Human Eye
 

Quote:

Frank Roberts Said:

"Can you flash foward into the future where cameras will be able to capture latitude similiar to the human eye?"


Erm.., the humam eye is a vastly complex organ, and the eye and brain (Vision Centre) operate togther to "percieve" images, and does raise philosopcal questions that have troubled man from the dawn of history ---------"To Whom does the eye inside belong"

But that aside as a crude compasion the resolution of the eye is aprrox 72 Mega Pixel

from http://clarkvision.com/imagedetail/eye-resolution.html

Human eyes have about a 1,000,000,000:1 absolute range from fully adapted dark vision to fully adapted snow conditions at noon on the equator. But when we are adapted to a normal working range we see about a 30,000:1 range of brightness values.

from http://www.digitalsecrets.net/secret...micRanger.html

Juan,

The picture really made me want to see a moving clip.
Is there any chance of posting a 5 - 10 second clip?

Thanks,

Luis

Sure, i'll take one tonight.

Note that the output is still 10-bit RGB because my software is flaking out on me :P. I'll also have to make the frame size small like with this last frame to save space, my account is almost full.

Can't imagine what it will be like at 12 bit.
I just put the frame grap into after effects to even out the gamma curve (seeing as it seems so dark initially).

I was able to pull all the detail out the shadows without any visible noise addition.

The incredible thing to me was, when viewing such a shot on the histogram, that after pushing it as much as I did, I saw no signs of banding.

Usually, with 4:1:1 DV footage, you would see large gaps of missing information on the histogram after pushing the levels to much.

I can't wait to see a sample clip.

Very exciting.
Keep it up Juan,

-Luis

Juan, when you do the clip, any chance the footage could involve panning in a really bright, contrasty area? Just curious. Best- Frank

Sure, that last frame was taken indoors with two work lights. It's not optimal lighting, but it gets relatively high contrast images.

And luis, no that is not a urine sample <g> that's where I keep all the DVX screws and small parts while i'm working on it :)

Juan,

Do you happen to remember what fstop you were at when you took that still?

And what kind of wattage are those worklights?

I'm trying to get an idea of the sensitivity of these chips, and what kind of light we'll need compared to shooting DV with the DVX.

Thanks,

Luis

Sorry to post again so quick.

I was looking over the still frame again (still shocked as to how well it reacts to some color work) and it looks like you might still have some noisy pixels.

I see a small group of them above your laptop monitor, near the top in the center of the frame. They weren't completely apparent at first, but as I played with the hue in the frame, they became pretty obvious.

-Luis


EDIT:

I realize now that the easiest way to see the pixels is to turn the color saturation all the way down, making the image black and white.

You should see the small group of black pixels against the white wall above the laptop in the photo.

Impressive
 

I just found this thread late last night and read the whole thing. This is great! The results are amazing! I thought it wouldn't be possible from disscusions that started when the DVX came out. Then when I thought I heard Jan say it only did 4:1:1 off the DSP I lost all hope. Guess I heard wrong.

I am really excited at all the possibilites of the new DVX filmstream. If those errant pixel problems are taken care of, it will be perfect! 2:35.1 will look so much better too.

Now I've started my plans to build a box that will hold support rails, a hard drive, PCB and an adaptor to power it from another Panasonic battery.

Luis,

My software was bugging out so that frame was captured using the old method, that's why there are some speckles present...i didn't notice them when i captured it though...

Juan

That's good to hear. For a minute I thought you were back to
square one.

I didn't notice them right away either, probably because they
were up against the white wall. I've scrutinized the frame, and
haven't found any other speckles in there.


One other small request. When you have a moment (which I'm
sure you don't have many) could you give us a quick summary of
the different capture options that will be available. So many have
been batted around, I just want to make sure I know which are
staying in the design and which are going.

(i.e. will SDI still be implemented? etc.)

I think a quick list would help a lot of people.
In fact, we may want to start it out in a new thread, seeing as
the camera prototype is now completed. Perhaps the new
thread could cover the contruction of the box, and capture of the
footage.

This thread seems to be getting kind of hard to search, and it
seems people (myself included) keep asking questions that have
already been covered.


-Luis

Juan,

About 3 weeks ago, you said that you were going to redo the 35mm to DV to RAW comparison. Any progress on that?

Speaking of questions that may or may not have already been asked heres one I think is VERY important for this project.

Can we see screen grabs both from a normal DVX100 and a uncompressed DVX100 AFTER its been thru DVD's standard MPEG2 compression?

Dont get me wrong, Im a perfectionist so uncompressed screen grabs are great to look at but what I would really be interested in is see what the difference is THE REAL world application that most of use will use it for - DVD.

Film outs are totally different story and no doubt uncompressed footage would look much better...but Im also a realist and know most people arent doing film outs all the time. All numbers and specs aside Im very interested in seeing what is can do for 99.9% of our projects that are going to DVD.

