How does RED stack up to the F900

[I Bloom]

Like I don't know the answer to this question already,

No instead I'm looking for some help in communicating the advantages of RED clearly to non technical clients so for example can I say:

1. It shoots a 4K Bayer Pattern image. This means each frame of a movie shot with this camera is larger than 4000 x 2000 pixels in size, equivilent to an 8 megapixel digital still camera. Roughly 4 times the size of conventional High Definition Video.
2. The image is captured by a single sensor that is equal in size to a Super35mm film gate. It also mounts standard PL mount 35mm lenses. This means that all the depth of field characteristics of 35mm film, as well as the advantages of PL mount cinema style lenses are readily available.
3. The camera has over 11 stops of dynamic range. Twice that of conventional High Definition (????). Making it one of the first digital cameras that is suited for shooting exteriors in bright sunlight.
4. RED has the option to use an advanced type of compression called Wavelet transform. This technique is more effective at reducing the amount of data load during post production while maintaining a higher level of percieved visual quality than conventional digital video compression.
5. RED shoots a 12bits per channel in RGB color space. This makes it possible to do incredible color correction work in post that was previously only possible with scanned film.

I'm thinking about wording and about accuracy. About using phrases like 4 times the size of etc. I'm just interested to hear your thoughts.

IBloom

[Michael Schrengohst]

Someone I am sure that is more in the know than me can shed some
light on this. The reason we are going for the RED is three fold.
1.) Higher resolution 2.) IT based workflow. 3.) The ability to record
RAW. I also shoot Digital stills and shooting RAW affords you an image
that you can process in post more so than you could with HDCAM.

HDCAM is tape based and some producers want to cling to that
for a comfort factor. But once the RED starts showing footage
it will be hard to want to keep shooting tape (And 3:1:1 at that)

[David Mullen ASC]

I wouldn't say that the RED camera has twice the dynamic range of an F900, but it certain has more, which makes it more practical and more aesthetically pleasing, closer to a film-like response to light. I'd say that the F900 was around 8 or 9 stops or so, compared to RED's 11-ish.

You didn't mention cost in all of that -- it seems like even if you "just" wanted a final 10-bit Log 4:4:4 HD image (which the F900 can't do), the RED is possibly the most affordable way of getting that. The F900 can't even do 10-bit Linear 4:2:2.

Once the 12-bit RED RAW image is de-Bayered, it will probably be 10-bit Log RGB, which is basically what a film scan would be. So the comparison to the F900 would be more like 8-bit linear 3:1:1 versus 10-bit Log 4:4:4 (RGB), at whatever resolution you wanted up to 4K.

Also, the F900 can't shot faster than 30P or 60i.

The simplest way to put it is that the RED is much closer to 35mm quality than the F900. If your clients or producers want 35mm quality, they are going to have a much easier time achieving it with something like the RED than with the F900. And it may even cost them less than the F900 to do it.

I mean, even IF they only wanted to end up with an HDCAM tape, they'd still get better quality with the RED.

[Chris Gearhart]

Yeah, have them shoot onto HDCAM SR for 4:4:4 to help them get over the inertial hump, then tell them you can cut out the ingest step after a few shoots and improve the image to boot.

[Graeme Nattress]

F900 on HDCAM tape is not 8bit linear though. It's 8bit gamma encoded, which is an entirely different beast.

We have three descriptors for the encoding:

Log = a log curve, like gamma in a way, but more extreme
Gamma = a power law curve, usually around 2.2, often with a small linear segment near zero like defined in REC709
Linear = linear light, no curve at all

If we adopt the habit of referring to gamma encoded video as "linear", we only confuse matters. Let's leave the word linear to mean a straight line transfer curve.

Graeme

[Michael Hastings]

F900 on HDCAM tape is not 8bit linear though. It's 8bit gamma encoded, which is an entirely different beast.
Graeme

Is that a good or bad beast?

[Graeme Nattress]

A much better beast. All video is generally gamma encoded though - would hardly work without it!

[David Mullen ASC]

This is where it gets confusing, terminology-wise -- isn't "LOG" then just another form of gamma encoding? So how do you describe the difference when using the Viper or F23 is "LOG" mode versus... "what" mode? What's the opposite term for "LOG" in context to the two modes that these cameras shoot?

[Graeme Nattress]

LOG is a different mathematical curve, but yes, it does tend to look similar, if more extreme than gamma encoding.

I'd describe them as using LOG mode or "video gamma" mode.

Graeme

[Chris Gearhart]

Well you learn something new every day.

So the Viper's filmstream mode is nothing more than a gamma curve?! It sure doesn't look like one--very flat. Are you sure it's not something more data-transfer-esque?

