4:4:4 12-bit Uncompressed DVX100

[Laurence Maher]

Juan,

There are many rumors ciruculating that by August the Canon XL-1 HD version will be on the market. Do you think your mod will be applicable to that camera?

Also, your mod will work with a mac editor right?

[Justin Walter]

Of course, Juan can do anything. Now all we need is someone to sneak into Canon corporate headquarters and steal the specs so Juan can start planning his attack strategy.

[John Cabrera]

Juan,

Please trust me on this... the pixels are not NTSC. Do the test for yourself. In order for the image to match up proportionately with the DV version (and I don't mean the DV version after NTSC aspect has been applied, but DV 720x480), it must be compressed horizontally by about 94.7. Then that image which will have the proportion of 720 x 480 will need to have NTSC applied. Do you want me to show you what I mean? I can save a few different JPGs and email them to you.

There is something different going on here with this camera in terms of encoding to DV and I'm sure the secret of the unaligned cannels is in there somewhere.

John

[John Cabrera]

Also Juan,

If this green screen test is going to serve the purposes of the channel test as well, please place a few more object in the scene and have some detail close to the right of the screen and some close to the left.

John

[Juan P. Pertierra]

I dropped the images in photoshop, and i got the exact same results as you, John. You're absolutely right. The lens is in fact doing something similar to an anamorphic adapter but in the horizontal direction. It probably does that in order to better use the surface of the CCD. Even though the image has to be squeezed to get 720 pixel across, you still get a sharper image in the horizontal direction.

There is still a big advantage to getting the full raw image, since any post production processing probably yields better results when applied to the cleanest, largest resolution image available, even if it is going to be squeezed down to NTSC DV frame size.

I guess the secret to the DVX's sharp image lies in this fact. They optically stretched an image to fit a wide CCD sensor, and capture more horizontal detail.

I will start the green screen test now, and place some more reference objects around. It's still light out so i have light coming from the sliding door, but if i can't get good results now i'll do it in a few hours.

Juan

[Juan P. Pertierra]

Greenscreen Test
~~~~~~~~~~

The green screen is sure much better than the blue one. Once again I did DV and RAW captures of the exact same thing.

In addition to the original object, I added two spray cans on the edge of the visible DV frame to add a bit of reference.

I also made the objects darker such that the DV frame is not clipping as much as the previous one.

I think the results are great...i was able to isolate the objects in the RAW scene perfectly, the only exception being the few speckles in the green screen which is to be expected. Once again, the speckles are only a by-product of the experimental setup.

Here is the 4:4:4 10-bit RGB(in layers) frame:
http://expert.cc.purdue.edu/~pertierr/green0.psd

Here is the DV frame:
http://expert.cc.purdue.edu/~pertierr/green0_DV.tif

[Randall Larsen]

I found a place on the web where you could buy a service manual for the DVX100. Perhaps the arcane design of the DVX100 is not disclosed in full; however, I think the service manual might be a good place to look to see if there are answers.

Of course John Cabreras detective work and Juan's diligent development is appreciated.

[Thomas Smet]

oh my god is this thing amazing. Good job on the green screen test Juan.

I found it interesting that right now to me it seems like no pixels are cropped on the left of the image. The D on the can on the left seems to be in the same exact place in both images which means only pixels are cropped on the right.

[Thomas Smet]

I think the distortion is even worse.

I aligned both images and got everything to match on the left side. I still have not figured out the exact distortion to match the right side but on top of that the can on the right was 3 pixels to low when the left was perfect. It appears that both dimensions of the image have a little distortion.

I tried the 94.7 image scale and then cropped 10 pixels and that didn't give a perfect match.

[Damion Luaiye]

Hey all - been rooting for you from the sidelines.

Wondering if it's the right time to ask about the possible dimentions of this mod? mount sizes? location of controls and cable ports?

Randall - not sure what your online source is, but I just ordered a DVX manual from Panasonic Services Corp (800 833-9626) - $57.57 w/ S&H.

- Damion

[Tim Brown]

http://www.uemforums.com/2pop/ubbthr...&fpart=1#41828

It's a link to a post by a former Panasonic engineer who was lead on repairs for the DVX, who is advertising his services. If there is no resolution to the alignment problem, maybe he could offer some advice?

[Juan P. Pertierra]

I've had the service manual for about a year now, and there's nothing that helps. The thing to note is that it is a ~service~ manual, and they only give you the bare info to make adjustments relevant to the DV functionality. How the lenses are setup and how the CCD's are implemented is part of the design and irrelevant for the use the camera was intended for.

Like I said before, they even left out some stuff that I had to find out on my own.

[Randall Larsen]

Juan,

Green screen looks great.

Sorry to hear the service manual is not more helpful on the alignment question. Perhaps Tim's suggestion you seek advice from Fox Valley Video Repair is in order. They might know something that would be helpful.

Damion:
thanks for supplying the telephone no. for ordering the manual. My source was one of those manuals for everything services.

I am confident Juan can reengineer the DVX100 as required but it sure would be nice not to have to figure out how it all works from scratch.

ALL:
I am facing a perhaps equal reengineering challenge.
I bought an Ikegami HDL-37 High Definition camera head w/o a ccu. I am hoping the service manual will tell me at least what pulses are required to drive the head. I don't want to reengineer the pulse electronics to drive the ccd. The head needs
HDTV Horizontal Drive with timing pulses and Vertical Drive and power.

This Ikegami camera head provides only analog RGB. So I have to build an A toD or use AJA's HD10A to get HD-SDI out.
I am considering HD-SDI and Gigabit ethernet chips from National Semiconductor.

Does anyone know of a chip that makes HDTV tri-level sync or just Horizontal and Vertical drive?

[Stephen van Vuuren]

Nice Green screen shots:

Several weeks ago, we discussed drives for capture. I thought this article was interesting in getting great performance from cheaper drives:

http://www.pcworld.com/news/article/0,aid,116076,00.asp

[Juan P. Pertierra]

Well, I just took the plunge and ordered an anamorphic adapter from B&H. I figured this is something that I was going to use anyway, i shoot everything 16x9, and it will be a must for testing this thing.

I will post anamorphic RAW frames as soon as i get it.

Right now I am concentrating on the easier of the two connection methods, capturing to a PC/Mac. The direct-to-PC application encapsulates some functionality required for the drive application, so it's a good place to start testing.

Everything is standard OHCI, so you don't need a driver but rather just run the program I have written with the camera hooked up. You can see a preview of the image, and capture directly to any drive on the system. My internal powerbook drive will probably not be fast enough, but we'll find out soon enough....

Juan