Brett,

I don't know that that test is really necessary.
Most people would already agree that uncompressed 10 bit
footage looks better going to MPG2 than DV footage does. I've
done both and I can tell you that DV looks worse everytime. Of
course it all depends on the compression program you are using,
and the skill of the person doing the encoding, but you'd be hard
pressed to get a cleaner encoding from DV.

You can try it yourself with the stills that have already been
posted from Juan.

There is a reason most people master onto D1, or Digibeta at the
lowest end, when encoding movies shot on 35mm for DVD.

Either way, if Juan posts a 5-10 second clip in the near future, as
he said he would, perhaps he can also roll a tape in the camera
so we can compare the DV and uncompressed footage. That
would be good to test how the footage stands up to the entire
post process. I guess it can't hurt.

-Luis

Brett Erskine said:

"Can we see screen grabs both from a normal DVX100 and a uncompressed DVX100 AFTER its been thru DVD's standard MPEG2 compression?"

If you only apply MPEG2 compression onto the uncompressed DVX100 frame and not the compressed frame, then it wouldn't be a fair comparison.

But it doesn't matter anyways, because the uncompressed version will look better no matter what, since it has only been through one compression (MPEG2) rather than two compressions (DV and MPEG2)

This is why it's better if we see the actual uncompressed frames.

Okay, some quick questions and then some hypotheticals.

I looked over the boards quickly and couldn't find this asked before, but am I correct in assuming, since your only taping video straight from the chip, there is no audio included in the mod, thenceforth you will always have to record to MiniDV/Firwire and resynce the PCM audio from that, into your Mod-Captured footage?

Also, I was reading some recent posts concerning loosing the in-camera effects, which is part of what concerned me previously, though it's great to hear the white balancing is still intact. I've been recently pondering over this in regards to the XL2, seeing all those neato in camera features in posses. One thing I've considered is you could always keep a record of all your settings and/or record to MiniDV for comparison.. to help your match any in-camera settings you want to replicate in post. It's not the most convenient thing in the world... although not too high a price to pay for uncompressed 4:4:4. Oh and that also brings up my question.. what activates the recording process on the mod, assuming it's separate from the cameras record activation? And if separate, can the two be synced?

And lastly, to appease my hypothetical brain... if you –could- tap into the pre-DV YUV signal and say.. put it on a wire or whatnot.. then you had a method to convert it to RGB.. say like a standard DVI output... your mod then could be applied from there, correct?

Hey Juan, I haven't posted in a while, but I'm still keenly interested in your mod. Looking forward to your clips =)

To answer Michael's questions:

1.Correct, there is no audio handled in the mod.(I gotta stop calling this a mod!) But it is trivial to simply make sure the DV tape is going, and then match the DV audio to the RAW frames. Remember, they are the exact same frames, so you don't even need a snap board(or whatever it's called), just any kind of quick motion will allow you to find two matching frames and synch them. Then delete the DV video and you are left with the audio.

2.Loosing the in-camera processing is the (LOW) price to pay for pristine data straight from the A/D's. Same thing with the Viper.

3.Not sure what you mean, but we wouldn't want to tap into the YUV data on the DVX, because it is lower-precision and a stage after the clean RGB data. My device does have a DVI output for monitoring purposes.

As far as the MPEG-2 compression comparison, if you are interested you can just grab any of the DV-RAW comparisons I've posted on this thread and compress BOTH with your favorite MPEG-2 compressor.

Like it has been said by Kin, the correct comparison is to compress both frames.

I'll try and take that clip tonight! :)

I dont want to buy a camera and then loose all the features that I paid for, 4:2:2 uncompressed is still better than DV and its better than just dropping samples. 4:4:4 uncompressed may give film-like latitude but it also means that postproduction will take longer and it will take up more disk space.

Filming 4:4:4 uncompressed is only an option.

Yea, it's true that you're going to lose some features and in-camera effects, but you can always shoot in regular DV if you want.

If I had a camera with potentials like that, I'd kick myself for not using it to the max. (even if it means I lose a few lousy in-cam effects)

As far as editing cpu speed and hard disk space, you could always use the DV tape as an offline edit source that will later be replaced by the uncompressed footage. Of course this would have to be rough edits with no color correction, but would save time and space none the less.

Also, a majority of the camera effects and features could be duplicated in post. Sure that will result in more render time in the edit bay, but sometimes you have to give a little to get a little. Personally, I feel the gain of uncompressed video captured straight to hard disk is far more significant than loosing the in camera features. So you're giving a little to get a whole lot.

Juan has done an excellent job of "unlocking" these amazing abilities of the DVX while at the same time preserving those included stock. Like Kin said, with a camera capable of this much, it would be disappointing to use it for anything less than its highest potential.

