EX1 and RED playing nice together...

So one last question Jim. For keying footage in a controlled environment for a HD deliverable does the EX1 give you quality keys 90% of the time compared to more expensive options?
I really appreciate your test!
 
Nik Manning said:
So one last question Jim. For keying footage in a controlled environment for a HD deliverable does the EX1 give you quality keys 90% of the time compared to more expensive options?
I really appreciate your test!
Click to expand...

Hi Nik... you pre-qualified your question just enough that I can answer with a resounding YES!

Sound general thoughts now that I've had the EX1 a bit...

1.) A wish the lens at full wide had a simpler distortion profile... it's not a simple first order distortion, and requires a bit of effort to un-distort the footage if you need perfectly rectilinear images for motion tracking. Having said that, I worked with a 38K Canon zoom on an F900 a few months ago that was just as bad at about the same FOV. And that camera/lens combo needed constant checking for backfocus drift every time the camera had been on for a long time, or suspended in a different configuration.

2.) The on-board codec is WAY better than HDV, and keys nicely with a little chroma bluring. It's about on-par with HDCAM, IMO. I said that early on and still stand behind it. The HD-SDI out pushes the camera into another arena, as shown in this test with the RED. I relish the chance to work with the full 10bit 109 IRE potential with something like Decklink CineForm or Prores. Or the flexibility of the Convergent Designs XDR unit.

3.) I HATE the button and menu selection on the EX1. Having said that, we spent a good half hour trying to replace the joystick knob on one of the three REDs. That little sucker has poorly engineered .9mm allen head set screw holding on a very important gidget.

4.) The EX1 gives you no "wiggle room" for compositing if you need to reposition or reframe. The RED does. You can resize the RED footage quite a bit, do things like that lens undistortion filtering, etc. and the image will hold up for a 1080 or 2K post workflow. Of course the same is true not only for the EX1, but any 1080p camera at any price. At least it does fully fill that 1080 full raster with image. The HVX200 struggled to fill a 1280 by 1080 raster with detail, and my tests indicated it only had 1280 by 800 or so actual resolution.

5.) The EX1, like any other video camera, has a fragile image that has limited room for color correction. We can debate the stop range of the RED and EX1, and I think properly handled they're similar, but the RED pushes all those decisions off to post that will impact the usability of that dynamic range. The EX1 bakes it in and you're stuck with it.

Regards,
 
Jim, what about the shutter speed.

Some people suggest to use higher shutter speed for bluescreen work, to eliminate motion blur as much as possible.

Do you find this has any significant impact, in practice?

and... how come the hair (the toughest part to key) is keyed so perfectly on EX1 image (right side), vs. showing a very visible red edge in RED footage (no pun intended)?

Cheers,
 
Nothing makes a key look more fake than a hard edge - motion blur is very important to a convincing key.

Therefore it's important to use keying software than handles transparency and softness well. Keylight is my favorite choice.
 
Jim Arthurs said:
5.) The EX1, like any other video camera, has a fragile image that has limited room for color correction. We can debate the stop range of the RED and EX1, and I think properly handled they're similar, but the RED pushes all those decisions off to post that will impact the usability of that dynamic range. The EX1 bakes it in and you're stuck with it.
Click to expand...

Thanks for pointing this out, Jim. As the owner of the RED camera used in this comparison, I just want to put some brackets around this statement.

This test is most useful as a keying test, but I think when you see the two images side by side like this, it is very tempting to jump to conclusions about overall image quality, and I would really caution against making this mistake. That is a poor use of this test.

The RED doesn't have in-camera image quality adjustments -- it is all expected to happen in the RAW post-processing. That's the point of RAW -- total image flexibility. And this is where the image really shines. This is just the unprocessed negative, the RAW image, here. There's no in-camera gamma or black or knee adjustments, like you are seeing with the EX-1.

All of the secret sauce is added later, and none of it is added here.

At the end of the day, these are both excellent cameras. And I think anyone who finds the price of the RED a bit too rich, would be well served with an EX-1 or -3 purchase. But I'm rapidly falling in love with the RAW workflow and and total image flexibility it affords. For image purists, there is no better camera, in my opinion. But for the money, the EX cams are a great value. For a versatile, efficient lightweight 1080p camera, it is great.

But if the higher resolution rates and the revolutionary flexibility of shooting RAW is important to you, it can't match the RED.
 
Well, one thing is for sure, the EX1 is the ultimate RED add-on! (seen here setup to test rolling shutter shear between the two cameras)

EX1_piggybackRED.jpg


Planet E is on target with her thoughts on the whole workflow issue.

