Oh yeah, and that is with the anamorphic, but it also has the overscan area since it is the full CCD, so you will actually get something like 1.80-1.85 aspect ratio.

Juan

Ah yes, Juan I forgot about that test you did. It's been a while. Thanks for reminding me of it. It looks really good aside from the white balance issue, which makes it difficult to compare dynamic range.

Interesting about the optimal white balance. I hope you figure it out. Sounds promising.

Hi,

Something odd is going on there. Is it just that the blue channel is running into clip much faster than the others, or is there something odd we've all overlooked?

Phil

Ive been away for about 6 months. Can anyone post a quick report on the progress? Last I heard was that Juan had captured a single frame from test jumpers on a chip. Any working models yet?

74 pages...sheesh!

John, I believe Juan is working on a web page that should answer all of your questions, now I'm not sure exactly when he will post it (he's trying to get the info accurate before he releases the website to the public) but hopefully it will be in the next couple days, I think september 1st would be a good goal...

In case that doesn't happen, the results are that juan has now captured a sequence of frames and almost completely elliminated the noise in them. He has re activated his white balance and the main thing we're all waiting and discussing right now is how much lattitude one can get with the camera. (is it plus 4 stops or plus 5, and how much did it have to begin with, 6?)

Hopefully that brings you up to speed enough until www.444yourdvx.com (or whatever it will be called) is up until then we just have to stick with www.waitforyourdvxtobe444d.com

Okay Juan, now WAY WAY way back on like the first few pages, people thought the DVX could make a high def image. I remember you saying it can't, except maybe with some pixel shift. The DVX and the XL2 have horizontal pixel shift. How would I be able to use there pixel shift to make a higher res image using your mod?

The DVX already uses the pixel shift so you cannot use it again. Juan has shown us some images captured however that were blown up to 2k images and they look damn good. A program such as s-spline pro can make it even better. I however have used just Photoshop and even went up to a 4k image size with it still looking pretty damn good although a hair soft. Just having raw RGB pixels alone make the blowups look very good. It is sort of like if you have a 2 megapixel digital camera and it says you can get 8x10 photos from it. Well those 8x10 photos would be less than 150 dpi which isn't as good as getting a 4 x 6 at 300 dpi. However because digital cameras use RGB pixels the blowup looks much better that if you were to try and do it with a video frame. Juan's mod is basically allowing us to turn the DVX into a 1 megapixel digital camera that shoots 24 frames per second at perfect raw RGB.

So to answer your question you can blow up the footage to be 1920 x 1080 HD rez and it will look great except for a slight softness compared to actual HD footage. Some may actually like this softness because it makes the footage seem a little more film like instead of HD like.

I second Thomas' statement.

Up-rezzing the full quality output of my device will simply give you the BEST possible enhancement without actually upgrading the sensors or other hardware. The only limitation is how good the up-rez algorithm is, and S-Spline is pretty darn good.

The full color sampling and high color precision transposes to the HD up-rezzed image. Remember that most of the HD footage we see is decimated, just like any footage that originates on a compressed format. Even the tape outputs on the high-end HD cinema cameras is decimated.

By starting with an undecimated SD image, you can obtain color comparable to a HD image that originates on HD chip(s). The downside, of course, is detail. S-Spline does a great job at enhancing detail but detail is simply limited by the resolution of the sensor.

My experience so far, however, is that S-Spline generates some awesome HD images when based on RAW images from my device, and it even introduces some subtle grain-like noise in the up-rezzed image which looks remarkably like film grain.

I've been looking without success for a location with a film printer, such that I can print a couple of frames to projection film and see what that looks like.

Juan

Juan,

If you want a film-out, I would recommend putting the frames on a CD, taking the disc to a photo finishing place and asking them to make slides. I am pretty sure most places can do that.

Would that work for what you are trying to accomplish?

Hi,

I think everyone's overlooking the fact that the DVX-100 is a 4:3 camera and under most circumstances you'd want a 16:9 image, or thereabouts, for your filmout or HD presentation. Not to say that this won't look a lot better than a DVX would normally in any case, but it's still a 40% cutoff. Might be time to reach for the mni35 with anamorphic primes.

Phil

<<<-- Originally posted by Thomas Smet : The DVX already uses the pixel shift so you cannot use it again. Juan has shown us some images captured however that were blown up to 2k images and they look damn good. A program such as s-spline pro can make it even better. I however have used just Photoshop and even went up to a 4k image size with it still looking pretty damn good although a hair soft. Just having raw RGB pixels alone make the blowups look very good. It is sort of like if you have a 2 megapixel digital camera and it says you can get 8x10 photos from it. Well those 8x10 photos would be less than 150 dpi which isn't as good as getting a 4 x 6 at 300 dpi. However because digital cameras use RGB pixels the blowup looks much better that if you were to try and do it with a video frame. Juan's mod is basically allowing us to turn the DVX into a 1 megapixel digital camera that shoots 24 frames per second at perfect raw RGB.

