Yes, that is the setup I own and just used on a 9 day shoot.

I had assumed the same thing, that going through the optics of the DVX would handicap the system, having used it extensively with the XL1s. However, I was surprised to see how much quality was retained with the DVX. It looks fantastic.

Hi,

I noticed the gigantic amount of overexposure (compared, that is, to the DV) that the setup seems to require to get the best out of the raw images. On the other hand, images exposed at the DV level do pump up quite well and you can then end up with a large dynamic range. You just end up putting a huge negative gamma curve onto it.

If you exposed such that the raw looked bright enough straight off the camera, you'd be throwing away a lot of luminance resolution in clipping.

Also: people constantly rate video cameras at between 300 and 400ASA. This directly contradicts my experience and I believe it's a lot to do with the misapplication of film-style exposure and metering techniques to video. To wit: a few weeks ago I was shooting some makeup tests on 35mm stills (Fuji Superia 1600) and video (Panasonic AG-DVC200, a reasonable semi-broadcast video camera.) The stills were underexposed at 1/15 of a second and F2.8 whereas I was some way from fully open (probably around F4) on the PAL video at 1/50 to get a reasonable video exposure. Lighting was a mix of small 500 1K tungsten and sub-100W fluorescent with a 575W HQI backlight. So, small light, but the (very fast) film was struggling and the (entirely average) video was fine. This also happened to me on a night exterior in Amsterdam where the unit photographer was unable to shoot without flash on 800 stock yet I was shooting quite happily under the street lighting without gain.

Draw from this what you will, but I have formed the opinion that video cameras are effectively a lot faster than standard characterisation techniques make them out to be - at least equivalent to ISO800 and probably more.

Phil

The same happened to me in a recent work.
We shoot a whole feature film in Buenos Aires (you will be able to see it at next San Sebastian festival) with a DVX100 at night just using the sodium street lights, nothing more...
We calculate that going film, we should have used at least 500 ASA to obtain simillar results...

For what it's worth...

The DVX100 has been tested by Berry Green and Jarrad Land from dvxusers.com. They find the DVX100 has equivalent ASA rating of 640. Both are very competent, knowledgeable individuals and I have no trouble believing the accuracy of their tests.

-Rodger

I think the same.640 ASA sounds right...

Yeah, my assessment of the DVX at 320 ASA was fairly casual and besides, the exposure curve of video vs film means that EI ratings are fairly meaningless. As was noted here, the underexposure characteristics are quite different, video will "see" a lot more into the shadows in the same way that it doesn't see into the highlights. It's just part of why I don't use a light meter with video unless I am pre-lighting a set without the camera in hand.

Also, the mode being used with the DVX would dictate the EI. Moving between different cine modes, you can easily note shifts in exposure.

This is not to dispute Barry and Jarred's findings, it's just interpretive. If the camera had a -3 db setting like the XL1, I would use that as much as I could to try to keep the noise down.

Does anyone have an image of what the DVX looks like with the 35mm adapter on it? The current device occupies the bottom of the camera and is as big as the body of the camera itself. The probe cabling can be routed sideways along the bottom, but i'm not sure how long the cabling can be made without signal degradation.

Juan

by the way Charles...

The A/D's in the DVX supports gains from -6dB to +42dB. Whether the camera actually uses the full range, i cannot tell from the schematic. The gain setting is set through a serial interface on each A/D so modifying this would be extremely tricky.

Juan

Hi,

> If the camera had a -3 db setting like the XL1, I would use that
> as much as I could to try to keep the noise down.

Well, looking at these data captures, it would seem that "default" would be at least +6dB and possibly much more, considering how under they look. I presume this is probably so Panasonic can scream about the low light sensitivity in the ads.

Phil

I'm inclined to agree with you, Phil.

Juan, check out our sponsor, ZGC, (http://www.zgc.com/zgc.nsf/c7a682995edb4e7585256b4d001ebd57/b20461fb956bf08085256a8400801da2?OpenDocument) for a good picture of the DVX and Mini35.

