Low light...

[Gary Blake]

Not at all.
The setups are enough similar to get a decent roundabout comparison of noise control between the two, especially in my "video still" comparison where the difference in fstop is in the negligible range.

This was done for everybody saying that the M-X shot easily outdid DSLRs.
That obviously isn't the case concerning high iso noise.

 

[David Battistella]

RED 758 Candle Light test.

So I put a candle in front of my color checker chart today and shot it:
It's pretty amazing to see how much the Jim's recent images blow away the existing camera.

I'll post the R3D's too.

50mm Nikon Prime open to 2.8
ISO 2000 on RED
Build 21.4.1 (NEW Color science)

REDSPACE GAMMA
REDSPACE COLOR

The files marked TUNGSTEN AND DAYLIGHT are snapshots to show you the exact output settings from RED ALERT.

The 4K Tiffs and the r3d's files can be downloaded from here:
R3d's will take an extra half hour.


http://www.f8films.com/client/red/

Username: red
Password: red

Please note that the username and password are case sensitive.

 

[Gary Blake]

I scooted back
And used Tungsten white balance
and shot at 2.8

and the 5D at ISO6400

 

[Michael T O'Connor]

Does anyone with a 7d/5d have a color checker? I think that would create the fairest comparison. The current comparisons are at best comparing apples to oranges.

 

[Gary Blake]

I wish I had a color checker.
But comparing noise in a candlelit photo this is an apt comparison.
Noise isn't going to magically disappear if I shoot a colorful picture of a duck riding a horse.

 

[Ruby Quincunx]

The difference would be negligible for a high iso noise comparison.

But for a high-ISO noise comparison with practical relevance to motion picture photography the difference is not negligible, since you would not be using shutter to control exposure. That's the hidden variable. Hence, apples & oranges.

RQ

 

[Michael T O'Connor]

Where's Sanjin when you need him? I seem to remember him having a 5d or 7d and a color checker....

 

[Gary Blake]

But for a high-ISO noise comparison with practical relevance to motion picture photography the difference is not negligible, since you would not be using shutter to control exposure. That's the hidden variable. Hence, apples & oranges.

RQ

Aperture isn't going to change high iso noise, though. :mellow:

 

[Jannard]

Gary... sorry. I know your intentions are good. As you stated, this is not a scientific test. It isn't even a non-scientific test. How big in diameter was your candle? What type of wax? Lumens? How far away from the subject? Without a light meter to tell you ISO/exposure... how do you know equal amounts of light hit the subject? And what was the color temperature of the light? The lack of light falloff indicates that either there was room light, the candle was stronger and farther away from the subject, or you cropped in. If the latter... how much? 1 stop makes a HUGE difference in noise.

Canon's low light capabilities are great. No question. The video is a line-skipping (read every 4th line) H.264 output.

I posted that we weren't done... and I probably would get in trouble for posting these images. I did. Your post helped me get yelled at. A comparison is not a comparison unless most things are equal. And we have no way to know. Let alone that our image will improve as we are in the middle of development.

Why not wait until we ship some early EPICs and let someone do a real side-by-side test? It will lead to some clear conclusions...

Jim

 

[Scott Miller]

I love when the owner of a company gets yelled at Shows how much faith you obviously have in your team.

 

[Jannard]

David... there is no question that M-X is way better at handling noise than M. However... a couple of things to point out. I downloaded the top r3d and opened in RAND (Graeme's new internal Alpha app). This really is ISO 4000. And it should be balanced for 1940K and 25 tint. Since NR is not a sin, if you add a default amount, you will see a much better image. Certainly not the quality of where we are now with M-X... but not so bad as you posted.

Post tools are there to help you when you need it. It should be noted that all DSLRs use NR and sharpening in camera. The higher the ISO, the more they apply. It makes sense. Why not? Except we do not use either in a RED ONE. So there may be a time when you need to pull them out of the bag.

Jim

 

[Gary Blake]

Both candles look to be regular tea lights and a stop change in aperture makes absolutely no difference in noise if shutter speed is adjusted to keep exposure the same.

But you're right, it isn't scientific. I've scrutinized the noise in tens of thousands of images and can easily tell when a photo from my camera is in a certain small ISO range regardless of lighting conditions and subject.

The fact of the matter is that the lighting is so similar in these two shots that they can be compared. The M-X image looks like a 5DII shot in the ISO4000-5000 range which is just incredible.

I'm not trying to root for Canon or for RED, simply making an observation about the rough high ISO noise capability of the M-X. It looks to be better than the 7D and worse than the 1DIV which puts it about on par with the 5D II. Of course it has better dynamic range and it isn't finished yet, so like I said, no conclusions should be drawn.

I still think photos posted are open to investigation. I'm not going to wait until a final product is released just like I'm not going to make any final conclusions until one is released.