Or am I just way misunderstanding things?

[Graeme Nattress]

Sure, go and read the filmstream specs. It looks flat as you're recording the entire dynamic range of the camera, whereas most cameras in their video gamma mode put a lot of "contrast" into the image with their transfer curve to make it look pretty, which is fine, unless you want the whole range for grading later.

[Chris Gearhart]

Yeah, I thought of that after I wrote it. So, it's more a matter of how the range is "stored" in the signal--how much space is given to each particular stop?

Edit: Sorry to get the thread off track a bit. But it does help one understand just how short the F900 will fall with respect to dynamic range compared to the F23, or the Viper or Red, with the latter excelling in almost every other area above the former two.

[David Mullen ASC]

A video camera naturally produces a linear signal off of the sensors; normally it is processed (gamma encoded) to look correct, contrast-wise, for broadcast and for viewing on a video monitor.

The Viper and the F23 processes the signal to emulate the look of a LOG file of a film scan, which is much flatter, because that's better optimized for a transfer to film within the limitations of 10-bit. But that's why you need to find a way of converting the LOG image being recorded into something that looks correct for monitor viewing on the set and in editorial, usually by use of a LUT somewhere in the chain. Or use a camera that sends out a gamma encoded signal to the monitors but a separate LOG signal (or RAW in the case of RED) to the recorder.

Even if you shoot film, ultimately you won't be broadcasting a LOG image, but one where the dynamic range has been compressed into broadcast specs for electronic display, with the proper contrast. But if you start out with more information to begin with, you can create a better image for broadcast, hence the argument for using the RED camera over the F900 even if you just end up with a 4:2:2 master.

When you have an 8-bit format like HDCAM, there are problems with trying to squeeze more stops of exposure information into 8-bits, because each step has to cover a larger exposure change, thus creating banding problems over smooth gradients.

[mdo]

The eye can see a 1% change in light intensity. With 8 bits, that gives you a 2.5:1 ratio, or 2.5 increments in which to encode that 1% shift. Theoretically possible with a perfect gamma curve, but not really possible as a practical matter. If you miss on the gamma curve even a smidgeon, you've got banding. With 10 bits, the ratio is approximately 10:1, which gives 4X more room for conversions and adjustments. Far less chance of having different values appear to be the same due to lack of bits.

[Antoine Fabi]

Hi Graeme,

Very interesting..really.

So i assume that recording a "pretty" gamma curve is essential if you record in to 8 bit codec, to prevent the image from stepping gradients in post.
With a 12 bit codec...well, you have plenty of margin to keep a smooth image, smooth gradients even after intensive CC.

But how can we compare them visually (linear, log) ?

[I Bloom]

I wouldn't say that the RED camera has twice the dynamic range of an F900, but it certain has more, which makes it more practical and more aesthetically pleasing, closer to a film-like response to light. I'd say that the F900 was around 8 or 9 stops or so, compared to RED's 11-ish.

How does RED stack up to various DSLR's in terms of dynamic range?

[David Mullen ASC]

[QUOTE=Antoine Fabi;56764]
So i assume that recording a "pretty" gamma curve is essential if you record in to 8 bit codec, to prevent the image from stepping gradients in post./QUOTE]

Plus they lop off some of the dynamic range to keep it within a range that the processor / codec can handle. And there are still some banding problems that pop up with HDCAM anyway...

There are quality limits to 8-bit that are hard to get around.

A Log curve helps to allow a wider dynamic range to be handled in post, hence why 12-bit RAW is usually converted to 10-bit Log RGB, more or less similar in dynamic range.

[Chris Gearhart]

If I recall, DSLRs used to be limited to around 5-6 stops. Don't know where they are now, but I would imagine they are still under 8 stops tops.

Any still photog gurus?

[J. Bernard Vallon]

If I recall, DSLRs used to be limited to around 5-6 stops. Don't know where they are now, but I would imagine they are still under 8 stops tops.

Any still photog gurus?

The numbers ive seen are way above 5-6.

Remember your bit depth of your RAW file must be at least as big as your DR, but too much bigger wastes data space.

I use a d2x, and its definitely on the low end of DR, probably around 10 stops (12 bit). I've heard the fuji s5 hits greater than 12 stops (14 bit), and the more expensive cannons are up there as well. Evin i'm sure could give us more accurate numbers.

[Chris Gearhart]

Ahh. I was assuming 8-10 bit max. Makes sense, though, as technology barrels along.

[Gavin Greenwalt]

You can pretty safely assume that whatever the best digital cinema camera is achieving, still cameras are a bit ahead. It's a lot easier to eek performance out of a sensor that only needs to shoot 5 frames a second than 60.