda's output is uprocessed RAW data from sensor through USB2 only(no FireWire).
All the corections/demosaics should be maded in software later.However you can set on the "Magic Box" to output 10bcc instead of 12.

Ronald,
In "classic 3CDD" sensors there is no pixel shift-I'm very curious about that XL2 documented "CCD Pixel Shift, horizintal".
At least I beleave the shift in DVX100 is because Panasonic have plans to release a consumer HD camera using DVX100
as a base and ONLY implement the demosaic algorithm in the DSP chip...

[Régine Weinberg]

Well all XL did have pixelshift to do something like 16:9 as I do know, it was mentioned since the first XL 1 that this is not a true 16: 9 biy it has been theremaybe Cannon will do something like Pana, but Cannon is all the time a closed box

http://consumer.usa.canon.com/ir/con...odelid=7471#f0

they do it only with the green snsor, with the GL's as well and the XL2 to

[Wayne Morellini]

Yes, Canon used this to lower the number of pixels needed and increase low light ability. Also this fits the way DV pixels format are recorded. I do not know it's performance compared to straight three chip image quality. I defintely am interested in it compared to single chip bayer, because the pixels collect RGB light for all pixels and do not suffer Bayers throw away 2/3rds of the colours for each pixel mentality.

Realistically (especially on 4:2:0 imagary) you are going to get better HD range and low light ability, even when using SD progressive chips. So this fits 720p and HDV 1440*1080 prety well.

[Juan P. Pertierra]

Hello everyone,

The green channel CCD contains a great deal(~half) of the RGB image information. If the XL2 has ~only~ horizontal shift, then the resolution can be doubled only in the horizontal direction, which wouldn't be great if we can't get all the vertical CCD real-estate to fire.

On the DVX the green CCD is physically shifted in both directions.

The pixel shift can be used to increase the resolution depending on the physical arrangement. In the XL1s for example, the ~280k pixel CCD's are used with pixel shift to get a higher resolution image.

On the DVX the resulting image resolution is past 720P just because of the size of the sensors.

[Wayne Morellini]

Had a look at globalmediapro, and the price of the XL2 is way cheaper than in Australia, where it is so expensive to make it rediculouse for a prosumer SD camera.

If what you say is true, it is exactly the same as Canon did to get 270Kp to act like 410Kp. Actually their patent should reference any older patents the incamera pixel shift decoding technology is based upon. A good place to start.

A cheaper way to do all this is if somebody could hack the firmware of the cameras to download the raw through the 400Mb/s firewire. Possibly even SD cameras may be enticed to give up a bigger frame.

[Juan P. Pertierra]

Wayne:

You cannot 'hack the firmware' on any of these cameras without HEAVILY modifying the board and desoldering/soldering components. Most of these cameras use programmable logic which is DESIGNED to protect the IP of both the camera manufacturer and the IC manufacturer.

If the camera does happen to have a feature to do something which is 'hidden' then that's something different. But making it do something it wasn't designed to do will always take modificiations. And what you're talking about will take heavy modifications. It's a million times easier to just take the signals out and record them.

Juan

[Wayne Morellini]

As long as it is within reason it is viable. I imagine that there is methords for service centres to upgrade the firmware to repair bugs, it can almost be garranteed on all but he cheapest cameras (in matter of fact I know for certain that this is possible on at least one camera). Firewire will take it. If we look at the security taken on Playstation 2's (and many consoles) to prevent hacking, some mods were so simple as rewire a few traces to bypass the security, or find a software methord to run programs, I don't imagine cameras are normally going to be any more difficult. I have also heard of one of these cheap cameras running 3D games off the LCD, after being reprogrammed. The biggest problem, not having a program manual, is working out the memory map/timing, and obtaining different specs for the different parts. The existing firmware should show you how much of it works. I don't mean to be direspectfull, but I think it can work. But in the end a software upgrade could be sold for $100-$200. I guess I am just getting impatient waiting for variouse projects/cameras to come through, and am open to cheap options.

Thanks for your comment

Wayne.

[Juan P. Pertierra]

Wayne,

I really understand what you're trying to say, but I think you have a fundamental misunderstanding about how at least the DVX100 works. It is not a playstation, it's very different. There are entire set-hardware stages between the raw data and the firewire. You'd have to literally build a new circuit, and it would so difficult that, like i said, piping the signals out is an easy task compared to it.

If the whole camera ran off a microcontroller or microprocessor, then that'd be different. but it does not. What you're saying would involve changing the circuit VERY significantly(i'm talking 100's or 1000's of jumpers, external IC's and circuitry, SMD soldering galore), that if you had those skills why not just build your own camera?

Juan