Say guys, I copied this from another thread:

Not sure if this will help or not but I found a codec for free that supports 16 bit per channel as well as an alpha channel making a 64 bit video codec. It works on mac and pc with just quicktime 5. Best of all it is free. They even have a lossless codec that can get 6:1 compression with no loss but that codec is $99.00. I know it isn't 12 bit per channel but it might be an easier way for people to manage files opposed to a series of stills. Besides right now the tiff files will need to be 16 bit anyways.


http://www.digitalanarchy.com/micro/micro_none16.html

how did my post get into a different thread? For me it shows up right in this thread. I also posted a followup.

latitude
 

is the latitude of the normal dvx100 5 stops or 6?

what is the lat of the uncompressed version?

35mm cinema film is 13 stops?

thank you,

jake r.

has this thread died?


Anyways if you are out there Juan I was curious as to how hard would it be to just get 8 bit uncompressed 4:4:4 off of the mod. I have been doing some tests and except for really fine color correction 16 or 12 bit video does seem a little over kill. Whenever I convert these sample frames down to 8 bit they still look amazing better than anything I have ever seen from a video camera. I did some keys at 16 and 8 bit and I can't tell the difference. Color correction did suffer a little bit but not much. The only time you would really need that much range for color is if you really screwed up shooting and you had crappy footage.

If this isn't possible then I guess I could just convert to 8 bit when everything is captured.

In my AE work, sometime moving to 16 bit really helps banding and gradations, other times, not at all. It really depends on the source material. I agree that 8-bit is easier, but if 12 or 16 bit is possible, I would still like to have the option to start their and move down.

Hi,

Nope, not dead, i've just been working non-stop, i guess it's a good time for an update.

I have the camera closed and completely assembled for the first time in over 6 months. The flex cable slides out the bottom perfectly. I have the hardware prototype all hooked up and i'm right now just working on the software configuration of the FPGA, which is what's left to be done. Once again, all the probing done to the camera was done with absolutely NO SOLDERING.

My plan once I get it done, is to get one or two units together and lend them out to people who can try them out and give me feedback on any suggestions or added features.

About the 8-bit video, that's easy. Hadling 12-bit is actually much harder because it doesn't align to word boundaries. My hardware allows the user to capture several different color precisions to save space.

I understand what you mean about the 8-bit video. However, the problem with DV footage is not really that it is 8-bit, but more like they compressed the 8-bit values to a smaller dynamic range, loosing latitude. With the XRS-1(random name for my device) you will still get the full latitude of the CCD, just decreased precision to 8-bits when set to that mode. At full quality it will give you uncompressed 16-bit RGB color files with 12-bit RGB precision.

I've been busy with this, but if anyone wants to see any more tests, just let me know what you want to see and i'll take them.

I still have all the ORIGINAL frames from the latitude test, so if you want I can color-correct some of the RAW ones at different exposures for comparison with the DV frames. Remember, all the RAW frames in the latitude test were not color corrected, so they don't even have white balance adjustments done.

Juan

Hey Juan,

If you need people to be beta testers for your mod... I would be more than happy to donate my DVX to the cause. :)

latitude
 

So, what is the latitude of the
uncompressed dvx-100?

Thank you,

Jake Richardson

I'm probably not the right person to answer the question in photographic terms, i'm more of an engineer. But from looking at the latitude test results, it seems that there is a 2 F-stop difference between the optimal scene exposure for DV and RAW.

Can anyone else look at the images and see if you agree with this? I can post the original frames if needed.

latitude
 

Thank you Juan.

Are you saying that your uncompressed dvx-100 has an additional two stops of latitude?

'Origin's exposure latitude is comparable to the best film stocks and offers at least 12 stops of linear response for an astonishing range in the whites as well as the blacks.
http://www.dalsa.com/dc/design/dc_sensor.asp

if this site says that the best film stocks are around 12 stops it would be interesting to know the latitude of your camera.

thank you,

Hey Juan have you had a chance to check out this video codec?


http://www.digitalanarchy.com/micro/micro_none16.html

I am not even sure if it will work for real time recording but it seems like a very nice codec. There are two codecs. One is free and just does uncompressed 16 bit video. The other is $99.00 and does 16 bit video with a lossless compression that can get anywhere from 3:1 to 12:1 compression depending on the material.