I would think that a rear-mounted system would allow the camera to become a better handheld rig by itself (allowing for shoulder mounting) and still keep it compatible with the Mini35, etc. Although this would eliminate the rear viewfinder from use. Perhaps having the flexibility to mount both ways?

<<<-- Originally posted by Charles Papert : <<<-- Originally posted by Brent Douglas : II've been thinking that if you attach the JPWonderBox (ahem) to the bottom of the camera will this still work with the Mini 35 adapter? I'm not so sure how the adapter mounts cause I don't have one (or a DVX for that matter :( ) but they look like they mount on tthe bottom. I thought this would be important cause I recon if anyone is crazy enough to spend $6000US + lenses on the camera to get (from what i've heard) a marginally better picture and slightly shallower depth of field then every single one is going to be jumping on the bandwagon for this baby here.>>>

Brent, as one of the "crazies" who has spent the $10K it actually costs for a Mini35, I can tell you that the adaptor indeed mounts directly to the bottom of the DVX100a, so a breakout box would indeed need to be relocated for use with the Mini35. And while I'm at it, I can also tell you that the ability to use cine lenses introduces factors and possibilities unknown to most video users; focal lengths longer or wider than stock video lenses: better resolution characteristics and flare handling, not mention the world of specialized lenses such as swing-and-tilts, 2:1 anamorphic lenses, boroscopes, etc etc etc...and as far as "slightly shallower" depth of field, it's far more than slight. And yes, this particular crazy will probably jump on the bandwagon for this system once it becomes practical, so I'm watching from the sidelines and looking forward to seeing some frame grabs.
-->>>


Sorry Charles, I didn't mean to be offensive with the word crazy. The meaning is probably changed in context (Australia to US). Perhaps I should have used enthusiastic instead? I myself plan to get one (and a camera to go with it :P) when I can afford it which is why I mentioned it but I did also say I did not know that much about them. I probably should not have posted without researching what i was talking about more. I only did because no one else did and as a prospective buyer (one day) of both products I was hoping they would be compatible. I was not intending to slight the Mini35 or it's owners at all.


Brent

Hi,

My concern is how easy this stuff is going to be to postproduce, especially in a more professional context (this may concern Mr. Papert as well.) I'd very much advocate at least a basic form of metadata, such as tagging each frame with a timecode. This is how digital effects are done with DPX sequences for feature production, and it would allow you to more or less easily conform your high-bandwidth uncompressed material to a cut based on either DV downconversions or tape that you also ran in the camera. Sound is also an issue.

Sure, there are ways to fiddle around all this stuff in FCP or Premiere, but I think at least a few bytes of metadata per frame would be useful. I imagine the problem would be obtaining it - I'm guessing that the DVX100 lacks an easily-read digital timecode output, where the 1394 output need not apply for "easily read."

Phil

Hi read through the entire thread with some excitement.
Like WOW you are one brave dude.
I am a software/hardware graphics person.
I was wondering if catagorize your device as
a Digial I/O interface that transfers image data to the
hard disk?
Also saw you mention that you could do this to other cameras
and I was wondering how since each camera uses slightly
different I/O interfaces dont they?

I find it MOST intresting. Were ever able to
eliminate the noise? I grabbed as many images as I could.
Sorry for all the questions just excited to see you push the
envelope.

Thanks,
Bob

Phil,

Providing timecode is not a problem, the issue is how to provide it. We discussed using DPX files but a lot of people voted against it, just because TIFF seems to be more commonly supported.

Implementing DPX/Cineon files is not hard...i could include a feature to pick your format and have the timecode included with DPX files. However, i don't think programs like FCP could import them? This was a while ago, don't quite remember what conclusions we reached.

So far Shake seems to handle the raw files perfectly. FCP handles them as well. The only thing to remember is to set a compressed format for rendering while working on the clip, otherwise the files are too huge to handle.

