4:4:4 12-bit Uncompressed DVX100

[Nicholi Brossia]

Yes Luis, you are correct.
The image must be flipped both horizontally and vertically. Just for further reference, this is also referred to as inverting (flip vertically about the x-axis) and reverting (flip horizontally about the y-axis) or, like Luis said, rotated 180 degrees.

Juan,
Am I understanding correctly that the user will be able to switch between the normal and flipped modes? If so, this would be excellent for use with switching between 35mm adapter projects and other projects recorded with just the native camcorder.

Either way, Juan, you're doing excellent work.

[Juan P. Pertierra]

Yep, that's correct, the user will be able to decide the orientation completely, that is there are check boxes for flipping horizontally and vertically, as well as rotation up to 270 degrees.

In other news, i know nothing about color correction so this is mostly for fun. It took a similar shot of the lamp and attempted to make it look as much as i could as the film output.

Now, the main problem I had was that since I do not have control over exposure because it is disconnected, all i have is the ND filters, and one of them was too bright and the other too dark. I started with the dark one and had to push it A LOT. I basically clipped a lot of the low end in the levels and it shows in how saturated part of the image is...still if i got this far with this dark an image, once i have the camera closed and i can vary the exposure, i can get adjust the light going into the camera to be in the correct range...

http://expert.cc.purdue.edu/~pertierr/sample1.tif
http://expert.cc.purdue.edu/~pertierr/sample2.tif

Juan

[Juan P. Pertierra]

A noise free 4:4:4 10-bit uncompressed -> HD frame, with some of my own color correction...

http://expert.cc.purdue.edu/~pertierr/cap10_RAW.tif

[Nicholi Brossia]

I don't mean to be annoying, but I've noticed a bit of chromatic abberration in cap10 RAW and sample2, but sample1 has no chromatic abberration at all. It looks perfect.

Could this be due to your color corrections?

I'm sure your initial recording is perfect considering sample2 looks so great, but I figured this was worth mentioning.


EDIT: It just dawned on me that you are capturing the red, green, and blue channels seperately and then layering them in post. I'm guessing the "abberration" is probably being just a hair off on the alignment of one of the layers. This seems like a valid subject though, so I'll leave the whole post here.

[Juan P. Pertierra]

what kind of chromatic aberration is it? if you're speaking of something present in the histogram, it wouldn't suprise me, because the frame was put through S-Spline pro for up-rezzing which could have some unexpected effects.

All I did on cap10 other than uprezzing was some level adjustment.

[Nicholi Brossia]

I'll use sample2 as an example. I've opened the file up in Photoshop. Even at 100%, I notice a very thin green line around the base of the lampshade. If I zoom in to 200%, it shows about a 1 pixel "shift" of green to the bottom, and red to the top. After seeing that one part, its apparent throughout the rest of the image. I figured the RGB layers just aren't aligned right. Could this be correct?

As far as uprezzing, both cap10 (uprezzed) and sample2 show the same abberration effect. I guess whatever these two have in common, but different from sample1 will be the culprit.

[Juan P. Pertierra]

Yup, that's alignment...try one of the other Photoshop PSD files that i have uploaded, and see what alignment you can come up with by moving the R,G,B layers...it's tricky! :)

Juan

[Nicholi Brossia]

Yeah, that is tricky. Good thing you can program the computer to do all the precision detail work, because I know I couldn't :).

[Isaac Brody]

Juan,

I noticed a few stray noise specks in Cap10_raw. I was adjusting the levels and they cropped up when I brightened the file. The specks are green and appear in the sky. Do you see them?

[Juan P. Pertierra]

Yup! I haven't gotten the new cap yet, i've been working on finishing the software to put out a clip...

Juan

[Juan P. Pertierra]

Intersting things first...i have good reason to think that the CCD's are actually 811x495, however several columns seem to be covered up for some reason,(black level adjustment?). Given the barrel distortion of the lenses with the current configuration, it probably would work well even if there was a wider image available, but it'd be nice.

I am sooo close...i've got the program spitting out clips perfectly, except for one thing. There is a somewhat periodic shift in the raw output frames of 0 or +1 rows, such that right now the video seems to be shaking slightly...i'm trying to figure out how to recognize the shifted frames so I can compensate for this, and then i'll be done.

Except for the slight shakes, the video looks awesome! Pretty CPU intensive though!

[Randall Larsen]

Could the shakes be due to vertical subsampling?

Dalsa has an application note on this AN07.

This is my first post to DVi; however, I have been following this thread for a while.

I would like to add my name to the waiting list for Juan's modification and software.

It will be great to escape the limitations of the DV format. I don't think these sort of limitations are what Orson Welles had in mind when he said "The enemy of art is the absence of limitations."
[thanks to Isaac Brody for quoting this].

I worked with Orson on some projects from 1978 on. He often remarked that Black and White movies were better than Color movies and Radio was better than movies. Art is often in what is left to the audiences imagination.

Jaggies and poor chroma due to the limitations of DV are not what Orson had in mind to promote art. He shot his TV projects on 35mm film because he didn't like the limitations of 16mm or video.

Excuse my rambling here (somewhat off topic):
Now that Juan has almost solved the problem of outputing raw files to disk from the DVX100. There are still some remaining opportunities for research. JVC will release a 3 chip High Def
box camera the 870 in July. It is said to have dual SDI outputs.
However for effects and compositing this camera would benefit from a raw output. Juan if you are willing to take up the gauntlet here may JVC would even loan you a camera!