So to answer your question you can blow up the footage to be 1920 x 1080 HD rez and it will look great except for a slight softness compared to actual HD footage. Some may actually like this softness because it makes the footage seem a little more film like instead of HD like. -->>>

So, your basicly saying that when a camcorder like, lets say, an XL1s uses pixel-shift it then uprezzes it to 720x480. that doesn't quite make sense to me. But then again I never quite understood how shifting the green CCd 1/2 a pixel to the right thus increasing its sampling points would allow a CCd with 270,000 pixels to provide a DV image. If anything I kind of assumed pixel shift would sort of work as and uprezzing thing.

Phil&Mark

Phil:
You can always use the anamorphic adapter, which is what I am using. Since you can get the full width of the CCD, you actually get a ~1.83 ratio out of it.

I am working hard on re-designing the box and choosing the production circuitry such that both the box and mini35 can be mounted at the same time.

Does anyone have a picture of the ~bottom~ of the DVX with the mini35 on it?

Mark:
The best way I can explain pixel shift, is by looking at how single CCD color cameras work. Each pixel sensor has a color filter on it, such that the raw image is not a black and white image with absolute brightness(Y) for each pixel, but rather a mosaic of colors, with the magnitude of a specific color content for each pixel. Adjacent pixels of different colors can then be used to approximate the color of that pixel.

To the best of my knowledge, pixel shift is this same idea but with 3 CCD's...by simply offsetting CCD's slightly from each other, you can have a pixel element of, say, the green channel capturing a detail which lies in a element division of the other CCD's. The data can be combined into a larger image, with better detail, although as you might expect the color rendition is slightly affected because technically each element of each sensor doesn't capture the exact same detail color.

Now that I lay it out like this, it would be interesting to somehow have an up-rez program like S-Spline take 3 RAW RGB layers and take CCD shift into consideration. What i've been doing so far is aligning the layers, but if there is indeed intentional CCD shift in the DVX, that info can be used to create ~even better~ HD images.

Juan

Hi,

Please excuse the repetition, but I'd like to ask again what you are doing about a filesystem on the hard disk recorder. Does your device comprise a microprocessor system running a full-blown operating system or have you implemented a common filesystem in firmware? I would be impressed if you had!

Phil

I'm using FAT32, in configurable logic.

Hi,

I am duly impressed.

- Phil

Hi Juan
Mini 35 compatibility really is a must I think.
It's well worth the hard work it's putting on you, and I am sure many
will bless you if you succeed.
Can't wait
Take care
Emmanuel

Well, i've done some significant changes, and the box should now fit snuggly behind the mini35 adapter. I haven't played with a mini35, but I'm assuming that using the tripod mount is out of the question, since any vertical offset may cause the mini35 to not align properly with the lens? Does this make sense?

I'm going to have to figure out some other way of attaching it to the bottom of the DVX.

Juan

Yes it does--did you get the picture of the Mini35 setup I sent you Juan? So, just curious--by "behind" the adaptor, what does this mean exactly?

Charles:

Yes, i received the picture and it was very helpful, sorry i didn't get back to you.

If you look at the original box in the picture i posted, and imagine splitting the box in half along the length of the camera, the new box occupies what would be the rear half, slightly behind the tripod mount.

Juan

Out of curiosity, I've started a discussion thread:

4:4:4 Uncompressed 1080i HDR-FX1? (http://www.dvinfo.net/conf/showthread.php?s=&threadid=31571)

I was playing around with edge enhancement on samples Juan has posted so far and I have to say, this stuff has got some juice in it! The uncompressed images can literally be sharpened into oblivion after uprezzing and they still look great. I posted a sample of Juan's cat for your viewing pleasure here:

http://www.hot.ee/whiteroom/sharpening_comparison.psd

This is a 3-layer .psd file. I uprezzed the original image to hi-def, but cropped it in the end to keep the file smaller. Comparison to original size included. Mind you, this is an old example, so pay no attention to the speckles. You know they'll be gone in the final version.

Cheers,
Chris

I don't remember if I posted this or not:

http://web.ics.purdue.edu/~pertierr/cap11_HD.tif

I showed it to a senior proffessor who does a lot of 4k graphics and HD work. he was very impressed with it, he couldn't beleive it came from a DVX100. He has a macbeth chart and is interested in doing some proper green screen tests in the next few weeks.

