Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Creative Cow Article: Panavision, FUD and...HDR?

Since when did the viewing resolution have to be the acquisition resolution? That was never the case with 35mm. How can we ignore that aspect? It's always been the case with video, and it's always been a problem - never any room or resolutionto crop / move / etc.

Say, sitting at home now, I sit on the edge of being able to fully see what my 50" 1080p plasma screen can resolve. It's a nice image. But does it fully fill my FOV? To do so, I've have to sit uncomfortably close to it. Now say I had a 4k 100" screen (not likely to happen, but this is a thought experiment) I can now have a more immersive FOV, and still be able to fully see all the detail it offers.

Graeme
 
Unless I'm reading this wrong, their new sensor has a 60dB SNR... the RED ONE's sensor is 66dB. No word on the newer sensors from RED yet, but I must imagine that it's higher.

Obviously SNR isn't the only determining factor... but it is a big one.
 
I love Creative Cow, but keep in mind who their biggest advertisers are. You know they must get a ton of pressure to "take these Red upstarts down a notch." :)
 
ronx said:
6k is better than 4k, period, whatever method is used, fact. 28k will be better still no doubt..

And the sensor layout used, a secondary issue in comparision..
Click to expand...

I disagree.

Lets say the magic delivery medium is 4k.

The Genesis RGB strip and pixel binning technique currently uses 6k camera to delivery a 2k image. Continuing on the same path they will need a 50mega pixel 12k camera to delivery a final 4k image. The photo-sites will be very small when crammed into a S35 space. One of the nice things about the Genesis is how it handles low light which will be reduced by this cramming.

Now with a Bayer pattern, in order to capture the 4k MTF then you are looking at roughly a 6k sensor. I feel than a 6k CMOS will happen long before a 12k CCD.

We can argue which is better. My point is you said:

ronx said:
6k is better than 4k, period, whatever method is used, fact. 28k will be better still no doubt..

And the sensor layout used, a secondary issue in comparision..
Click to expand...

And in my opinion this is not true.

It's a shame that Panavision feels the need for smear tactics. All that really needs to be said is, "This is how we do it, this is how they do it, this is why we think our way is better."

Subversive tactics designed to breed rooted animosity with out learned decision making is just regretful.

The sad part is, it's not their cameras that have turned me off to them. It's their attitude. I have shot many a job with Panavision cameras. However, seeing the way they have conducted themselves as a company makes me want to stay away.

HDR is going to happen. The fact that Pani is looking into is good.
 
5k is sufficient, with a decent OLPF and demosaic to have a full measured 4k. 6k just adds gravy :-) Or more room to move. How short sighted is it to think that the acquisition resolution can be as small as the delivery resolution?

Graeme
 
J. Eric Camp said:
I disagree.

Lets say the magic delivery medium is 4k.

The Genesis RGB strip and pixel binning technique currently uses 6k camera to delivery a 2k image. Continuing on the same path they will need a 50mega pixel 12k camera to delivery a final 4k image. The photo-sites will be very small when crammed into a S35 space. One of the nice things about the Genesis is how it handles low light which will be reduced by this cramming.

Now with a Bayer pattern, in order to capture the 4k MTF then you are looking at roughly a 6k sensor. I feel than a 6k CMOS will happen long before a 12k CCD.

We can argue which is better. My point is you said:



And in my opinion this is not true.

It's a shame that Panavision feels the need for smear tactics. All that really needs to be said is, "This is how we do it, this is how they do it, this is why we think our way is better."

Subversive tactics designed to breed rooted animosity with out learned decision making is just regretful.

The sad part is, it's not their cameras that have turned me off to them. It's their attitude. I have shot many a job with Panavision cameras. However, seeing the way they have conducted themselves as a company makes me want to stay away.

HDR is going to happen. The fact that Pani is looking into is good.
Click to expand...

You've totally missed the point of my post, and quoted the same bit twice bizarrely? I know bayer is better value for your buck, and I understand the bigger pixels argument too, but..

My point is, bayer is fine, "now give us access to it and let us debayer and archive it how we want"

Graeme Nattress said:
5k is sufficient, with a decent OLPF and demosaic to have a full measured 4k. 6k just adds gravy :-) Or more room to move. How short sighted is it to think that the acquisition resolution can be as small as the delivery resolution?
Graeme
Click to expand...

