Jump to content

Recent testing of the new M-X sensor


David Mullen ASC

Recommended Posts

  • Premium Member

I'm testing a Red One with the new M-X sensor right now in Chicago, for a TV pilot I'll be shooting in March.

 

My initial impression is that the camera is vastly improved over what I worked with in the summer of 2008 on "Manure" and "Stay Cool".

 

I tested the camera in 3200K light (a SMPTE 11-step chart, and then a face with a MacBeth chart), at different ASA ratings from 250 to 3200 ASA.

 

The noise is extremely low all the way to 1000 ASA, and even above that, it's fairly mild. But what's important to note is how the noise LOOKS -- in 3200K, it no longer has that chunky blue noise look of the old Red sensor in earlier builds, it's a tight, smooth noise that is more like mild film grain -- the noise at 1000 ASA looks like the grain of a 200 ASA film stock.

 

The roll-off into the clipped highlights is smoother now, less artificial looking, and overall there seems to be wider dynamic range.

 

RedColor works better as a viewing format and RedGamma works great as an output format, much better than RedLog. The image has a nice film-like smoothness to it.

 

But what pleased me the most was the performance in low-levels of tungsten light, which was an important step.

 

2000 ASA was certainly usable, even 3200 ASA wasn't bad -- the jumps in noisiness as you rated the camera faster were less extreme than when boosting the gain on a video camera. Practically speaking, the new sensor makes the camera at least a stop faster than the F35 / Genesis due to the low noise floor, even if the actual sensitivity of the sensor isn't necessarily faster. In other words, 800 ASA on the M-X sensor reminds me of 400 ASA on a Genesis camera.

Link to comment
Share on other sites

  • Replies 76
  • Created
  • Last Reply

Top Posters In This Topic

The noise is extremely low all the way to 1000 ASA, and even above that, it's fairly mild. But what's important to note is how the noise LOOKS -- in 3200K, it no longer has that chunky blue noise look of the old Red sensor in earlier builds, it's a tight, smooth noise that is more like mild film grain -- the noise at 1000 ASA looks like the grain of a 200 ASA film stock.

 

Impressive to hear, especially from a reliable source such as yourself.

 

2 questions:

 

what do you think you'll find yourself rating the camera at?

 

and do you think the new sensor essentially eliminates the need to overexpose by almost a stop to get a clean image?

 

I found myself rating the old sensor at about 200 asa when shooting at 320, ie - that's the lightmeter setting that seemed to deliver the best images in post in most situations. I know the red is it's own beast, but I do find the ASA setting useful for prelighting and communicating with my gaffer.

 

Thanks!

Link to comment
Share on other sites

  • Premium Member

Let's get a cheap shot out of the way first - it wouldn't have been terribly hard to outdo the original in terms of noise performance or dynamic range.

 

From what little I've seen, though, the more recent revision is indeed a vast improvement. No chance of any stills, I suppose - or better yet one of the R3Ds?

 

P

Link to comment
Share on other sites

My initial impression is that the camera is vastly improved over what I worked with in the summer of 2008 on "Manure" and "Stay Cool".

So where does that leave all the prople who invested in a RED One?

5-6,000 people can't be wrong?

 

But this is what always happens.

Every generation of cameras is gnat's whisker short of being indistinguishable from reality

Then the next one is always Vastly better.

Great if your the rentee, Sh*thouse if your'e the rentor.

Oh well, 'twas ever thus.

Link to comment
Share on other sites

So where does that leave all the prople who invested in a RED One?

5-6,000 people can't be wrong?

 

But this is what always happens.

Every generation of cameras is gnat's whisker short of being indistinguishable from reality

Then the next one is always Vastly better.

Great if your the rentee, Sh*thouse if your'e the rentor.

Oh well, 'twas ever thus.

 

 

You can upgrade the sensor.

 

So yeah, it leaves them in a pretty ok position.

Link to comment
Share on other sites

  • Premium Member
Great if your the rentee, Sh*thouse if your'e the rentor.

Oh well, 'twas ever thus.

 

Yup, that's a major part of the value that the rental houses bring to the table. They eat the depreciation and take the risk on obsolescence. They also keep it working, hop right out with a new one if we bust something, and store it safely when we don't need it. They know how it really works, and tell us what we need to know. That's why we use them rather than buying stuff. Equipment rental is a difficult and honorable business.

 

 

 

 

-- J.S.

Link to comment
Share on other sites

  • Premium Member

Thanks for the report David, this is pretty exciting stuff. I have several questions:

 

1. Were you monitoring in RedColor/RedGamma out of the camera as well?

 

2. What are your impressions on the improvements in contrast/color accuracy/dynamic range for monitoring on set and making exposure decisions off the HD-SDI monitor? Of course, we can still use RAW View metering for exposure, but I'm curious if the new color science is improved to the point that we will have WYSIWYG on set.