 

[Paul Schneider]

Paul-
No. The Red M-X image is the largest resolution posted and the Canon 5D II image is scaled down a little larger than that and cropped to be the same size.


EDIT:
Here's the video from the mkII (watch in HD)
http://www.youtube.com/watch?v=HkmjTFarOLQ

Thanks. I do think that it is important to see the Raw image. The red will inevitably look better as it is uncompressed verses the Canon solution. That said, to truly just the noise at such a high ISO its tough. My 7d is GRAINY at 3200 and has some banding and compression artifacting as well, but when scaled down slightly these abnormalities disappear. This is shown in to some extent in your scaled 5d footage as well.

The red chart looks really, really good...just curious to see how good. Would love to see the raw up on this forum.

 

[Gary Blake]

Paul is on point.
7D scaled to fit would be noticeably noisy just like the 5DII at 12,800 would be noticeably noisy.

But releasing a 1200x512 image isn't the best way to show noise control.

 

[Graeme Nattress]

Which is secret, so if anyone is reading this, please ignore any reference. Wave hands. This is not the app you're looking for....

Graeme

 

[Darren Orange]

Decepticons, the RAND shard has been located...

 

[Roberto Lequeux]

7d iso4000 - noise

7d iso4000 - noise

This whole thing is just to give a rough idea of what to expect in terms of noise and compare that aspect of the 7D (S35 sensor!) to the latest M-X ISO4000 motion posted in recon.

To give a rough idea of how much light I "used" I added a picture. The focus is the two soda cans on the left. There is also very faint and soft tungsten fill coming from the right, but as you see it is completely overpowered by the monitors. The second image attached shows the settings used to render the .mp4 out of Vegas 9. Notice it is an 8Mbps H.264 and 540p 16:9, exactly half 1080p. It also crops in 50% half way to show a 1:1 crop of the 1080p*. Unfortunately it is shaky, but it doesn't really matter cause even with heaps more light than the video posted in recon the 7D falls apart at ISO4000... crap is still crap when stable:

7D ISO4000 (31.5 MB)

Hehe, then look at this one right away. I'm pretty sure this is 1/48th and f/2.8:

ISO 4000 moving..



http://redgrabs.com/flashlight.mov

 

[Gary Blake]

Yikes.
The 7D is really scary at ISO4000.

Thanks for taking the time to do that.
The 5D II is so much nicer.

http://www.youtube.com/watch?v=HkmjTFarOLQ

 

[Brook Willard]

4,000 ASA Test

4,000 ASA Test

Here is the ISO 4000 file in 2.2K (Jarred will post it). To recap:

This is a RED ONE with the new Mysterium-X sensor.

I turned out all the lights in the garage so it was pitch black... really. Then I lit the chart with my iPhone and free app called "Just Light"... not very bright. You can download it yourself and do this same test. Don't cheat... you need a pitch black room.

We used Graeme's new test app (RAND) with our new color space (REDcolor), our new gamma (REDgamma) and a touch of Graeme's new secret NR sauce (very mild)... like milk chocolate.

Next week I'll do the same test at ISO 800 to show you what the new sensor can REALLY do.

Jim

2.2k MOV:

http://red.cachefly.net/2.2k.mov

Here's the Canon 7D @ 4,000 ASA, 4100˚, 1/50th shutter, 23.98, T3.5 [The IS 18-55 kit lens is the only thing I have here right now] using the iPhone app Just Light. I don't have any charts here, so I chose a RED shirt since it's something I'm sure a lot of people here own.

This is the original file minus audio.

Download here.


Now granted we're comparing a $22,500 camera to a $1,300 camera, but it's a lot more telling than I was expecting. This makes me more excited than anything... because the 7D's low light characteristics blow away the RED ONE's in many situations.

 

[Roberto Lequeux]

Yikes.
The 7D is really scary at ISO4000.

Thanks for taking the time to do that.
The 5D II is so much nicer.

http://www.youtube.com/watch?v=HkmjTFarOLQ

But the 5D is a FF35 sensor, so it isn't a direct comparison of the sensor and recording technology. AND both the 5D and the 7D already have noise reduction built in. AND both the 5D and the 7D give you juuust a little more than 720p worth of actual resolution!

And I dunno about you, but it looks like with great noise reduction and some light in the shadows we might be seeing "decent" 2k out of Red at ISO4000!!??!!!??!!??!!

Then again I should also note that the iPhone while smaller it is much closer to the areas of the chart we are focusing on. Any DPs here that can estimate what my two 500cd/m2 24" LCDs are hitting the cans with?
Then again it doesn't matter because the severely underexposed areas of the M-X video look clean, and the noise in the 7D creeps all the way up and is ruining all the mids. Right?