Sorry if I keep going on about this. I wasn't sure if any of my old posts worked or not.

I haven't tried it yet, but I have read the specs and it seems like it will work for what we are doing. I guess as long as FCP and other programs don't have trouble reading the files with the codec, it's all good. It will help with the large number of files and sync'ing of sound.

Juan

Jake:

I'm not an expert in photography or exposure, but intuitively, since there are about two stops of difference from the optimum lighting for RAW and DV, i would assume that the raw footage has about two stops of additional dynamic range.

Can someone else post and say if this is correct or not?

MicroCosm Codec
 

Just a quick remark about the MicroCosm codec: although it's compression ratio for lossless compression is impressive, it's far from real-time. Both reading from and writing to the MicroCosm codec takes up a lot of CPU resources.

A better place to look might be the SheerVideo codec, which is supposed to do realtime compression and decompression of up to 10-bit (beta) video streams. http://www.bitjazz.com/

Support for 16-bit is supposedly on it's way.

Bar3nd

Well, this is where offline editing comes in.
Use the codec for storage, and convert somehow to a smaller version of the file for editing.

Stephen

Not a lot of time for me to be on the forum lately.. but i'm glad to see this is coming along.

Juan, when you're ready to start developing for the XL1, if there is anything i can do (loan you one, etc) just let me know.. i can't wait to get one of your mods.

Adam:

If by the time i am done with the DVX design none of the HD cameras have come out, I plan to apply this to the XL1s.

The one thing I DO need is the service manual. Once again, if anyone here has it please email me.

I can apply the exact same device I have on the DVX to the XL1 with a small modification of the probing hardware, but I need the service manual to actually do it.

Juan:

I think the results on the XL1s would be interesting, although there is a risk they may not be useful. Canon's low pixel count and narrower latitude combined with lots of pixel shift might make the RAW image less than desirable.

Or the converse, it might make dramatic improvement depending on how much of Canon's processing occurs on the CCD block and how much in the codec assembling.

I am not so sure a mod for the xl1 would really be worth it. First of all I have a XL1 and like it so this isn't an attack on the XL1.

1. The chips in the XL1 are not the best and have a low pixel count.

2. The XL1 is an interlaced camera.

3. The XL1 runs at 30 frames per second unless you have a Pal camera then you can get 25.

4. The XL1 image is already soft without blowing it up. If you upscale an xl1 image it gets even worse. The detail just isn't there compared to the dvx100 or higher end cameras.

The XL1 with mod would still give you for the most part the same type of image qulaity that you have with the XL1 now. You would just have more color depth and accuracy. The same issues such as converting 30 fps to 24 would still be there. De-interlacing would still be there. The same soft picture with low pixel CCD's would still be there.

The reason the mod works well with the DVX100 is because it is already a great camera on it's own. Having a true 24P camera with CCD's with a high pixel count make the camera already high quality. Really the only downside to the image from the DVX is that the CCD's arent bigger than 1/3' and that the camera forces the image into a DV codec. The mod fixes the problem of getting rid of the dv codec and giving the true quality of the camera.

I would actually take a normal dvx100 over an XL1 with the mod anyday due to the other aspects that make good image quality.

Besides the XL1 is getting a little old and by the time Juan would have time to make the mod the "XL2" will be out and everybody will be trying to get rid of their XL1's. Now a mod for the XL2 is a different story.

Pardon if I've missed any past posts on this and am being repetitive.. but as of far I can't find any.

The hype on the threat has been about getting 4:4:4 RGB out of DVX100, which peaks my curiosity just as much as the next guy. But how about 4:2:2? It seems like getting 4:2:2 (lets say 8-bit at least) out of a DV camera should be extensively easier.

I think Juan might have mentioned a type of switch to his mod that’ll allow you to go from 4:4:4 to 4:2:2, which is what I’m more practically interested in using, as it’ll take up less disk space, will be more software/hardware compatible, but will still be an awesome potential over DV.

If anything, might I inquire what would specifically be involved as far as trying to tap into a cameras pre-DV compressed signal and getting the video onto a PC. Is it a simple process of physically tapping into a cameras pre-DV Y/C stream and routing it through firewire then capturing it with anything that supports 4:2:2.. or (as I'm assuming) are then other technical dilemmas that are most likely beyond my intellect.