 

3. Do you find that the ISO on the camera now matches your light meter more accurately than the Red One?

 

Hi John and Luke,

 

Isn't the sensor upgrade around $5k? That's nothing for a rental house, and a significant chunk of change (though still quite reasonable) for the private owner... I'm sure most Red owners are pretty happy with this deal. Sony and Panasonic aren't gonna give you what is essentially a new camera for $5k!

Link to comment
Share on other sites

I'm testing a Red One with the new M-X sensor right now in Chicago, for a TV pilot I'll be shooting in March.

 

My initial impression is that the camera is vastly improved over what I worked with in the summer of 2008 on "Manure" and "Stay Cool".

 

I tested the camera in 3200K light (a SMPTE 11-step chart, and then a face with a MacBeth chart), at different ASA ratings from 250 to 3200 ASA.

 

The noise is extremely low all the way to 1000 ASA, and even above that, it's fairly mild. But what's important to note is how the noise LOOKS -- in 3200K, it no longer has that chunky blue noise look of the old Red sensor in earlier builds, it's a tight, smooth noise that is more like mild film grain -- the noise at 1000 ASA looks like the grain of a 200 ASA film stock.

 

The roll-off into the clipped highlights is smoother now, less artificial looking, and overall there seems to be wider dynamic range.

 

RedColor works better as a viewing format and RedGamma works great as an output format, much better than RedLog. The image has a nice film-like smoothness to it.

 

But what pleased me the most was the performance in low-levels of tungsten light, which was an important step.

 

2000 ASA was certainly usable, even 3200 ASA wasn't bad -- the jumps in noisiness as you rated the camera faster were less extreme than when boosting the gain on a video camera. Practically speaking, the new sensor makes the camera at least a stop faster than the F35 / Genesis due to the low noise floor, even if the actual sensitivity of the sensor isn't necessarily faster. In other words, 800 ASA on the M-X sensor reminds me of 400 ASA on a Genesis camera.

 

David... it sounds like your experiences matches that of others that have shot the new sensor. Do you have any negatives to report? We are anxious to improve.

 

Jim

Link to comment
Share on other sites

  • Premium Member
David... it sounds like your experiences matches that of others that have shot the new sensor. Do you have any negatives to report? We are anxious to improve.

 

Jim

 

I still don't care much for the EVF -- we were doing a slow pan at night and I asked the AC to do a stop-pull, and with all the bulky coats we are wearing, his sleeve bumped those three buttons on the EVF and switched the camera into RAW view, which since I was shooting at 1600 ASA, made everything go dark so I cut to find out what had happened. On my last RED shoots, we just disabled those buttons but they really are in the wrong place, it's too easy to grab the EVF there. And in extreme lighting situations, the EVF image starts acting weird, banding problems on snow, almost solarization around a bright sky, etc.

 

I know that's a minor thing unconnected to the image.

 

At night shooting into car headlights, on primes with no filters, there is a ghost image of the headlights near the top right of frame, sort of pinkish. I assume there is some sort of internal reflection going on between the IR/OPLF in front of the sensor and the back of the lens maybe.

 

Those were the only two annoyances, other than we discovered a thumbprint on the sensor once we got into night shooting. I suppose I could fingerprint everyone and track it down... either that or every Red goes out with Jim Jannard's personal thumbprint of approval on it! I mark this down as due to the rush to get the camera from NYC into Chicago and testing it immediately, causing a lack of basic prep on everyone's part.

 

IR problem seems minimized now and I didn't catch much of that sensor protection effect around points of light.

 

I'm still waiting to see all of the footage on a big HD monitor at a post house.

Link to comment
Share on other sites

At night shooting into car headlights, on primes with no filters, there is a ghost image of the headlights near the top right of frame, sort of pinkish. I assume there is some sort of internal reflection going on between the IR/OPLF in front of the sensor and the back of the lens maybe.

 

Something like this David? http://reduser.net/forum/showpost.php?p=54...mp;postcount=15

 

Is this a serious issue to be concerned about?

Link to comment
Share on other sites

  • Premium Member
Something like this David? http://reduser.net/forum/showpost.php?p=54...mp;postcount=15

 

Is this a serious issue to be concerned about?

 

I can't access that page for some reason, but it's not a color cast problem, it's "narciscism" I think the term is, an optical issue.

Link to comment
Share on other sites

I can't access that page for some reason, but it's not a color cast problem, it's "narciscism" I think the term is, an optical issue.

 

Here is the image from MX test:

 

dyufxi.jpg

 

There was some speculation that it was light reflecting off the sensor. I am not saying this is the case. I don't know.