Exactly, I've never thought of the R1 as 2k,3k or 4k. It is what it is.

IMO , 4k bayer is great for getting a 2k or 1080 image from, I couldn't care less what that guy from Panavision thinks, I just wish Red would feel the same way, then give us our RAW data..

I made my mind up by shooting, analysing, and programming..

I remember Graeme or Rob telling people to downsample their 4k to 1080p and apply gaussian blur, then downsample that to SD, again with blur to get acceptable downsamples. That must be especially fun with products that only allow users to choose radii at 0.1 increments, but anyway.. I realise why this was adviced, but I also "laughed my tits off" that this was the current situation.

Thats when I realised, we should be allowed to disagree and take our own route..

You can feed my downsampler a 28k image today btw, and it will give you SD (or any resolution, odd or even, down to 1pixel squared). Every one of those 250mp would get used, with the gaussian weighting calculated automatically, all done in floating point. And 4kHD mode, wtf, shoot at a lower resolution so that realtime hardware can handle the downsample.. well I guess some people just use Pro-res to master from too, time is money, I can understand that, but I don't have to believe its the only way to go..

If I wanted a really good laugh, I'd compare the output from my downsampler to what Photoshop produces.

If you want a nice interface, multiple image formats supported with hardware preview support, and file browsing that doesn't get confused when a multi-CF card reader gets plugged in after the program is loaded, then my software becomes the joke.. but I use it to downsample the only output I can get from my Red, TIFFs..

And no, I don't know if the Genesis records audio, but if it does, you can probably export it..

What someone at Panavision wants to do, or say, is fine by me. I agree with much of it, and people do need educating possibly, but then again, many won't care.

Panavision don't sell cameras.

The Genesis has more Photosites than the Red, the SI2K ("Different Product" as Jim once described it) has an affordable RAW meta-data driven workflow, both facts IMO
 
Ronx, the advice to blur before scaling is because many apps that people do scaling in, like FCP, have in-adequate downsampling filters. If you scale in REDAlert, for instance, the scaling filters are quite nice, and you can choose which you like. Or you could use Shake which has nice, selectable downsample filters.

I don't mind you laughing your tits off, but please direct your humour to where it belongs, which are the NLE and compositing apps or whatever that have inadequate downsample filters. :-)

Graeme
 
Graeme Nattress said:
Ronx, the advice to blur before scaling is because many apps that people do scaling in, like FCP, have in-adequate downsampling filters. If you scale in REDAlert, for instance, the scaling filters are quite nice, and you can choose which you like. Or you could use Shake which has nice, selectable downsample filters.

I don't mind you laughing your tits off, but please direct your humour to where it belongs, which are the NLE and compositing apps or whatever that have inadequate downsample filters. :-)

Graeme
Click to expand...

Your doing it again. Why are you telling me what I already know? Why are you defending yourself, my humour was directed exactly where you suggested it should be? I know why you adviced it, as i said..

This ceaseless handholding must stop..

Yeah, "quite nice" downsamplers are, er "quite nice"
 
The internet as a forum for communication is limited. The written word does not read like the person speaking the same words. I'm happy where your humour was directed. It is the correct place.

Graeme
 
Graeme Nattress said:
In a RGB stripe CFA camera, we also know that the three colours of photosites are not co-sited, but that is often ignored. The spatial separation is not used to extract real resolution from the data.

Graeme
Click to expand...

I watched a demo of the Sony F35 camera at the Nor-Cal Digital Cinema Society meeting in December. Although it does produce excellent images with really good dynamic range and low noise, there are artifacts in tight gray pattern such as a wire mesh strainer. Normally, not an issue for most people. However, the recording is limited to 10 bit HD-SDI. You cannot record to a RAW format, so you must specify burn in the gamma curve that you want at the time of recording. It's also a $200,000 camera without glass, so you could buy quite a bit of Red gear for the same price.

Because the camera does show color fringing on very fine grey patterns, it's clear that this RGB strip patterns aren't perfect. It could be fixed with extra image processing inside the camera, but more image detail would be lost.
 
Thanks for that info Brad. I've not been able to point a Genesis at a zone plate to see how it behaves. Perhaps one day someone will do that for me, and I can really see how the RGB Stripe CFA behaves.