There have been several changes, but we are now ready to assemble a final production version of the device. We've done our best to keep costs down for the initial version. Some features such as SDI are a bit costy to implement, so if there is a great demand for it, we will provide a new version or add-on module.

The final design uses USB2.0 instead of Firewire800. It will still work directly to a drive, but USB2.0 allows for a smaller device(and thus compatibility with the Mini35), and it also allows for the use of a firewire800 drive when hooked up to a laptop. Before, the firewire800 port would be occupied with the device. it makes more sense to use USB2.0 because it has enough bandwidth to handle the full quality data. Drivers will be provided so you can use your laptop as a preview screen while you record. The device also has a VGA output port so it can be hooked to any screen which supports 800x600 72Hz. We are using a tiny flat panel from a POS system.

Juan

so when can we see the web site.

Out of curiosity Juan, what is the maxium resolution you are getting from the DVX100?

773x494 pixels per frame, 36-bits of color information per pixel.

Only 493, would the PAL version get 540?

I was wondering, like the Canons and new Sony, is it possible to offset the CCD's to get interpolated 854*480 picture to match the widescreen flat TV's (using anamorphic adaptor)? Or further to archieve interpolated square pixel resolution equivalent to widesceen 2.39:1 format (what would that be 480*1148, 540*1291 PAL)? Once it went through upscaling film conversion it would be perfect for cinema production.

Without modification, you can't move the CCD's for a specific configuration. However, in the DVX the CCD's are already offset, i beleive specifically the green sensor is offset somewhat. My software doesn't yet take this into consideration, but it is very probably that I can get an even better HD image by somehow using the pixel shift.

Juan

Juan, is it possible to flip the image inside the mod? Many people are building their own Mini35 systems and these usually project a flipped image on the CCD. I am building one of my own for steadycam work, because the PSTechnik adaptor is just too heavy.

Pretty please with sugar on top,
Chris

Yup, it already does that.

Juan... many moons ago you toyed with the idea of putting a cooling device (a'la astronomy) behind the CCDs to reduce thermal noise.

Are you still toying with that idea in the future or do you have your hands full w/too many other things?

Hi,

Depending on the mechanical construction, it could be as simple as strapping a peltier stack on the back of the block carrier and cooling it. Noise would be an issue, as would battery consumption. However, this may be worth pursuing, particularly in light of the poor low light performance with this modification - well, no poorer than it ever was, but you get my drift. I'll go out on a limb and bet that at least the two LSB of that 12-bit data is noise, and probably three or four.

Phil

Hello there !
Sorry Juan and everyone, but I am way too lazy to search the whole thread
in order to find out if DVI is still in or not.
I wish it is, cause those new Apple displays sure are tempting monitoring creatures.
Keep it up
Emmanuel

If you cool the CCD's be carefull of condesation build up on the cooled surfaces. PC's were having that problem with their peltier coolers, the water would short out parts of the board, I don't know how they over came it.

<<<-- Originally posted by Juan P. Pertierra : 773x494 pixels per frame, 36-bits of color information per pixel. -->>>

At that resolution, if my math is right, you're look at about 1.6 megs per frame. Even at 24p, that's nearly 40 megabytes per second.

Have you actually found any USB2 drives that can actually sustain that? Even the best USB2 drives I've seen have only sustained about half that in benchmarks.

Has there actually been any testing of "full motion" capturing instead of single frames?

I have done full motion capture directly to a Western Digital IDE drive, i listed the model number here before but it was nothing fancy, i bought it at circuit city.

Whether a specific USB2.0 hard drive enclosure supports the maximum speed of USB2.0 that remains to be seen. My LaCie drive works fine.

Juan

Hi,

It's probably worth noting here that a filesystem implemented in firmware doesn't suffer from the same problems as one that's part of an operating system - mainly that it's one of a large number of tasks, each of which is trying to claim bus and processor time. This filesystem doesn't even need to know about directories, only needs to know how to write one kind of file, and many other shortcuts which will increase its speed.

That said, I'm fairly surprised that a standard IDE drive will write 40Mb/sec, even in raw sectors.

Phil

A quick question to clear something up, if you don't mind, Juan: when you say you're using USB 2.0, is that as a camera-to-drive interface, or drive-to-computer? Or both?

I was concerned (like I could afford this setup in the first place) that USB 2.0 wouldn't be fast enough to capture directly from the camera without dropping frames. Capturing regular DV footage from my camera directly to a USB 2 LaCie drive started dropping frames if I so much as opened a window...though I suppose there are other factors at play in my situation.

What exactly will this USB connection be used for?