Bob:
My device handles raw digital data, it works with any A/D in every camera i've encountered so far.

Juan

Hi,

That's why I suggested the filename thing. Normally in film postproduction files are given a name based on the total frame count represented by the timecode.

01:14:03:21 is 106,701 frames from 00:00:00:00 at 24fps. An example filename might be myshot0106701.tif

You might choose to allow the user to define a prefix.

Because the largest valid timecode is 23:59:59:29 (the length of one day) and the largest number of frames per second in a standard motion picture format is 30, this number is always <2,592,000 and never more than seven digits.

Of course on NTSC DVX-100s this becomes a nightmare of trying to derive the same timecode from the 30-frame camera as will eventually be created by Final Cut once it's unwound the 3:2 pulldown, but I leave that one up to you!

DPX options would be good. I can see a camera assistant with a laptop wirelessly linked to the device tapping in the notes for each shot.

Phil

metadata seems to be a buzzword I've heard a lot these days, I agree with Phil's advice and think that if this camera wanted to be more useful to compositors and CGpost effects, it would be very useful to inculed lens numbers (f stops, focal length, focus setting etc. But I have no idea if that is possible. I wonder if it will be with the xl2's sdk? (I'm beginning to think that the canon SDK will end up being like the 3d lens...)

That data is already encoded on DV tape (at least on my GL1 -- it records at least the time of day, the gain setting, the Fstop, and the zoom level). So if you record DV along with RAW, you should be able to reclaim all of that data.

Yeah, all the data is in DV tape as well, so as long as the tape is rolling the data can be used for the raw data.

The only issue here is that it would need user intervention in order to sync the raw and DV footage. There is a way around this, to sync the device record to the camera record, but it will take another probe.

For now i'm going to assume the user will pick which two frames correspond to each other. Once i'm done with the prototype I can add that feature. I'm sure there will still be some bugs to work out as well.

Juan

This is some CRAAAAAYAYYAYAYAYYYYYYYYYYYYY stuff man. All I request is that this magnificent thing end up being compatable with FCP . . . becasue I just got a mac system, so PLEEEAAASSSEEE.......

Juan,
As I cruised the internet looking for parts.
I gather that if you can get A/D output from a CCD you
deliver the digital data into a Raw RGB format.
I am a little fuzzy what you are sending it to.
A hard disk? Or a firewire stream?

Then I can use that data recorded to my hard disk
and with my computer?

I am sure you are busy if you can answer thats OK
Maybe someone else can answer in your place.

Just trying to fill in the gaps.
Why I am asking? I was thinking of looking for an OLD unwanted
SVHS camera and using your method to get the data into
a much more exceptable format.

Thanks,
Bob

The data is put on a firewire 800 interface. The firewire port on the device can be plugged either to:
A: A PC/Mac for direct streaming to the internal hard drive
B: Any standard Firewire 800 hard disk directly.

Although it doesn't matter how good the stream of data is, i think putting it on SVHS is probably worse than using the DV output.

Juan

Here's what the prototype looks like right now.

I've been able to map every function so far to the LCD interface, so there has been little need for any other buttons. The firewire interface is on the back.

I've already found things that will be changed for the production version, such as the size of the box might be slightly shorter.

http://expert.cc.purdue.edu/~pertierr/proto1.jpg

I'm a little fuzzy on the interface too.
So it plugs into the computer, OK.
How does it gain access to the drives?
I assume you are writing a PC-Mac application to send the data to the drive. You obviously have to tell it what drive to write to, at a minimum!

Yes?

-Les

<<<-- Originally posted by Juan P. Pertierra : The data is put on a firewire 800 interface. The firewire port on the device can be plugged either to:
A: A PC/Mac for direct streaming to the internal hard drive
B: Any standard Firewire 800 hard disk directly.

Although it doesn't matter how good the stream of data is, i think putting it on SVHS is probably worse than using the DV output.

Juan -->>>

That's right Les. It works exactly like when you do DV capture directly to your hard drive, the only difference is that you will be using my program. You tell it where to put the files.