Michael:

it is actually simpler to get 4:4:4 because that's what the camera natively captures. You would intuitively think that it should not be a lot of work to throw away samples and get 4:2:2, but the problem is that the camera captures 4:4:4 >>RGB<<. This means that to get any other YUV decimation I have to first convert RGB to YUV, and then discard samples. This is why it is just simpler to store the RGB that comes out of the camera.

About what is involved, there are two parts to it. First of all is finding where to get the signals and how to get them cleanly out of the camera, It sounds simple, but when you are trying to tap into ~40 pins of several 0.5mm surface mount chips it's not simple at all. Add to that the space constraint when the the camera case is completely closed and your options are narrowed down to custom work.

The second part of it is the actual device which interfaces this digital stream with whatever you want to record on, which is in itself, the hardest part of the project.

Now, if you didn't care about portability or price, you can get an expensive scientific digital capture card for your PC and capture the data RAW. But that means your camera is tied to your desktop PC, and the camera case open. Pretty useless, unless you won't be moving the camera at all.

If you want it to be portable, you basically need to design a small computer system in a box that interfaces the digital data to firewire or a drive.

And, you can't really use a small PC board like a PC104 because not one of them can handle the raw digital input at the fast rate the camera puts it out.

Hope this helps...

Regarding monitoring of the signal I think you should just go for a standard DVI interface.

As the captured signal is not white balanced and has no gamma correction it would be unfit for a CRT based SDI monitor.

By using DVI you could keep the signal in 4:4:4 RGB and retain the full resolution. You would have to truncate the signal to 8 bit of cause. And you could use a cheap 1024x768 LCD panel as the monitor.

I took a look at that Sheervideo codec and it seems like it might work a little better than the microcosm codec from Digital Anarchy. From the website it seems this codec will be fast enough to capture and edit video with. We may even at some point in the future make drivers for editing systems to give us realtime video out through a decklink card or another card like it.

One of the downsides to the codec however is that it costs $150.00 and is only in beta right now for the windows version.

I for one do not mind spending $150 for a codec but some may not want to if the codec is going to cost $150 and right now only give them 10 bit video and very limited support for Windows. The Microcosm codec does cost $100.00 but their none16 codec is free and is 16bit.

I am sure in the near future the windows support will be fully added as well as 16 bit.

I figured the codec was going to be slow for the mod but it sure is hard to beat the price of free.

Juan--Two stops is about right--at least in the digital slr world. For instance, I can take a raw image from my Canon 1D and grab an extra two stops or so when processing.

Thanks David. I guess if the standard latitude of the DV output is about 6 stops, the RAW latitude should be about 8 stops?

Anders:
That's a very interesting idea, i'm going to look into the DVI specs...

American Cinematographer (April '03) found the DVX output to be about 8 stops, so if we are getting 2 more stops... could we be looking at a 10 stop DV camera? Woo hoo!!!

Interesting...what is a sure way to test this? Can I somehow to a comparison with my film camera?

Juan,

The only way I know of would be with a lightmeter. I wouldn't recommend buying one just for this test, but if Purdue has photography classes or filmmaking classes, you can probably check one out for free from the school. Shouldn't be too hard to learn how to use it. The tech who checks it out to you may be able to give you a quick tutorial. If school is closed, your next best bet for rental is probably in Indy (depending on which campus you are at, Chicago may be closer).

On the other hand, if you are going to continue to use this mod yourself, buying a light meter may be the best way to go, depending on how good the monitoring situation is.

I have both a lightmeter a DVX100 pal and a waveformmonitor so I should be able to test it
Don't know if I have the time though.

Anders,

I was actually recommending to Juan that he test the exposure range on his modified DVX. We can probably trust Michael Phillips' assertion in American Cinematographer that the DVX normally has an exposure range of 8 stops, but thanks for the offer anyway.

If the 8-stop figure is to be trusted, then i think that just by comparing the DV output and the RAW output in the latitude test we can get the 2-fstop number. It's actually a bit over 2-fstops, but if we round it down to 2, we get 10-fstops of latitude.