I would mention that I did add the ubiquitous light "S" curve to the RED footage in REDCINE, else it wouldn't visually look as expected. And, as I have mentioned before, I could have eeked out a bit more dynamic range from the EX1, using 100 + IRE information and recording the camera's full 10bit vrs the 8bit my capture card limited me to.

Although I've already put this pictures out there with earlier links, folks might not want to wade through high megabyte layered Photoshop files to check out the original frames. Here they are as 100% Jpegs;

RED; downsized, Mitchell filter from 4K to 1080p.

http://ftp.datausa.com/imageshoppe/outgoing/EX1/RED_downsized1080_LUCKY.jpg

EX1; From 8bit uncompressed.

http://ftp.datausa.com/imageshoppe/outgoing/EX1/EX1_LUCKY.jpg

As I have mentioned before, you will note that the RED image is cleaner in all color channels, and that the blue channel doesn't display some of the 4:2:2 issues that the EX1 does. The RED image also benefits from the greater than 2:1 downsizing to increase "natural" sharpness and fruther reduce noise. Check the RED's red channel and you can see fine hairs that are simply not there on the EX1's red channel.

And as mentioned before, these were shot under daylight balanced lighting, which is RED's real playground, in terms of a noise free image.

Finally, folks have been asking how well the EX1's native codec fairs against the uncompressed footage. I haven't had time to process the full motion, but I have set up the key and here is how that looks;

http://ftp.datausa.com/imageshoppe/outgoing/EX1/LUCKY_EX1_uncomp_vrsCODEC.jpg

I did do chroma smoothing to blend out the 4:2:0 steppy chroma, but the results aren't bad. I will process the full shot and share when I have the time.

Regards,
 
Jim,

How much difference would it make on a blue or green screen to be shooting 3200 instead of daylight. It's far more cost effective lighting wise to me and I have until now not heard of daylight being quieter. I would have expected the opposite actually. is it different for the Ex-1 or say an HDCAM or Varicam, rather than the Red.

Lenny Levy
 
I've heard the RED is cleaner in 5600K verses 3200K for a while. This was also noted from a friend that has the RED ONE.

Also, I've seen on the RED forum some tests that Jim had performed a while back with the RED comparing 5600K verses 3200K noise levels.

I believe the EX1 noise levels are less at 3200K verses 5600K. It would be interesting if Jim had some data on this test.
 
I'm sure daylight is quieter than tungsten because the blue channel is darker and noisier with tungsten. Since blue is the noisiest channel, I'm sure that's why it's noticeable.
 
Definitely; we've seen tests of this and the difference in noise in the blue channel is substantial. The RED is balanced for daylight. When you light with tungsten lighting, the system has to gain up the blue channel, bringing in a lot of noise.
 
Isn't it actually something like 5000K?

It was weird, the stuff I shot was all balanced at 5600K, HMI and daylight balanced for the most part, but didn't look it's best until 5000K.

That said, the friggin' footage is always incredible from that camera.
 
Here's the EX1 under daylight and tungsten, or rather 5200K and 3200K, as that's what I got as the solution on the white balance. This is a grey card, exposed at exactly 50%. Focus is thrown out so there's no object/surface texture to spoil the test. Uncompressed 8bit as the source. -3db.

http://ftp.datausa.com/imageshoppe/outgoing/EX1/EX1_8bit_5200K_VRS_3200K.png

The top half of each image is a very heavy levels adjustment to "pop" out any variations. As an aside, it's clear that the white balance was a bit off on the daylight side, or perhaps my cheaper daylight biax tubes aren't good as the GE 3200K ones are.

Take the image into Photoshop or your picture editor of choice and check out each of the color channels.

The EX1 has less noise in daylight than tungsten, as you can tell by the smoother quality of the grain. Most noticable in the blue channel.

It's not night and day different, and it would be interesting to do this alongside a RED to see how that all shakes out.
 
Kholi said:
Isn't it actually something like 5000K?
Click to expand...

I think you're right, if I remember correctly from the RED forum. And I think the general gist was that it might be a good idea to partially correct toward daylight as a good compromise, using in front of lens filtration. Search on my name and "daylight " and there should be lots of posts over there from last year.

Several of us were very concerned about this topic, most of the heavy hitter folks with RAW experience told us it wouldn't matter... then of course we come to find out it truly is an issue. This is an area where the latest firmware builds are making progress.