When hooked up to a drive, the device acts as a PC, and writes files directly onto the drive whenever the device is recording.

Juan

"The firewire port on the device can be plugged either to... Any standard Firewire 800 hard disk directly"

oh my god, i love you.
flat out-- best, most inexpensive, most convenient solution possible.

and if you're still looking for a name, how about "RawBox". if you want it, then i relinquish all rights to you. in exchange, feel free to give me a friendly discount (insert smiley emoticon).

were you ever able to find a way to take care of those "dead" pixels? (sorry if i missed a post that answers this)

do you plan to sell them as a mod (send you a camera), or as new cameras (you order, mod, and sell them complete and ready)?

thanks, dude.

The 'dead pixels' were noise inserted in my experimental setup. In the actual prototype there is no noise because the wire lenghts are extremely short, and the device pulls a tiny amount of current from the camera. The capture card I used in the experimental setup pulls a larger current so the long wires not only caused noise in the RAW feed, but the same speckles were visible in the DV output because of the current drain.

The prototype literally acts like a extremely high impedance probe, such that the camera doesn't even know it is there.

Currently I am only planning to 'modify' cameras owned by other people. It is possible that I might end up selling complete systems as well, but I don't know for sure as of this moment.

Hope this answers it!
Juan

Juan,
I was looking at the same thing someone else here was looking at.
http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=2490

This seems to be intresting. If it can deliver any kind of color fidelity at that resoltution and the output be all digital then thats something intresting ;) Combine with a lens that is.

Bob

hey Juan, a few more questions-- though if they'll be answered in a few days on your website then feel free to ignore them.

is the box permanently attached to the bottom of the camera (i know you mentioned that the line has to be very short)? like can it be disconnected and left at home if raw capture isn't needed?

dare i ask what the ballpark cost is going to be and your forseeable production threshold (ie. how many a month do you plan to do)?

are the files from the 12-bit or 10-bit step in the camera?

also, are you familiar with the whole linear(HDR) vs. sRGB issue? i won't even bother getting into that, but i bring it up because some guys just released semi-shareware aftereffects plugins called eLin that allows for sRGB-to-linear (and back) conversion. if anyone's familiar with the HDR/sRGB thing, then like me you'll see these plugins as a pretty big deal. and though those plugins weren't intended for something like this, they may be very useful for the raw files, since they have an unnatural "put 10/12-bit into a 16-bit file" white point (some of the eLin plugins are for handling 10-bit cineon files). also, based on what i've seen from your stills and what you've mentioned about white balance and latitude, the raw files are likely to be linear (though not true HDR) images. i've yet to try it, or even do more than skim the eLin documentation, but i plan to do some tests with your older stills in a few weeks when i have time. the download is at

http://redgiantsoftware.com/elin.html

and some very useful info on HDR vs. sRGB is at

http://www.cinenet.net/~spitzak/conversion/index.html

warning: this stuff may put some people to sleep. and the eLin documentation i skimmed makes this seem like THE most technical-oriented aftereffects plugins i've ever seen. if you can post some recent stills of both raw and simultameous dv, then i'll be happy to try and see if eLin will be especially useful for your mod.

thanks again.

1. The top lid of the box has a tripod mount, along with rubber bumpers just like your tripod. The top lid is first mounted on the bottom of the camera, and the probe cables fall through an opening in the lid. The rest of the box is the attached to the lid using some screws. So yes, it is completely removable although it takes a bit of care not to force the probe cables.

2.When I have the exact figure i will post it here...

3.The data is obtained from the most pristine digital stage, which is right after the data is very first sampled from the CCD's. This is 12-bit RGB data(36-bit color), and it is simply the BEST possible quality you can get out of the hardware.

I'll read on the link, however there is essentially no difference between a 12-bit sample, and the same sample padded with zeros to 16 bits. It's the same number. Kinda like adding zeros after a decimal point.

