Flashing with Varicon?

[Guest kalkarman]

You're shooting night time, in the desert, with the red, you have a clear sky with a full moon. That's it.

 

Could it at all be helpful to "flash" using the Varicon in front of the lens? If it were film, that would open up the blacks, and get you a bit more detail. Is this at all relevant with RED's non-destructive, Camera RAW philosophy?

 

Should I do it?

 

Thanks,

 

Kal

[John Brawley]

You're shooting night time, in the desert, with the red, you have a clear sky with a full moon. That's it.

 

Could it at all be helpful to "flash" using the Varicon in front of the lens? If it were film, that would open up the blacks, and get you a bit more detail. Is this at all relevant with RED's non-destructive, Camera RAW philosophy?

 

Should I do it?

 

Thanks,

 

Kal

 

While flashing or controlled flaring (which is what the varicon is doing) works for getting more detail into the black, if it's night and in the desert, then flashing won't add detail to darkness. If it's black then it's black and there's no coming back. I don't think you'll get much out of a full moon, even at 2000 ASA@ 24 FPS.

 

jb

[Phil Rhodes]

> RED's non-destructive, Camera RAW philosophy

 

The non-destructive camera RAW philosophy that compresses the image 12:1.

 

Yes.

 

That non-destructive camera RAW philosophy.

 

P

[Guest kalkarman]

> RED's non-destructive, Camera RAW philosophy

 

The non-destructive camera RAW philosophy that compresses the image 12:1.

 

Yes.

 

That non-destructive camera RAW philosophy.

 

P

 

 

 

Crap. That's news to me. What's up with that? So much for all the hoopla. Can you elaborate, or paste some links for me to delve deeper into this? Are there any settings at which the frames are 1:1 4K Camera Raw (for real) ?

 

 

K

[Phil Rhodes]

> Crap. That's news to me.

 

I am constantly flabbergasted at the number of people Red have managed to completely hoodwink with this; half the planet seems to think Red is the second coming of deity-of-choice and have somehow obtained a waiver to the laws of physics. Gah!

 

OK, rant over. They're using wavelet compression at about 12:1. Wavelet compression is clever; cleverer than the compression used in something like JPEG in that the image just becomes progressively softer rather than blockier, but it's not a panacea that automatically makes compression OK - actually it just makes it harder to evaluate.

 

Compare:

 

- Red claim to be recording several minutes of 4K images to a $200 compactflash card.

 

- Something like a Viper, Sony F23, F35, an Arri D20 or Panavision Genesis records images of lower claimed resolution and frequently use large, expensive recording devices involving hard disk arrays, HDCAM-SR tapes or hideously expensive multi-hundred-gigabyte flash devices.

 

Now consider: either all these huge companies overlooked something which makes it somehow possible for Red to record "raw" images to low cost hardware, or they're just using a gigantic technical cheat for the sake of convenience. Red themselves claim that by "raw" they mean that they have not demosaiced the image coming off the camera's single chip, and this is entirely true; although what it actually means is another technical convenience for Red inasmuch as you are required to do what would traditionally be in-camera processing on computer equipment you provide, which is very slow and cumbersome. But that consideration aside, the use of the word "raw" to describe images which are anything but is, in my opinion, an outright lie designed to cause exactly the sort of confusion and overestimation of the abilities of the thing which you have suffered.

 

> Are there any settings at which the frames are 1:1 4K Camera Raw (for real) ?

 

No. There was to be a raw port, but was cancelled. My feeling is that it would have revealed too much about the shortcomings of their image sensor, which is based on an IBM stills camera design and may be being driven perhaps somewhat faster than its constituent parts were ever designed to go, but that's necessarily conjecture.

 

It is not really capable of 4K images in any case. In more conventional terms, what they have is a 12 megapixel sensor which is about 4000 pixels across. DSLR stills cameras are generally rated in megapixels specifically because of the difficulty of relating a bayer-pattern imager of given dimensions to a demosaiced, viewable RGB bitmap. Again, I consider this mixing of terminology to be misleading at best; they're using terminology intended for imagers with co-sited or virtually co-sited RGB pixels, such as a film scanner, to describe an imager which is nothing like that.