There is talk that JVC will furnish raw outputs on their mid-price pro HD camera to be released in 2005.

The new JVC cameras use a 2/3" cmos chip from a Rockwell spinoff. [Development kits cost $10,000]. What is the size of the
DVX100 chip 1/2"? This brings up another annoying problem with DV camera's they don't have 35mm depth of field characteristics. PS+technik has an expensive solution previously mentioned in this thread.

I used to have 1960s RCA TK-41 image orthicon cameras that had a 35mm image area. They used motion picture lenses. There was no moving ground glass. They did make use of a field lense to bring the image (snorkel fashion) back to the image orthicon tubes.

Perhaps I don't know optics well enough but I would think a field lenses could incorporate image reduction and perhaps an increase in "speed." by concentrating the light. Now maybe that wouldn't preserve the depth of field relationships? The problem with mini-35 is that you lose two stops.

DALSA has a 35mm image area CMOS sensor. To me using a sensor that has 35mm image area would be ideal. Now if different parts of the sensor had to be driven separately and the image assembled in a frame buffer that would be OK.

I also understand that ARRI is working on a 35mm image area
sensor.

These solutions will cost big bucks! In the meantime Juan has enabled HD capable results from a $2500 camera! Bravo!

Incidently my retro solution to avoid DV jaggies has been to shoot 600 horizontal res. analog NTSC on a sony BVH500 1" machine and to record the 9bit output from an ampex Zeus TBC uncompressed to the G5. This doesn't avoid the cross color and dot crawl problems of NTSC but i don't get Jaggies. I am thinking of recording rgb on three machines to outdo betacam sp.

I can't wait to get my DVX100 modified. This is definitely the way cool solution to avoiding DV artifacts!

[Jon Yurek]

i have good reason to think that the CCD's are actually 811x495

I think that this and the fact that you're getting the image shifting problems point to the possibility that not all of the CCDs on DVXs are the same size. It's entirely possible that there's variation between cameras of the same model and to cover for that, the manufacturer purposely made them larger than necessary. Your capture software may have to be adjustable and (maybe?) auto-detecting of how big the frame is.

[Stephen van Vuuren]

Lots in interesting stuff:

Great idea about adding the image flipping problem of the 35mm adaptors to this. I agree that a combo of 35mm adaptors and 4:4:4 (there's got to be a good name is their somewhere) would provide a very non-DV looking image.

And it would make far more sense to drop the high price for buying or renting the prebuilt 35mm adaptors using Juan's cam then compressing the image back through DV codecs.

Juan:

I recall reading somewhere (perhaps in Steve Mullens article on the DVX100) that pixel shift was employed. Unfortunately, I don't remember how or why, but it was a variation on the pixel shift in the XL1 CCDs.

Randall:

Welcome aboard. I just recently got the restored Citizen Kane and how can you not love the cinematography? Also, our local filmhouse is actually doing a Welles retrospective though no word on what prints etc..

Welles could be a whole topic from this thread - but I agree with your point that he would have not been a fan of the DV codec or any artificial limitation. The DV codec was great to break off from the generational, non-digital aspects of previous video. But we've rapidly gone from the days of having to carefully tweak a machine to capture DV data to easily being able to capture HD or uncompressed.

I'm hoping Juan's efforts will help hasten the end of all these lossy, artifact inducing aquistion codecs.

[Randall Larsen]

Stephen and Listmembers,

Thanks for the welcome. Juan's hope of providing a $50 mod for the DVX100A to allow raw output or 12 bit processed output is awesome. We all know we would be willing to pay $1000 for this.

Juan are you still using the Adlink PCI-7300A. I am thinking of trying to duplicate your experiment. However the PCI-7300A carries a list price of $880.00 without cables or probes.

This device has a maximum clock rate of 20MHZ. So it won't handle the data rate of 75MHZ on the JVC camera I am interested in tapping in the future.

I assume that you started with Tektronix klip-chip probes. Did you
use the ADLINK Scsi terminal block to connect to the Adlink?

Altera has a logic analyser probe for the Apex 20K but I am sure
the kit is pricey.

http://www.altera.com/literature/ds/dssignal.pdf

On the other hand it might allow some reverse engineering of
Panasonic's program if that were necessary.

This probe probably is too big to fit inside the camera?

I guess I would like you to tell us where you are getting your APEX20K probe made. I can understand why you might not want to reveal that at this stage. However, I am not too shy to ask.

It might be interesting to make a "open source" project out of this. You might actually make more money writing books and furnishing support to users than you will selling mods. Just a thought.

As someone pointed out earlier. Manufacturers are not going to be too happy when they see $3000 cameras out performing their $100,000 SDI, DVCpro and HDV models.

For recording, Blu-ray DVD disk recorders might be a good option to hard disk recording (archive copy). Of course 2 gig and 4 gig solid state cards are now available from panasonic.

A 2 gig/sec Fibre Channel interface might be nice if it wasn't so expensive to implement. I would be interested in making a dual head 3D rig that might require higher than firewire 800 bandwidth to connect. I guess two firewire 800s would work.

off topic comments:

Stephen, Hope you enjoy Citizen Kane. Welles probably takes too much credit as auteur on that film. He had some great collaborators. I believe it was Greg Toland who shot and lit most of that picture. Welles was a perfectionist though so if he didn't like the lighting he would have sent everyone home till it was fixed. He certainly wouldn't have put up with DV artifacts!