The data is linear, corresponding to the response of the CCD's. The white value should be close to correct as long as the White Balance is adjusted.

Juan

Oh, and another interesting fact about the GAIN.

I ran some tests, and it seems the gain setting for the CCD's is set to a constant value. The Gain setting on the camera is actually done by the internal circuitry, and after the data has been captured.

So, it would be interesting to be able to control the Gain setting of the CCD's since they have a wide range of adjustment. This is hard to do without modification though...

<<<-- Originally posted by Bob Tasa : Juan,
I was looking at the same thing someone else here was looking at.
http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=2490

This seems to be intresting. If it can deliver any kind of color fidelity at that resoltution and the output be all digital then thats something intresting ;) Combine with a lens that is.

Bob -->>>

Hi Bob,

The ES-2093 (http://www.edmundoptics.com/onlinecatalog/displayproduct.cfm?productID=2498) seems better (real 24 frames/s) and more expensive.

Does one ccd is enough to get a good picture quality ? If, it was, I could really be interested to build my own camera.
Otherwise, I will be one the Juan's customer for his incredible mod.

flax

What is the pixel dimensions using the mod? I remember it being posted in the thread but looking through 69 pages for it is rather daunting. And would the pixel yield be the same if the mod is performed on a PAL DVX or would it be higher? Thanx.

The resolution is 773x494. The PAL DVX will yield more lines. I am unsure of whether it will yield more horizontal resolution or not.

Juan

Can this mod be done with a GL2? Or an XL2?

When will the site be ready?

Ron: the mod could be done with any camera basically. The
question is: will it be made for other camera's.

Personally I would wait to see the project on the DVX finished
and others have their DVX's modified, tested and reported back
before thinking about adapting another camera.

This stuff isn't simple.

Ditto what Rob said.

The website is done, I am now adding the content and images. I'm shooting for tomorrow or tuesday. Hey, you can all get a preview if you can guess the name of the company and .com it. :)

Juan

Im going crazy trying to figure it out and I dont even own a dvx.


Stephen

Tried to guess but no luck so far.

So when will the link be up for the web site?

No luck here either. Although I have fumbled onto some fairly interesting sites... and apparently, the P. stands for Pablo :).

Hello all. Forgive me if this has been asked before, but how much HDD space does the output of this camera consume? Let's assume the highest possible quality settings (4:4:4 12-bit uncompressed). How many GBs would an hour take? MiniDV is about 13GB per hour. How about other, slightly compressed settings?

Also, you mention that it records to FW800 drives... no chance of using FW400? Too slow?

Thanks again for all the hard work. Can't wait to see the website!

At full quality 4:4:4 36-bit RGB 24P, with no compression, it consumes about 40MB/sec. I am working however on lossless compression such that the number can be greatly reduced while mantaining full quality. If you drop the color precision down to 30-bit RGB it comes to ~33MB/sec uncompressed.

Now that you mention it, there ~is~ a chance that this might all work over Firewire400 after all. I went with FW800 just to be on the safe side, but the total bandwidth was always a bit under the FW400 bandwidth. Given that the host drive can handle the full 400Mbps rate, it just might work with standard firewire drives.

Do note, however, that i've seen older firewire 400 devices that do not operate at the full rate, plus the main point is that the actual drive that is in the Firewire casing can handle the bandwidth.

I'm confident that all new drives can handle it fine.

Juan(Pablo) :)

144GB/hour . . . so nothing lest than a terrabyte oughta do. . . :)

Where's that URL, Juan . . .??? Need some help with the web design? I can help ya out . . .

Hello Juan and all the lovely think tank!

This is my first post here although I've been checking this thread almost every day since early February (hey, better late than never).

Anyways, fantastic work... I don't care if hell freezes over, I'm gonna buy this thing.

I've done a lot of color grading work (both film and video) and I've had lots of fun playing around with the stills you've posted. A graded image off the mod looks surprisingly pleasant at 2k and 4k resolutions, it doesn't feel forced at all and the extra bits give a lot to 'tamper' with. Beautiful...