Over at the 10-bit 4:4:4 and HD cinema camera threads they have discovered drives that do 50MB/s sustained and even one that does 72MB/s. Problem is I can';t remember what they are (and have to go out now), but they're there. If you want to do a search, I think Juan M. M. Fiebelkorn mentioned them about a month or two back, and they were Western Digital. These are likely to be server drives.

USB2.0 is fast enough to transfer all the data. The bottleneck is when you introduce software based systems such as a computer. Phil makes the point about having only hardware interaction.

For example, my desktop computer is about 2 years old. If I tried to capture the full data in real time using standard functions i'd run into trouble. However, the experimental digital capture card I'm using has specialized drivers which use DMA and use a direct path between the PCI Bus and the hard drives. I can capture full motion up the point where my 120GB drive is full with no problems whatsoever.

About the pixel shift...it's not really a 'shift' of the CCD's per-se. It rather seems that the green CCD is placed farther away from the prism by a tiny amount such that there is a larger projected image on the chip, and the optical information that lies between elements in the other CCD's lies ON elements in the green sensor. Afaik the sensors are not moved in the x/y direction willingly.

And no, the HD images I have posted do not take advantage of the pixel shift. I actually 'eliminated it' by resizing the images to match and aligning them which will actually yield worse detail. I'm still not sure about what procedure to use to take advantage of the shift....it seems like I would have to up-rez R,G,B channels separately in a bicubic manner, align them and THEN apply S-spline.

Anybody have a suggestion?

Juan

Use the same demosaiking algorithms used for single sensor cameras.
Just need some little modifications to work with that structure.
The drives Wayne is talking about are the WD Raptor.
They are not server disks.
SATA interface, 10,000 RPM, 72 MB/s.
Two flavors , 36 and 74 GB.
Around $200 (74GB).

keep in mind that this a total guess, but... maybe in order to take advantage of pixel shift, you would have to be converting to YUV, like the DSP does (at least i assume it does). if the R, G, and B channels are somehow converted to YUV individually first, and then the Y (luminance) channel of the original G channel is combined with the Y channel of the other two, then maybe it somehow improves image detail. or maybe it's the U and V channels from the G channel that somehow are combined with the other U, V channels, though it seems like that wouldn't make sense. but regardless, i would assume that the utilization of some other color space is a part of the whole pixel shift technology thing.

i read somewhere that the Arriscan film scanner uses pixel shift to obtain a 6k frame using a 3k scanner. maybe someone science-brained can look up the patent info (if that's even possible).

juan, can you post a new raw frame straight from the camera? i have no idea if any of the other frames i have from you have had the green channel adjusted or not.

i did a quick search and found these...

http://www.cse.ogi.edu/~linchuan/demosaic/
http://www.cse.ogi.edu/~linchuan/demosaic/adobe-raw.php
http://www-ise.stanford.edu/~tingchen/main.htm

i have no idea if those sites help any, but i know they make my head hurt. maybe someone can write one of those authors a charming email and ask for advice.

Ok I Tried a search but with no result, so anybody remembers if DVI is in or out ?
I think this could be quiet important.
Take Care
Emmanuel

I forgot a little thing....
I guess that kind of setup would require the use of an anamorphic adapter unless you want to resize the image to anything near the 4:3 aspect ratio after the interpolation...
Cause if the CCD is 794x4XX you will end with 1588x4XX (anything around a 3.3 ratio)

DVI is normally output from computers and input on veiwing equipment. I thought I found an AGP 3D card with input, but apparently was wrong on that. We discussed this over at the cinema camera technical thread (I think) and I did post a DVI input board there.

Juan P, thanks for your answers.

If you look at the cinema camera thread (and the obsura cam wiki) you'll get an idea of just how much capture performance you can squeeze out of a computer in HD, which is enough. USB2.0 has a lot more cpu performance hit than IDE and Firewire), and Rob Scott is getting much better capture peformance than expected as he learns more.


Thanks

Wayne.

Yeah, I gotcha. It's the middleman that causes such problems.

Thanks for the explanation!

Juan(M.):
Remmeber that the DVX's green channel is enlarged in both dimensions, so there is pixel shift in both axes...it's not only shifted horizontally but vertically as well...so the final aspect ratio should be about the same(?)

didn't know that, sorry.
Well, if that is the case, just two possibilities are left.
I could be an amazing high resolution image, or it can be really a mess :)

I have been reading this forum with great interest. Juan's work looks very cool. However, after waiting for over a month for the website to appear I am begining to wonder if this thing will ever come to fruition.

Any ETA as to when the site will launch?

Thanks

Please do realize that this is essentially a one-man operation. Opening a company of any size is a huge task, both time-consuming and monetarily. Not to mention the rest of the tasks such as actually BUILDING this.

As of right now I am working on assembling final production hardware.

Juan