Discussion. How would you do a mod like the DVX100 uncompressed mod, on this camera?


Wayne.

Hi,

Using a sodding great disk array, for starters.

But it'd be theoretically the same thing.

Phil

The device would be the same, but it might have to be modified for the much larger throughput of HD, as well as the drives.

Juan

Apparently the new Sony camera doesn't even use 1440x1080 chips like the HDV spec calls for. According to some its doing some sort of "upsampling" to achieve 1440. Maybe it's better to wait to see what JVC comes out with in December.

By the way Juan...would it be possible/practical to create an HD- SDI output from one of these HDV cameras that would interface with one of the new and cheap HD-SDI capture cards such as Black Magic Decklink HD? Or is component analog a better/easier option?

In fact I guess the JVC camera has a higher true resolution than the Sony's.
Anyway that is questionable, what have more spatial resolution one 1280x720 Bayer chip or three 960x1080 (which in fact ends at 960x540 ,Remember the Sony's is interlaced) chips?
Both has 4:2:0 color sampling so there is no color resolution advantage.
I guess the only advantage would be the better highlights management the Sony's surely will have.........

The Sony would have a number of advantages. it is not just 960*540*60, but 960*540*3*60, which is just as valid, as it interpolates, as bayer does. The light is spread accross three chips, without desentising, light wasting, filters, and the pixels are bigger than if they were true 1080's which gives them a spread much better than a bayer 720p 1/3inch chip, and much better sensitivity and accuracy. If we sat down and did the calculations, it would be interesting to see what size bayer chip we would need to match it. I also would not expect the full benefit to be taken advantage of, as it is a cheap prosumer camera (like the VX1000 was of lower grade compared to vx2000).

But yes, the interlace scheeme, and 4:2:0, does let it down when compared to a 60fps 720p cinema camera, but for the HDTV market most of the stuff is at most 60i. I think we can do much better with the cinema cameras, but there is something about a complete prebuilt system that can be made to record to disk (if it canbe raw moded) aswell as tape when needed, that is appealing to lots of people.

<<<-- it is not just 960*540*60, but 960*540*3*60, which is just as valid, as it interpolates, as bayer does. -->>>

Wayne: are you saying that it does not matter whether you have
a single chip bayer or a 3-chip system?

No, not at all, I'm saying that 3 chip expands your pixel area (3 sensor cells per pixel) and the pixels are bigger than they would be if they were true 1920 * 1080. This (without the bayer filter) gives better sensitivity and range of charge it can accumulate. But in this case the interpolation and possible other 4:2:0 etc dumbing down of the HDV model, brings it down compared to what we are doing. Otherwise we would need effective sensor areas much bigger than theirs, more pixels etc to compete with it, then I still haven't calculated how we get past the filter limited sensitivity.

I think one thing the machine vision box cameras offers, is probably very good filters, and access to pure data. In a lot of prosumer cameras they don't have correct gamma curves, 8-bit colour, 4:1:1/4:2:0, and they oversatuate, which flattens colour and luminance detail even more. I don't know how they do it but I suspect they take all the colour channels and move/or squeeze them up, maybe stretching from the bottom to maintain pure blacks, to give that flat, play school, crayon style, mini-dv look. So even a bayer 720p cinema camera might be better than a top of the line 3 chip prosumer mini-dv, even passed through the mini-dv codec, and work out cheaper in future.

But this is not what this thread is about, but generating interest in raw from the Sony, which camera is probably going to displace the DVX100 for indie. Some people will go for the Sony, some for the cinema camera project, and some for a raw Sony, to each his own (but without true 1080/720p 3 chip 1/3+inch, 50-100Mbs codec, I don't want to buy one).

Sony Makes it very hard to use the HD camera for filmmaking because its not progressive scan at all

Isn't that just a software issue though? Or are the chips not capable of capturing in that manner? If it's just software then I don't see much of a problem as an image would be taken directly off of the CCDs.

Hi,

The scan characteristics are specific to the CCD. Some can do both, some can do one or the other, but if you have an interlaced scan CCD, it's generally never going to do progressive. Some of the types which will theoretically do progressive would require massive changes to hardware beyond the ADCs if it wasn't already implemented in the camera.

Phil

Thanks for the info Phil :)

Something I forgot to mention before. I think the interlace scan is both a disadvantage and a benefit.

One it takes away from a clean progressive image, but it also reveals extra temporal/pixel detail about the image that canbe extracted during deinterlacing. Requires a lot of extra processing though, and preferably RAW 4:4:4 output to bypass the 4:2:0 interlace problems.

Although I have no way of verifying the info, reportedly the camera will have component ouput. I would expect this to be uncompressed.

This from avsforum.com:

"following Sony's press event in NY.
The change is related to the inputs and outputs."


"REVISED (Based on new/corrected information from Sony):
Features component video out (but NOT in)"

That would be a very good start, if they allow you to use them uncompressed during filming (unlike the JVC).

Well, about the interlaced thing for Cinema.
It is not a big problem to have 1080i source because it is easier to convert 60i to 24 fps than 30p..
So I guess it is a good candidate for such a work..