Juan, I'm sure your fanbase will grow fast after you get your web page running... screaming girls, paparazzi... you know the drill :)

OK, a couple of questions:

1) Did you get the RGB alignment problem sorted out? BTW, most prosumer 3ccd DV cams I've worked with have their channels very slightly misaligned. The compression usually hides it well, so it hasn't been that much of a
problem (until now). Fairly impossible to detect on a SDTV monitor, but uprezzing gives it away.

2) You have mentioned that it is still possible to record simultaneously on standard DV tape (to use as dailies/backup). As uncompressed capture requires wider aperture than standard recording (due to added stops), shouldn't the DV video be totally overexposed?

3) The Mini35 issue still bugs me. I'm gonna be shooting a movie next year (budget around $2m), and I would
definitely like to use your mod (or several of them). But Mini35 is pretty much a necessity here (gotta use
anamorphic primes). It would be good if they didn't null each other out. There seems to be enough space
between the support rods and the camera base attachment, but I'm not sure.


Thanks again and keep up the good work!
Chris

Not to speak too soon (yah, I know I am) but if this works out on the DVX, maybe he'll make one for the XL2, so you won't even have to bother with the anamorphic primes (can use more accessable standard primes) and you won't even have to worry about the down-sampling of all this prime Canon pixels to fit on a miniDV tape.

Hint, hint...

Also, I've been following this post fairly regularly, but I must have missed the part where the uncompressed image adds stops to the recording... Is there a simple answer to why that is?

Non-downsampled XL2 - now there's something intriguing! Now if only I could afford one!

Chris: This is sort of off topic, sorry, but what do you use to upsample video/images?

Chris:

1.Yes, the alignment problem can be perfectly fixed with some shifting and slight resizing. However, since I only have one DVX I leave open the possibility that the corrections that work for my DVX might not work on another, so my code allows the user to experiment in order to obtain the right alignment.

2.Depends on the scene and settings, but yes it does happen. There's no way around it unless I heavily modify the camera. Most of the shots i've posted however, didn't use the full dynamic range, and thus the DV output was still usable. I've found that to really get a nice film look to it, I under-expose the scene to avoid complete white-outs and then take advantage of the color resolution and apply a gamma curve.

3.That's a tricky one. I completely understand the need for it, but the problem is twofold: the wiring coming out the bottom has to be kept as short as possible, and in addition the box is pretty large so it would be hard to mount it elsewhere. The only option I can think of is a side-by-side arrangement. It will be huge, but it will allow the mini35 to be mounted.

Eric:

The DV process limits the dynamic range of the CCD's in order to comply with NTSC standards. My mod actually uses the full latitude of the CCD"s, which in my tests has resulted in about 2-4 additional F-stops of latitude.

Cheers,
Juan

Come on Juan!!!!!!!! What's the URL?!!!!!!!!

Thanks Juan. Sorry to go over this again if it already has been. Are the stops your mod adds on both sides evenly or does it stop down more than it opens up stops?

Regarding the mini35 mount, would it be feasible to to make your box wider, but shorter so that it could mount to the rods and fit between the camera and rods? (not sure what the longest wire length would be).

Hi Juan and everyone.

Lots of people think you're a lier and a psychopath.
For people interested about that go on cinematography.com :
http://www.cinematography.com/forum2004/index.php?showtopic=1887&st=15

I don't have a personal opinion about it.

I think what you've done is possible even if I don't understand what you don't give the so long awaited URL.

The only thing annoying me is your price.

8500 $ is too expensive.
Especially when you know that some amazing HD will appear soon.
So, is it your last price ?

---

Chris Rubin if you have a 2$ m dollars budget what you spend time with Juan's mod ?

Flax, all i'm gonna say is that you have the wrong person. I checked the link and they are talking about someone called Pete, who wants to do some low-cost HD solution.

And i haven't announced the price yet but it will be waaay lower than $8500 that's for sure.

Juan