This is really amazing stuff, Juan. Would somebody pay this guy? He needs some assistance!!

I wonder whether Panny should put him on the payroll. If this guy had a budget he might be further along.

Rock on bro!

"As far as editing cpu speed and hard disk space, you could always use the DV tape as an offline edit source that will later be replaced by the uncompressed footage. Of course this would have to be rough edits with no color correction, but would save time and space none the less."

Nicholi,
That sounds like a great idea, but what are you keeping the uncompressed footage on? D1 tape?

Usually offline edits are done to save hard drive space in post.
There is no point in doing offline editing if all your source material is captured to hard drive to begin with.

This is why I think the SDI out solution is ideal. Granted, I still want the 4:4:4 uncompressed option, as it would be great for some special effects shots, when compositing or greenscreen is planned. But otherwise, going SDI out into a computer would be much more feasible with long form projects.

-Luis

Yes, you're right that initial hard drive space wouldn't be saved since the information is being recorded directly to hard drive to begin with. However, please correct me if i'm misunderstanding, after the individual .TIF frames are recorded to the camcorder's hard drive, you still have to process those frames into video, thus taking up space on the computer's hard drive. By rough editing with DV first, you will get an idea of what RAW footage isn't necessary and won't have to process or store that "extra" uncompressed video, thus saving both time and drive space.

No matter what process you take, you will always need to capture the DV tape in order to get your audio. I figure its a good idea to take advantage of this less-drive-intensive format since you have to capture it anyway.

Of course, a lot of what I'm saying is entirely dependant on Juan's program and how it functions. If there is a way to pick and choose which .TIF image "clips" to process, then this offline system could work well.

I agree, SDI would be great with long form projects, especially studio shots and chroma keying. You would only keep the acceptable clips and wouldn't have to deal with processing or capturing later. The only downside I see here is having to carry a computer along with the camcorder. Again, this is all about what setup best suits your production.

"I agree, SDI would be great with long form projects, especially studio shots and chroma keying. You would only keep the acceptable clips and wouldn't have to deal with processing or capturing later. The only downside I see here is having to carry a computer along with the camcorder."


Actually, what I was saying was that 4:4:4 12 bit uncompressed would be best for chroma keying. You could record the rest of your footage through SDI if you didn't mind the loss of color information. But, for select shots you would still have the option of capturing 4:4:4.

And, SDI would not necessitate a computer (although I wouldn't mind carrying one around). If you didn't mind the 2:1 compression you could always go out to digibeta.

I realize this still isn't as ideal as uncompressed, but let's remember we're still talking about digibeta quality out of DVX. That is still amazing to me.

This is what I like the best about the modification. It will allow for various options which can fit numerous different production environments and situations. One capture solution obviously won't work for every scenario. Hell, who knows, there may still be times I'd want to use my DVX as simply a DV camera.

It's great to be able to get anything from DV to 4:4:4 uncompressed out of the same camera, and chances are I would use every single option at one point or another.

-luis

I know this isn't related to the dvx100a but going back to the XL2. Even if we can only get 960 x 480 with a mod that is a 2:1 aspect ratio. If we want a 2.35:1 cinematic aspect ratio we now would only need to crop 72 pixels from the resolution. That would be 36 from the top and 36 from the bottom. This gives us 960 x 408 or a 2.35 x 1 aspect ratio. This would be great if we plan on going to film with that type of aspect ratio because we keep most of the resolution without cropping to 720 x 272 like we would with regular 4x3 DV.

Thomas I get the idea of what you're saying but I think there might be a slight problem with your math, er... your formula.

I think that the pixels are not exactly 1x1 but 1x.9 which means that when you shoot in 4:3 mode you're shooting in 480x720 which gives an effective resolution of 480x (720x.9) or 480 x 648 (approx. 640) which reduces way the heck down to 4x3 (or 4:3)

Now, do the same math with 480x960 you get 480x(960x.9) or 480x864 which is 1X 1.8 which is almost 16x9 (16x9 being 480x853 which is 1x 1.77)

So in order to get cinema scope (1x2.35) or cinema flat (1x1.85) you'd have to remove 112 pixels from the top (864/2.35= 368 and 480-368=112) for scope and 13 pixels for cinema flat (864/1.85=467 and 480-467=13)

That being said it's still interesting because the dv format can't hold anymore then 720x480 (or 648x480) so in order for the xl2 to record it on to minidv tape then there must be some sort of algorithm that compresses the footage on to the tape and effectively (we hope) perserves the total image quality. However, if someone were to capture the entire CCD image (or at least the ones that are firing in the xl2's case) without using ANY compression or algorithms, they would be super duper cool, and would hopefully finish up with this silly dvx100 business and get onto working on one for the xl2, before everyone decides to go HDV and we need a RAID array to capture uncompressed information off the chips.