This actually sounds very close to what can be observed from the 35mm comparison. Everything looks similar except for an area in the green box which is washed out, but that is partly due to the fact that the two cameras were in different positions.

10-fstops...nice.

Juan

Actually Juan, you bring up a good point about the cameras shooting from different locations.

At the time you said that, since the camera was open, it was hard to move. Now that it is closed up, would it be possible to re-do the shoot with the DVX right in the same spot as the film camera? That would be the ultimate comparison.

My ultimate wish list in this shot would have some shadow areas that go almost black, and some areas of white that have detail right up to the point where they go pure white, and most of the rest of the scene somewhere in the middle. That would allow us to not only see how they compare, but also how much they can be pushed in relation to each other. And, of course, it should be a perfectly static scene, much like the previous one (for repeatability).

Thanks for everything, Juan. I'm 'bout to jump out of my skin this is so exciting.

I was just thinking about that. I'm going to do it tonight or tomorrow, then get the film developed and post the results.

One question though...the lens on the DVX and on my film SLR camera have different angle of view...should i still put them in the exact same location or should I have the film camera closer if it's wider, etc?

Although now that I have the anamorphic adapter, they might be similar....

Juan

Oh, yeah.

I would recommend trying to get the exact same framing, and similar perspective. Of course, this will require a different distance/focal length combination from the 35mm camera to the DVX, and it will never match exactly.

The main thing I was interested in was seeing the shots from the same angle. That way the angles on the lighting are as close as possible.

Also Juan, when you get the film "developed", you are just taking it in and having it put on disk, right?

When you do that, do you get an option as to what quality the pictures are, things like resolution, compression and such? Seems like it could make a difference. I would hate to find out that the 1-hour photo place compresses the pictures after they scan the negative, and said compression cuts down on the lattitude of the picture, giving us misleading results. Might be worth checking into.

Two things:

First of all, i looked into the DVI solution to replace the SDI and it seems like it is the way to go. Even though I already have the SDI FPGA configuration, I could really use the free gates, and DVI takes almost no gates to implement. This would allow the device to be connected to any flat screen and monitor the RAW video in 4:4:4 8-bit RGB.

About the film developing: Afaik, it is a 2K scan of the negative, this means that it is scanned with a CCD which reacts linearly to color. However the film reacts logarithmically, so we should still get the latitude of film. The images are JPG compressed but my best guess is that this has no effect on latitude.

It's a good comparison.

Juan

Regarding the XL1 as being a bad candidate.

Not to sound foolishly brand-loyal, but i really do love the XL1. The results are typically a bit soft, and they do have a low pixel count, but the colors are very vibrant, the artifacting is extremely minimal (because the picture isn't over sharpened like the DVX), and all of our stuff gets converted the 24p in post and color corrected heavily anyway.. so i would love the ability to color correct natively in 16bit space, rather than convert to 16bit from DV. I know for some (many?), the DVX is a better camera.. but I already get good results following the formula of 30i --> 24p --> color correction --> edit.. so if i can skip a step and already be working in the right color space, that would be nice. it would help increase canon's limited resolution, and be great for chromakey etc. not to menion mini35..

If the XL2 comes out anytime soon, and the pricerange is right, we'll buy one or two of those and maybe seen if juan's mod would be better suited to that. but for now, the XL1 is my winner.

i guess we'll wait and see when juan finished the DVX version and how well it works.

Update:

It works! The first stage of the prototype is working, capturing directly to my mac via Firewire 800. It is caseless for now, but in a week or so i will be able to post pictures of the prototype case on the camera. The final device will probably be machined out of aluminum, but I did this one in polycarbonate plastic to save time.

This week we will be testing the DVI interface as well.

Does anyone know if any of the 35mm 'adapters' (Aldus35, etc) discussed in these forums work with the anamorphic adapter? i.e. can i get 35mm DOF and 16:9 on the 4:3 chips?

Juan

Juan,

Congratulations!
Glad to hear the prototype is together!

I think the solution would be to get an anamorphic lens to put on the adapter, not adding the Panasonic anamorphic adapter to a standard 35mm lens.

You could put the DVX anamorphic adapter on a standard 35mm lens, but it would probably not be worth it due to the quality of the glass, and the adapter steps needed to attach it.

Just my 2 cents.

-Luis