Get 9.5 Stops of dynamic range from your DVX100

pertierr said:
A 100GB FW drive can be used as a 'cartridge' and holds over an hour of 4:4:4 uncompressed footage
Click to expand...
Perhaps, but let's be clear here: there's a difference between Andromeda's 4:4:4, and its high-def. It doesn't do both. You can have standard-def at 4:4:4, or high-def (at ?:?:?) but not high-def 4:4:4.

What is the effective color sampling when it's in high-def mode?
 
Ok if all this is true it is probably I would count it top of possible accessories. Barry shooting with Andromeda with squeze would then give a better resolution than standard dvx with ana lense?

Juan whats the cost?
 
Barry_Green said:
Perhaps, but let's be clear here: there's a difference between Andromeda's 4:4:4, and its high-def. It doesn't do both. You can have standard-def at 4:4:4, or high-def (at ?:?:?) but not high-def 4:4:4.

What is the effective color sampling when it's in high-def mode?
Click to expand...

Barry (I can be wrong but) Andromeda allows you to record HD (1540x990) at 12 bits 4:4:4

http://www.reel-stream.com/cap_hd
"...Andromeda, but also the ability to output 1540x990 resolution clips that are still uncompressed"

Bayer rocks !
A very good pdf : http://graphics.cs.msu.su/en/publications/text/prog2004lk.pdf
 
Juan, no, I meant the DVX. Can you provide best case exposure on the DVX in DV (not the HVX) and best case exposure with the Andromeda in low light were noise is to be expected?
 
pertierr said:
Oh, I agree, i'm not saying the DV footage(on the old images on the website, not the latest comparison) is not over exposed. All I'm saying is that for the purposes of evaluating if the Andromeda footage has a greater dynamic range than the DV output, a shot of any scene with a wide range of light intensities will work, regardless of whether it is clipping or not.

I'm addressing the argument that because the DV footage is overexposed, the two frames don't show that the Andromeda image captures a wider dynamic range. This is simply not true. The proof is in that more dynamic information made it to the Andromeda image than to the DV image. It's simple.

Once you determine that the in-camera processing cuts off X stops on the high end, that remains true for the same settings, regardless of what the scene is.

Cheers,
Juan
Click to expand...

This was my point really - those "overexposed" DV images (I am refering to the captures on your "ANDROMEDA HD CAPTURES" page BTW) made me think, "Hang on, is the same thing happening with this chart, and causing the whites to blow out unnecessarily?" Can you see what I mean?

I am not denying the increased latitude the mod brings, and fully understand that it is superior to the DV output... just some of your captures suggest that the DV output is worse than it really is.

Could you post a link to the latest comparisons please.

Thanks for all your explainations - sorry for being such a skeptic! :)
 
Are you guys having a "Juannie Roast" this morning? :)

They've been reasonably clear on these things before and they're getting better at it. In an attempt to diffuse the divisive direction of this thread, I'll mention a few things: SD or HD resolution has to do with different ways of interpreting the same data. So, with an uncompressed capture you could end up with 444/480P or 422/720P or 4??/1080P. There are various processing algorithms they've experimented with with varying levels of results. Remember, this is still largely an experimental camera being used by a handful of beta testers and of course, the developers. They could absolutely improve explaining things better to the general audience who has a knee jerk reaction of screaming, "BS!" as loudly as possible when they think they're being conned (presumably a result of not understanding how raw chip outputs are supposed to look)

mmm-
If they set the iris two stops below where they did, then the blacks would have clipped harder.
It doesn't change the fact that the blacks MOST CERTAINLY WOULD HAVE clipped!

Your argument is moot.
 
thisiswells said:
So, with an uncompressed capture you could end up with 444/480P or 422/720P or 4??/1080P.
Click to expand...

ThisisWells, I disagree with this sentence. You could end up with 444 in 1080p (with the Juan mod).
Of course, for pratical reason (NLE) you can decimate the 444.

Pixel shifting allows you to use bayer patterns to "find/guess" the missing color channel.
 
The X:X:X notation in our system refers to the sampling mode. In full bandwidth RGB (4:4:4) sampling mode, all the sensor elements(pixels) of the 3 CCD's are sampled. If you put this into a standard definition 770x492 image, then you get a full color RGB image.

Now, when the data is demosaiced into an HD image, it is no different than the way the image from the Drake, Kinetta, JVC-HD1, Arri D-20 is generated. Those cameras have one sensor and only record one color per pixel; either Red or Green or Blue, and then that mosaic image is processed into a full color image by approximating the missing color values. Yet those manufacturers claim 4:4:4 output, when it is physically impossible to get full color for each pixel on the single sensor.

I haven't made any claims other than the HD is uncompressed, simply because as an engineer I wouldn't call it 4:4:4, it's something above 4:2:2. But on the other hand, where's the people criticizing those manufacturers claiming 4:4:4 output with cameras that only have one bayer sensor. We're forced to advertise inferior specs for being technically honest. :)

Cheers,
Juan
 
See, I don't have any problems believing you because my father studied engineering at Purdue and I believe in the program. Good luck convincing everyone else.
 
thisiswells said:
mmm-
If they set the iris two stops below where they did, then the blacks would have clipped harder.
It doesn't change the fact that the blacks MOST CERTAINLY WOULD HAVE clipped!

Your argument is moot.
Click to expand...

I know, I was originally concerned that the test unfairly favoured the uncompressed output, like some of the images on the website do. Juan has explained that this was not the case though. I am quite happy (and impressed) now... but the old images still haunt my sleep... :)
 
This thread has made me decide to remove those images. They were one of the very first Andromeda shots, and where done in Vegas at NAB with the very first prototype, but I think the latest dynamic range chart serves the same purpose.
 
Not necessarily Juan. Take for example that balcony shot; I think it depicts very well the latitude difference between DV compression and raw output. When I first saw it, I immediately understood the advantage of the Andromeda.
 
Yeah, that one doesn't seem to illustrate much of anything, other than the size of the recorded image. If they were both shooting the same portrait, THEN it would be interesting.

Nonetheless, this is very exciting. 9.5 stops from a DVX is unreal! Can't wait to see the finished product available.
 
yeah it's tricky, most people didn't know that the Varicam records 960x720(as the HVX will as well). If we stretch it to 1280x720, then it only looks as good as the algorithm used to stretch it.

I'm really hoping to do a Varicam - Andromeda side by side sometime. I think the dynamic range of the Varicam is close to 9.5 stops so I wonder how it compares.

In the meantime, i think that image gets the axe as well.

Cheers,
Juan
 
You must log in or register to reply here.
Back
Top