It's funny, but the "problems" with CFA sensors are the "problems" with CFA sensors. There's always a compromise in the pattern you choose. Whatever pattern you pick there are tradeoffs.

Graeme
 
The Epic647 sensor size and resolution should allow wire mesh strainers to be resolved from 100ft away :)

Can't wait till a screen grab is posted of that one..

I scanned a spider directly using a scanner at 30,000 pixels square about five years, it obliged by staying very still for me, it wasn't very big.. framerate wasn't brilliant, and a definite moire..

2958_1233765627.jpg


photoshop downsample didn't help, but the only app. i had that could open it..
 
Graeme Nattress said:
Thanks for that info Brad. I've not been able to point a Genesis at a zone plate to see how it behaves. Perhaps one day someone will do that for me, and I can really see how the RGB Stripe CFA behaves.

Graeme
Click to expand...

One more very important thing. The F35 seems to have Sony's very distinctive over saturated color look. You see it across all the Sony cameras and is somewhat in movies shot using Sony cameras. I don't know how of this problem can be corrected in post but there are only 10 bits instead of 12 to work with. Over saturation is a bad thing for preserving image detail.
 
The practical thing to do, IMO, is to do your own test.

Point both cameras at a test pattern and/or real world scene. I suggest an eye chart test... because it's a reasonable test for resolution and resembles what you'd shoot in a movie. Zone plates are a little extreme... similar patterns rarely occur in what you shoot (except in fabrics, but the contrast is not as high). Or just be thorough and shoot all that stuff in your test.

Then, look at the images. There are going to be things that don't necessarily fit theoretical models about how the images "should" behave. Artifacts from the optics of the system (almost every system has color artifacts in practice, including 3-chip), any artifacts from the demosaic process (esp. if the algorithm is not that good), artifacts from the sampling structure, etc. etc.
Also look at how the cameras resolve text and fine detail- this will give an idea about what the real resolution is, and you can't game that test by doing sharpness tricks (which may be why HDCAM looks very sharp).

Would also be useful to test keying ability (and secondary color correction).

2- I suspect that if you ran the tests above, you would find that Red is noticeably better in resolution. (I'm not sure about keying... Red may or may not be better.)
 
BradWright said:
One more very important thing. The F35 seems to have Sony's very distinctive over saturated color look. You see it across all the Sony cameras and is somewhat in movies shot using Sony cameras. I don't know how of this problem can be corrected in post but there are only 10 bits instead of 12 to work with. Over saturation is a bad thing for preserving image detail.
Click to expand...

Actually you're missing a few numbers. Sony's 10 bit uses three photosites per pixel instead of 1 like RED. So it works out to 30 bits compared to 12 bits. F35 actually oversamples at 14 and converts it to 10 for recording. In short, 1.9 MP, Three Channels, 10 bit per channel pretty much meets Kodak"s DPX Standard for making DI's.

F35 take s 30 bit sampling (over sampled at 42) spread out over 1.9 MP while RED has a total of 12 bit subsamplng over 4k. Sony records at 440 mb or 880 Mb, compressing at 4:1 or 2:1. A DPX RED frame is 36MB. Red shoots at 1.5MB/frame. That's a compression of 24:1 or if you compare their record rate to the sensor it's 12:1. The former being a standard way of calculating compression because that's what you're actually going to.

Perhaps that saturated look you talk of is just more color info while RED's less saturated look is a thin somewhat overtaxed color space.

I can't wait for the new upgraded chip with it's higher bit rates. Hopefully, they'll focus less on rez and more on color, compression and noise. Only so much a codec can do.
 
Um, the more saturated colours appear, the smaller the colour gamut..... Just like the more contrasty the image, the smaller the dynamic range.
 
Sony's 10 bit uses three photosites per pixel instead of 1 like RED. So it works out to 30 bits compared to 12 bits
Click to expand...

Bit depth is just luminence values for that pixel...Sony only records 10bits...Red 12bits

Bit depth dosn't matter how many photosites per pixel, color resolution may be better with 3 photosites per pixel....but bit depth is on a per pixel basis.

Basically 10 bit would be 1024 values of gray for that channel.
And 30 bit would be an astronomically higher value of grays which I'm pretty sure the Sony cannot distinguish.
 
You must log in or register to reply here.
Back
Top