I hope that wasn't too technical, and more so I hope that my math was right, if anyone else want's to correct me please do.

I'll correct you. The business with the DVX100 isn't silly. He's working his tail off and a lot of us appreciate it. I don't own an XL2.

Frank is right, it's not silly, just an attempt at humor, I'd venture to say that there are VERY few of us that own an XL2 right now as well.

Sorry, I think I overreacted. You know what's crazy, it's funny how sometimes we all get so overwhelmed that we can't even sense when someone is just trying to be humorous. Nick, keep up the humor bro, its me who should be corrected. Best- Frank

going down to 368 is still much better than going down to 272. Also the 864 rez from the XL2 would be pure pixels and not stretched from an anamorphic 720x480 so it would look much better when blown up. If you have a 720x480 image that is anamorphic and correct it your 864x480 image is interpolated. Even though it looks good a raw 864x480 image would still look better.

Thomas, yeah, I agree it's still better then the previous option and a raw image would definitely be better. I wonder if another option might be to use a P+S technic with anamorphic lenses, and then expand it further in post?

It seems like there might be different versions of this device, with SDI/DVI, etc.

As far as the editing goes, the best way is to just treat it like film. Use the DV footage for editing, and create an EDL which renders the final product using the RAW frames. The RAW frames can be color corrected in your favorite program.

Juan

Or once everything is captured you could render the raw frames and the audio track into one of the 16 bit video codecs that were discussed on this forum and then you would have a video file of raw frames and audio to work with. If you use the microcosm codec for this purpose you could even get lossless compression to save some space. Once you convert everything you could then delete the raw frames and the dv video and save a lot space.

Juan, does that mean you'll be including a SMPTE time code, for the EDL's?

Timecode and other metadata
 

Hi,

Signed up specifically to ask this question - from the way you're doing this modification, the timecode data isn't readily available to you. However, it's pretty much an absolute requirement to be able to work with this stuff in postproduction. The simplest and most generic way to do it would be to insert the timecode into the filename of each frame as a frame count; this is the way film scanners tend to do it. This way you could roll tape in the camera as well and use that as your offline. This would save a lot of messing about processing offline versions of the 12-bit data.

Also, I don't get the purpose of SDI. You want to modify a 3K camera so it needs a deck costing 15 to record? Hard disk and DVI for monitoring if possible would be fine by me.

Phil

"Also, I don't get the purpose of SDI. You want to modify a 3K camera so it needs a deck costing 15 to record? Hard disk and DVI for monitoring if possible would be fine by me."


The purpose for SDI is ease of use for many people.

SDI is an industry standard. Also, it allows for the use of a deck
which is an advantage for some who feel more comfortable
with their footage on a tape than on a hard drive.

Secondly, a deck is not necessary, as many capture cards are
available that utilize SDI. Blackmagic Design makes an SDI
capture card for under three hunderd dollars.

Lastly, SDI allows for the easy capture of uncompressed 10 bit
4:2:2 video footage as opposed to 4:4:4 RAW frames. It would
be easier to work with, and would take up much less space.
This would make it much more feasible for long form projects,
like feature work.


Either way, from what Juan has said, it is not a zero sum game.
There is no reason we can't have SDI as well as the other
ptions. So, there is no reason to use the option, but it would
be great to have the option.

-Luis

PS.
The cheapest Digibeta camera on B&H costs 44K.
So I think it's worth modifying a 3k camera to get the
image of a 44K camera.

Great, just what we need, an interface that requires an expensive capture card. How do you expect to capture in the field? Please, Gigabit Ethernet or Firewire 800. They're built-in to many computers these days and easy/cheap to add to older computers.

- ben

"Great, just what we need, an interface that requires an expensive capture card"


Again, it is an interface that allows you to use industry standards if you choose to. Many people already have equipment with SDI.

And, once again, it is an option that you do not have to use.

I don't think anyone has suggested that SDI should replace the use of FW800 or anything else.

Much like the silly 'film grain' function included on the new XL2, don't use it if you don't like it.

-Luis

Hi,

> SDI is an industry standard.

Yes, I'm aware, I use it weekly.

> Secondly, a deck is not necessary, as many capture cards are
> available that utilize SDI. Blackmagic Design makes an SDI
> capture card for under three hunderd dollars.

Hang on. You want to overlook the drive on the camera and lug an entire computer around with you, to record the image data on its.... hard drive?

> The cheapest Digibeta camera on B&H costs 44K.
> So I think it's worth modifying a 3k camera to get the
> image of a 44K camera.

Yaright, okay, you're one of those people! I really don't have time to list the differences between the DVX-100 and the DVW-790, but suffice to say that no matter what you do to a DVX-100, it will never be a 790.

Phil