 

The upshot of this is that the thing presumably has an output resolution of between 2.8 and 3.2K, and I allow the 3.2 figure only on the basis of people I have respect for having done tests; the standard rule of thumb is to divide the number of photosites by 1.4 and this predicts a yeild somewhere under 3K. It is possible to trade off aliasing and colour fringing for apparent resolution and it may be no coincidence that all the published tests to date have used black-and-white test charts only.

 

So yeah, the $17,500 4K digital cinema camera is too good to be true. Sorry.

 

P

[David Mullen ASC]

Flashing at best can only bring up information that was one stop below the threshold of being captured, it doesn't do squat for anything a couple of stops darker than that threshold, plus most digital cameras like the RED have pretty good low-end detail, sort of buried in noise though -- flashing doesn't help too much in that regards, not if eventually you want the blacks to be black in the final color-correction. It may help slightly for some dim areas.

 

Where it would be more useful is in high contrast day situations, just as the DigiCons could help, which is lift the shadows slightly so you can underexpose the highlights slightly and perhaps gain another stop of information to work with. But that's it, maybe you gain a stop of info, IF that. Could be more like a half-stop.

 

There's no free lunch.

[Chris Kenny]

OK, rant over. They're using wavelet compression at about 12:1. Wavelet compression is clever; cleverer than the compression used in something like JPEG in that the image just becomes progressively softer rather than blockier, but it's not a panacea that automatically makes compression OK - actually it just makes it harder to evaluate.

 

What makes compression "OK" is that it's often better than any of the realistic alternatives. If you're going to discard information, a modern compression algorithm does it in a far smarter way than downscaling, or cramming your image into an 8-bit color space, or subsampling chroma. The Red preserves a larger amount of significant image data by recording wavelet compressed 4K bayer sensor data than it would downscaling to 1080p in-camera and recording an uncompressed 1080p RGB image -- and still manages to keep its data rates significantly below what uncompressed 1080p would require.

 

Of course, you would get a slightly better image if you recorded uncompressed 4K bayer data. But instead of recording to a $200 card a bit larger than a postage stamp, you'd have to record to a $20,000 RAID the size of a mini-fridge. Is that really a tradeoff you'd make on-set, for an image that would be indistinguishable under normal viewing conditions?

[Phil Rhodes]

Only once you can prove to me that "normal viewing conditions" is a phrase with a widely-accepted meaning...

[Chris Kenny]

Only once you can prove to me that "normal viewing conditions" is a phrase with a widely-accepted meaning...

 

I'm using it to mean pretty much anything other than extremely close viewing for the specific purpose of spotting artifacts.

 

Take a look at this image. I started off with an 8 megapixel DSLR image shot in raw mode (no lossy compression), and converted it to a TIFF via Adobe Camera Raw in Photoshop CS3. The TIFF file was 23 MB. I exported it to a JPEG at a reasonably high quality setting, and ended up with a 2.1 MB file -- about 11:1 compression. The file at that link is a losslessly compressed PNG file showing a 100% crop of the same area of the image from the original TIFF and from the JPEG.

 

Now, I'm quite familiar with what JPEG compression artifacts look like, and I can spot some in the JPEG version of the image... with the image displayed at 100% pixel size on a high-end LCD and my face 12" from the screen. But keep in mind:

 

1) Redcode is derived from JPEG200, which has less noticeable artifacts than JPEG.

2) I'm looking at a ~100 ppi display. A 4K display at ~100 ppi would be 41" wide -- 12" would be a comically close viewing distance.

3) It's true the images wouldn't look quite as similar if the source image was razor sharp. But bayer sensor images aren't all that sharp at 100%, so that's not relevant to this application.

 

Is a smaller quality difference than the difference between those two images really worth having to wrangle 12x the data? In my view, in almost all cases, the answer is a resounding "no". I'll note that every movie isn't shot on 70mm anamorphic -- despite that making a far larger difference than (well) compressed vs. uncompressed. At some level, you pass the point of diminishing returns. Fully uncompressed video acquisition is well out in diminishing returns territory with the technology available today.