FYI, on our bluescreen test shoot; the EX1 and at least one of the three REDs gave us about 5200K as our balance, under the lighting setup we had...

Regards,
 
Thanks for the clarifiying comments about the S curve, Jim. And for the picture of the EX-1 mounted on the RED...I was hoping to get a copy of that. The difference is further dramatized by the size of the Angeniuex 20-120mm lens on the RED, that thing is a beast.

Just to add another note -- much of the discussion of the blue channel noise may also shift once the Build 16 beta is released, because RED has stated that this build will be all about enhancing image quality and since the blue channel has come up repeatedly as one of RED's limiting parameters, I suspect this noise issue will be addressed at that time.

But a daylight balanced camera will still be a daylight balanced camera, so tungsten lighting will probably still not be optimal lighting for RED...we shall see.

We are probably weeks away from this release, so our current experiments may have a short shelf life.

If I bring some more new toys, next time, can I tempt you onto the playground again?...
 
planet E said:
Thanks for the clarifiying comments about the S curve, Jim. And for the picture of the EX-1 mounted on the RED...I was hoping to get a copy of that. The difference is further dramatized by the size of the Angeniuex 20-120mm lens on the RED, that thing is a beast.

Just to add another note -- much of the discussion of the blue channel noise may also shift once the Build 16 beta is released, because RED has stated that this build will be all about enhancing image quality and since the blue channel has come up repeatedly as one of RED's limiting parameters, I suspect this noise issue will be addressed at that time.

But a daylight balanced camera will still be a daylight balanced camera, so tungsten lighting will probably still not be optimal lighting for RED...we shall see.

We are probably weeks away from this release, so our current experiments may have a short shelf life.

If I bring some more new toys, next time, can I tempt you onto the playground again?...
Click to expand...

E, I'm assuming that you're going to make a go for the Sensor upgrade when it's released? Is that supposed to be tungsten balanced?

One thing I can't wrap my head around is using Daylight balanced lights or HMI and gels: how does this affect the RED footage as opposed to just using Tungsten to begin with?

To Jim: What's the best method for Downrezzing your RED 4k 2:1 footage? I've only gone through REDCine to do that and for some reason it's not getting too sharp once you move from 4k to say 1080? It should be pretty sharp, no?
 
Kholi said:
One thing I can't wrap my head around is using Daylight balanced lights or HMI and gels: how does this affect the RED footage as opposed to just using Tungsten to begin with?
Click to expand...
Tungsten lighting + RED camera = noisy footage in the blue channel.

Red is designed to "see" daylight-balanced light. If you feed it tungsten light, you're going to have to compensate electronically. That means adding a whole lot of gain to the blue channel. Which means you have a noisier blue channel. If you just feed it daylight in the first place, that means no gaining up any of the channels and cleaner footage.
 
No nonono I get that. I'm saying if you slap on some CTO, does it get noisier? I'm assuming SO because then you're moving back to the three-thousands as far as Kelvin is concerned.

So basically, any "orange" light sees much more noise.

And, if it goes back the other way and we get a Tungsten-loving RED sensor, does that mean Daylight will be bad for it?
 
Yes.

(but by "bad", we mean a little more noise.) Every sensor is balanced for a certain exposure, and needs compensation (optically or electronically) to deal with light of the other color temparature. That's just the way it is.
 
Jim and others,

My interests here are both economic and technical/aesthetic.

Lighting a stage for green screen takes a lot of light if its more than a talking head.

We tend to use space lights and all sorts of other tungsten to make it work. Switching to daylight could be a significant expense. So what I would like to know is how significant is this. Do I really need to bother?

Does it matter as much in green screen?

A starter question would be - is this the same for other non CMOS chips that we've been using for years - i.e. an HDCAM or for that matter an SDX900 or even a good old Beta D30. I remember a few years ago Barry was suggesting shooting an HVX outdoors with an 85 filter to get it back to tungstun so you didn't add electronic noise from the red channel. In fact all professional video cameras did this. This seems the opposite. I'm confused. Is the Ex-1 as "dayllght balanced" as a Red?

I guess going largely Kino is a a cheap solution though you would still have to gel your space lights somehow. Does anyone make blue spacelight material? Jim your suggestions of adding some blue filtration to the lens sounds like a reasonable idea if we could get the lighting up say an extra stop or so. Still cheaper than HMI's.

Would just a half or even a quarter blue make any substantial improvement?

Now do we need to shoot with a blue filter plus an IR486 to really get accurate clean pics?

Lenny Levy
 
Last edited:
You must log in or register to reply here.
Back
Top