Link to comment
Share on other sites

Those were the only two annoyances, other than we discovered a thumbprint on the sensor once we got into night shooting. I suppose I could fingerprint everyone and track it down... either that or every Red goes out with Jim Jannard's personal thumbprint of approval on it!

 

Well, there'd be precedent for a thumbprint of approval from the past. Every English musket used to get a "seal" of approval after it was proved a straight shot.

 

 

What would a thumbprint on the sensor look like, almost like a thumbprint on the negative from a loading error on a 35?

 

 

 

I don't think the Union would be too appreciative of thumbprinting the whole crew ;)

 

 

 

 

One more quick question for you, David: When you say the ISO on this camera at 1000 looked like 35mm film at ASA 200, is that in reference to SD, HD TV, the big screen, IMAX?

 

I've noticed that digital cameras produced less "noise" on even SD TV early on, but I tend to notice other, less desirable effects of theres on a film print or even DLP.

 

I don't think there's a reason to use 500T film pushed at all anymore, unless you want extreme chunky grain and high contrast.

Edited by Karl Borowski
Link to comment
Share on other sites

I still don't care much for the EVF -- we were doing a slow pan at night and I asked the AC to do a stop-pull, and with all the bulky coats we are wearing, his sleeve bumped those three buttons on the EVF and switched the camera into RAW view, which since I was shooting at 1600 ASA, made everything go dark so I cut to find out what had happened. On my last RED shoots, we just disabled those buttons but they really are in the wrong place, it's too easy to grab the EVF there. And in extreme lighting situations, the EVF image starts acting weird, banding problems on snow, almost solarization around a bright sky, etc.

 

I know that's a minor thing unconnected to the image.

 

At night shooting into car headlights, on primes with no filters, there is a ghost image of the headlights near the top right of frame, sort of pinkish. I assume there is some sort of internal reflection going on between the IR/OPLF in front of the sensor and the back of the lens maybe.

 

Those were the only two annoyances, other than we discovered a thumbprint on the sensor once we got into night shooting. I suppose I could fingerprint everyone and track it down... either that or every Red goes out with Jim Jannard's personal thumbprint of approval on it! I mark this down as due to the rush to get the camera from NYC into Chicago and testing it immediately, causing a lack of basic prep on everyone's part.

 

IR problem seems minimized now and I didn't catch much of that sensor protection effect around points of light.

 

I'm still waiting to see all of the footage on a big HD monitor at a post house.

 

Thanks... we had known calibration issues that have been improved in the newest camera firmware. We'll make sure that camera gets it.

The new EVF has many improvements (starting with being 1/3 the size)... hope to begin shipping that fairly soon. Buttons are better and has built in heater.

Fingerprint on the sensor (OLPF actually) isn't an official part of manufacturing... :-)

 

There are still many things to do to finish up the new sensor program... but we hope we are pretty close. I'll have the guys take another look at the posted headlight issue. If we could get the R3D of any problem scene it would be helpful.

 

Jim

Edited by Jim Jannard
Link to comment
Share on other sites

I'm testing a Red One with the new M-X sensor right now in Chicago, for a TV pilot I'll be shooting in March.

 

My initial impression is that the camera is vastly improved over what I worked with in the summer of 2008 on "Manure" and "Stay Cool".

 

This is partly what has always scared me about digital capture, you shot two movies on RED (Stay Cool and Manure) now those movies are "stuck" in the world of the inferior sensor. Meanwhile there is a new and improved sensor for your next project.

 

If you had shot Manure and Stay Cool on Kodak Vision or FUJI Eterna, there would be no difference if you also shot your current project on 35mm FUJI or Kodak stocks.

 

I am considering digital capture for this 3D movie I may be shooting soon, and I know what will happen, a week after I close down a newer and more improved "system" will be released. Never fails.

 

R,

Link to comment
Share on other sites

  • Premium Member
What would a thumbprint on the sensor look like, almost like a thumbprint on the negative from a loading error on a 35?

 

The bad news is that any defect on the sensor surface would be there 100% of the time, absolutely still. A fingerprint on film is only there for one frame -- or two if it crosses frame lines.

 

The good news is that on the Red -- if the OLPF position is the same as the previous sensor -- the outer surface of the OLPF is far enough from the image plane that the thumbprint would be way far out of focus. It would have to be a really big dirty print to hurt the image significantly. Even so, of course, it shouldn't be there. I'd check the Red instructions to see how to clean it.

 

Perhaps David can post a still showing how they discovered it.

 

 

 

 

 

-- J.S.

Link to comment
Share on other sites

  • Premium Member

We had to load some new software in the camera to solve a problem with the blacks going green when the ASA was pushed really high (above 1000 ASA), negating my tests in the morning, so we broke for lunch and Fletcher loaded the new software.