 

BTW, here is a scaled version of the full image, if you're wondering how much detail there was in the whole frame.

[Neil Duffy]

To me the most important thing is the dynamic range and frame rates. Anything over 2k is sufficient for almost any task today.

Edited June 23, 2008 by Neil Duffy

[Max Jacoby]

1) Redcode is derived from JPEG200, which has less noticeable artifacts than JPEG.

Surely you must be mistaken, Redcode is completely new and innovative and blows any existing compression out of the water ;)

[Freya Black]

To me the most important thing is the dynamic range and frame rates. Anything over 2k is sufficient for almost any task today.

 

That's probably a good attitude to have. I wonder sometimes when the obssesion with resolution will end. Many films have had 2k DI's in the past. Obviously that's becoming seen as being a bit nasty these days but while it was fairly standard people were somewhat okay with it. The Red is really intresting because it appears to give something around the 3k mark, so it already exceeds the 2k data resolution. However this isn't enough for some people who are aware that film can do a lot more in theory. I wonder if people would be happy if they had an 8k video camera, or something that really obviously exceeded the resolution of film. Would people then actually stop and ask what does this really mean?

 

To be honest you can probably get away with something well under 2k. George Lucas did.

 

However my impression of what Phil has been saying is not so much about there not being enough resolution but about ideas like honesty, and the truth and false advertising and misleading people. It's more about people having a real understanding about things, perhaps even some kind of scientific imperative.

 

It's not so much about what you can do with the camera or if it's good enough for a paticular job but about having some kind of understanding about it.

 

Sadly things here have been nasty and murky in the past and I'm really glad to see things here have been slowly finding their way to some sanity.

 

love

 

Freya

[Phil Rhodes]

Ms. Black's interpretation of my position is correct.

 

Rumour has it that "Redcode" actually is JPEG2000. There's no reason for it not to be, which rather begs the question why they aren't being more upfront about it. If they'd used a widely standardised format they might not have the post issues they do now.

 

P

[Keith Walters]

Ms. Black's interpretation of my position is correct.

 

Rumour has it that "Redcode" actually is JPEG2000. There's no reason for it not to be, which rather begs the question why they aren't being more upfront about it. If they'd used a widely standardised format they might not have the post issues they do now.

 

P

 

The data from a Bayer Sensor contains a lot of strongly repetitive content which would be highly amenable to compression, so I could allow the possibility that it is subjectively more efficient than say, De-Bayering and then compressing.

 

Besides, if they did it that way, there is no way they they could record on cheap semiconductor memory, which is a potentially major cost-saving feature. However, it is all rather starting to sound like what happens when you over-clock a PC; it sort of works, but you can't depend on it.

 

But you are correct, the three most important aspects are workflow, workflow and workflow.

(And dynamic range, dynamic range, dynamic range, and power consumption).

 

I notice the D-21 has the full gamut of output options, INCLUDING S-Video and composite, (meaning you can make nice DVDs for the clients). AND it has a respectable dynamic range. AND reasonable current consumption. AND it can use anamorphic lenses. I'm glad somebody listens.

[Chris Kenny]

I notice the D-21 has the full gamut of output options, INCLUDING S-Video and composite, (meaning you can make nice DVDs for the clients). AND it has a respectable dynamic range. AND reasonable current consumption. AND it can use anamorphic lenses. I'm glad somebody listens.

 

It also weighs twice as much, doesn't support frame rates as high in data mode, is lower resolution, and has no on-camera recording, and is somewhat outrageously expensive. I won't say it's not a better option for anyone, but it's clearly not a better option for everyone. And many of the things that make it less practical than the Red for some projects are pretty direct consequences of the fact that it works with uncompressed data.

 

People really need to understand that the world is more complex than "compressed = bad". There are a whole serious of trade-offs involved, and the sweet spot for all of the relevant considerations most often involves some level of compression, with today's technology.