<<<-- Originally posted by Charles Moore :
Features component video out (but NOT in)" -->>>

Yes, this is like what they did with the vx1000. I think it was "ilink out" but not in, stopped me from buying it, they upgraded it in the next model.

shooting 60i for eventual 24p output is totally doable nowadays with good results via affordable software, but it is a significant step added to the post process. i'm doing exactly that with a longform project right now and some of the potential drawbacks include:
- significant increase in labor & logistical management for compositing effects
- possible discrepencies in synch (though miniscule) if your sound designer is working from the pre-converted 60i edit
- possible aesthetic discrepencies from your 60i edit (ie. the field containing a quick gleam/flash/muzzle flare happens to be one of the fields dropped by the conversion software), though you can go back and reconvert the shot with the gleam.
- though the framerate conversion looks good and is probably invisible to 95% of all consumers, there are some visual flaws from the interpolation that show up in your footage.

some of the advantages are:
- option of cleanly "overcranking" any of your footage to 80% or 40% slow motion or "undercranking" to 160%, with very good results. aside from the obvious use for effects, it has actually been useful in editing, with a few shots where motion/movement continuity was flawed.
- if you plan to do any non-constant time remapping, 60i will yield much better results than 24p or 30p.

personally, after dealing with this firsthand, i would opt for shooting 24p, especially if your project involves a lot of compositing. and 30p flat out cannot convert to 24p without looking like a big mess (in my opinion).

but so far i've read some info online about the HDR-FX1 that implies that there's a strong possibility that it will be able to output at least some of the HDV-sanctioned 720progressive framerates of 24, 30, 50, and 60, but only through the component outputs (the HDR-FX1 reportedly won't lay the 720p footage to tape). i just hope that "raw mod" juan can get in there and pull those raw frames out for us, especially if those progressive framerates are actually coming off the ccds at 1080 and getting resized to 720 by the dsp. 14-bit, 4:4:4, 24fps, 1080p from a $3700 camera just sounds insane... call the crazy house, because i wanna move in.

how do you know the chips are 14 bit???
Not every conversion include discarding fields.Be carefull with that.
Every post-house has its own method.Remember quality depends on people not machines, most of the times....

WM9 encoded Sony HDV footage:

File location:

The URL: ftp://md-ftp.sonypictures.com

Username/Password (case-sensitive):
dude
Sweetn3ss

for more information on how it was shot:

http://www.creativecow.net/forum/read_post.php?postid=109480679582805&forumid=24

juan, good point about how not all 60i to 24p conversion techniques discard fields. i can't speak for the numerous proprietary methods developed by post houses, but i was speaking specifically about software available to anyone. and among those options, there seems to be three general approaches, and it seems that all of the software available utilizes one, or some sort of combination, of these; fractionally discarding fields, morph technology-based temporal interpolation, and frame blending. and all three have their own achilles heal when it comes gleams/flashes that appear (or peak) for one field.

Charles, those user and pass don´t work..

Juan, it does work, though you may need to try an FTP program rather than IE. Check out this thread. (http://www.dvinfo.net/conf/showthread.php?s=&postid=220827#post220827)

OK thank you very much

If the CCD's could be offset (if we can get raw footage), than 960*1080 becomes 2880*1080 (or some other configuration) and with an anamorphic adaptor it would be good for 2.39:1 shooting.

Wayne, I guess that when you offset the CCDs yo move just a little the green one, and in such a case what you would get would be 1920x1080.
The idea is that you put the green pixels on the interspace between the other CCD pixels.If you wanted to shift the 3 chips you won't have enough space to gain resolution.
Anyway offsetting the chips is almost impossible unless you have a really expensive equipment just for that kind of manipulations..

I guess that idea is fantastic for the DVX100, after Juan's discovery of the pixel shift scheme it has.
If that could be used you would be getting a fantastic 15XX x 500, and assuming the Bayer Mosaic is said to has around 75% the resolution of a 3 chip conf. we could assure it would be like a 1280x500.(assuming its 794 horizontal res.)

Yes, it would be a workshop only type conversion.

Could you do some sort of software offset or does the CCD HAVE to be physically moved?

You can do many things in software, but if the CCD's are not shifted then there is no way that software can 'know' what actual light was impinging between red/blue sensor elements. The data will simply not be there.

As all these sensors are attached to a prism. Moving them would be difficult. I imagine that there might be some servicing screws, or something, to re-align them, but more likely they are just straight fix them at manufacture.

But then working out allignment is another big problem,an home made allignment chart could be made to use to allign it.

But if you have to break a seal on a fixed unit then you would open up a whole new set of problems, with dust/fibre/air (will corrode chips, CCD's I don't know though). Then you would need a totally clean workshop type environment (vacume sealing would be even better but I would say only a handfull of places in the world have that).

Juan, I noticed that your DVX100 pictures are sized closer to HD picture frame, is that upscaling, or are you taking advantage of the pixel shift? Can you explain how good it is?

Forget about it Wayne, really.
Usually the CCDs are glued to the prism.At least on cheap models.
Tube cameras used to have positioning screws....
And yes you need a sealed environment (they look like those transparent places where babies are put they they are born before 9 months of pregnancy, if only I could explain myself a little bit clearer..)
I suggest everybody to continue this kind of conversations on the uncompressed DVX100 thread, cause this is not the right place.

Yes, I was just trying to warn people that it was likely to be too difficult for them to do.

I know the ones your talking about, the ones I meant were even more advanced they use for Hard Disk data recovery.

Thanks


Actually I do have a way to shift the image instead of the CCD's, and I can guarantee nobody will like it unless they have a detachable lens camera, but I'll throw it into the melting pot anyway.

It requires an adaptor like the 35mm adaptor, a mechanism/prism to seperately dis-align the primaries projected onto the screen that is then filmed. Too big, too fiddly, too easy to knock out of allignment.