 

When we got back, I was in a hurry to reshoot my SMPTE chip charts and whatnot so I noticed a faint pattern of rippled lines in the color squares in the top right corner of the charts, almost like a surface error in printing the chart - but not visible when you just looked at the chart itself. But having done that S.A.L.T. test of lenses last year, I knew that Ang. zooms were prone to a rippling edge around light sources moving through the frame (looks like fish gills), and since I had an Ang. zoom on the camera and was pointing it at a backlit chart, I just assumed it was some zoom lens artifact.

 

Then outside I was mostly shooting handheld through a dirty car window and when it got to be dusk, the streetlamps were flaring the lens through the windshield. But when I finally got out of the car and switched to Zeiss Ultra Primes, the flare / glow around streetlamps in the top corner of the frame was still there. Now I first thought it was maybe an EVF artifact, since I knew that the Ultra Primes did not halate that much, but that's when we discovered the fingerprint.

 

Later when I saw some frames being transcoded the next day, it was more obvious that the rippled pattern in the top corner of the frame was the fingerprint.

 

Anyway, you can lose a lot of sleep worrying about new technology making your old stuff look worse, so I don't think about it -- past movies reflect the technology of their time, digital or film -- I'm OK with that. Besides, early high-speed stocks were much grainier than they are today so even film technology has improved in the past three decades. Same goes for lenses.

 

"Manure" and "Stay Cool" are an odd case because essentially the producer-investor has taken them away from the director (for now) which has delayed them being finished & distributed, and "Manure" (now renamed "The Smell of Success") was re-timed a few months ago at Lightiron by the producer, without my input or the director's. I didn't get involved because I didn't want to take away any bargaining power that the director had. But Lightiron sent me a few frames of the finished job and one thing I noticed is that they seemed to have a better handle on converting the footage properly, and the new Color Science from Red seems to have helped. I wasn't too worried about not being involved with the timing only in the sense that the entire movie is art directed brown, there is no other color in the frame anyway, it can only be timed one way, color-wise (brightness-wise is another issue.) This new Color Science, plus the fact that Lightiron can work directly with the r3d files gives me hope that "Stay Cool" (which is highly saturated) will turn out well, though I would prefer to be there to supervise.

 

My point being that even though these movies were shot on the earlier Red sensor, they still can benefit from some new software developments. Plus in the case of "Manure" I had plenty of light for the old sensor anyway, and lit most of the movie with HMI's. It's "Stay Cool" that would have benefitted more from the new M-X sensor because I had some available light night scenes pushed to 800 ASA that have some blue noise problems.

Link to comment
Share on other sites

David's kind-of hit on it, but you're only partially correct, Richard.

 

If you shot part of even one movie on a certain stock and one of the film gods obsoleted it, you might even be stuck trying to match two different stocks for the same movie.

 

Or then there is emulsion batch differences, processing differences, exposure differences. No two frames of film are exactly alike.

 

 

Especially with the EXR stocks, high speed ones, they were MUCH grainier than Vision or Vision2/3. I think this may've been due to not being their true speed (sometimes I've heard stories that these films were even a full stop lower in the generation before EXR). Vision was less contrastier than EXR. V2 and V3 much less so. I haven't shot the new V2 500T stock, but even if it is more contrasty, probably still won't look like EXR.

 

 

What if you go back to shoot a movie, a sequel, prequel, want the same look and your stock is gone? Then again, digital cameras age change and die too. Sensors don't last forever.

Link to comment
Share on other sites

  • Premium Member

If you want to see a SMPTE chart shot in 3200K light, here are some frames of my 400 and 3200 ASA test. They are large-ish files though still compressed JPEGS, my website is pretty full of still frames these days so I didn't want to put the TIFF's I have.

 

400 ASA, 3200K

http://www.davidmullenasc.com/400%20ASA%20SMPTE.jpg

 

3200 ASA, 3200K

http://www.davidmullenasc.com/3200%20ASA%20SMPTE.jpg

Link to comment
Share on other sites

  • Premium Member

I notice some (what appears to be) fixed pattern noise in the lower left corner of the greyscale at both speeds; is that a weave effect in the printing of the chart (in which case it's a pretty feeble chart) or is that in the camera?

 

Looks like vertical bright and dark bands a few pixels wide.

 

The reason I suspect it might be the chart is that it isn't visible on the coloured swatches around the edge.

 

P

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Forum Sponsors

BOKEH RENTALS

Film Gears

Metropolis Post

New Pro Video - New and Used Equipment

Visual Products

Gamma Ray Digital Inc

Broadcast Solutions Inc

CineLab

CINELEASE

Cinematography Books and Gear



×
×
  • Create New...