Recent testing of the new M-X sensor

[David Mullen ASC]

There's a stain, I think. I'm looking at the pronounced vertical lines in the greyscale; is that really just a wear and tear issue?

 

P

 

Like I said, the chart has seen better days... I wouldn't try to make judgments about pattern noise from it. It had a slight texture to the printing. What's a good test for pattern noise? A grey scale full-frame shot out of focus?

[Freya Black]

 

I'm not keen on the blue fringing in this image. It's on one of the lampposts (middle one), in the car headlights and on the furthest of the two building lights. None of the other images seem to be exhibiting this effect.

 

I'm curious what is causing that?

 

love

 

Freya

[Freya Black]

Just to clarify some things. I understand that the M-X sensor is an upgrade to the existing Red One. My understanding is that this is a factory upgrade and the camera has to be returned to be upgraded.

 

What I'm wondering is if the upgraded camera comes with new software too? It seemed from your description David like it might?

I'm also wondering if the upgraded cameras require a different set of builds to the older red ones or if the sensor is basically independant of the camera software. Basically if you have a certain build will it run on both versions of the camera?

 

love

 

Freya

[Ramsey Nickell]

David-

 

Thank you for taking the time to post your test results. I'm not sure that everyone fully appreciates your generosity.

 

I certainly do.

 

Ramsey

[Phil Rhodes]

Yeah, I guess a plain grey field shot way out of focus.

 

You do have to be a bit careful because if they're using any optical components on the front of the sensor to maximise effective fill factor - which they probably are - the angle at which the light approaches the sensor matters, which obviously changes when, among other things, you pull focus.

 

P

[David Mullen ASC]

There is new software for the new sensor including a new build, and there is new debayering software for the new sensor, etc. A lot is new.

 

All lenses exhibit some degree of chromatic aberration and streetlamps that are partial spectrum sources can make it worse sometimes, it seems to me at least. Watch something shot on Cooke S4's and notice all the brassy-red fringing that goes on around lights and shiny metal.

 

I just watched "Thief", shot partially in Chicago on Zeiss Super Speeds I believe, and there is a lot of that blue CA around headlamps and whatnot.

 

I try to not get hung up on little optical artifacts like that, particular in a moving image rather than a still frame.

[Michael Collier]

Knowing nothing of the innerworkings of the camera's software, I will venture a guess:

 

I bet there would only be one software "build" that contains libraries and code for both versions. Either during the install, or just as a function of the chip talking to the main processor, the camera will determine which chip it has and which set of instructions to load. But that would mean large sections of code are specific to one chip or the other, so in the future I assume a problem in one camera may not necessarily present in the other version, given the same build.

 

of course this is just a guess.

 

David- Full frame out of focus would work. What I would like to see is a chart that has strips that fill the horizontal, alternating between 50% gray, vertical black and white sharpness lines, and gray scales. It would be nice to see what affect high detail areas have on the amplifier (since tightly alternating black/white lines effectively maximizes the frequency seen by the amp. noise might fall apart under high frequency areas, and it might be smoother where the detail is lower.

 

The big question I have is what part of the image is from chip sensitivity and what comes from software caressing. Was there any 'tracers' or ghosting similar to what software degraining does? Did the image feel solid, like it was a camera original, or did it have a feeling of "mushiness" that DNR gives?

[Jim Jannard]

Knowing nothing of the innerworkings of the camera's software, I will venture a guess:

 

I bet there would only be one software "build" that contains libraries and code for both versions. Either during the install, or just as a function of the chip talking to the main processor, the camera will determine which chip it has and which set of instructions to load. But that would mean large sections of code are specific to one chip or the other, so in the future I assume a problem in one camera may not necessarily present in the other version, given the same build.

 

of course this is just a guess.

 

You are correct... there is a camera firmware Build 30 coming that works with both M and M-X sensors in a RED ONE. REDCINE-X, along with the SDK that has gone out to 3rd parties, supports both M and M-X footage in the same apps.

 

Jim

[Freya Black]

You are correct... there is a camera firmware Build 30 coming that works with both M and M-X sensors in a RED ONE. REDCINE-X, along with the SDK that has gone out to 3rd parties, supports both M and M-X footage in the same apps.

 

Jim

 

Thanks guys. It only just occured to me that the hardware change would probably mean a software change too. The shared build seems like a great way of doing things, like having a basic system but a new driver for the new sensor or something.

 

I'm suprised this didn't even occur to me till David started talking about using it! I feel silly not to have realised this before.

 

love

 

Freya

[Freya Black]

All lenses exhibit some degree of chromatic aberration and streetlamps that are partial spectrum sources can make it worse sometimes, it seems to me at least. Watch something shot on Cooke S4's and notice all the brassy-red fringing that goes on around lights and shiny metal.

 

I definitely think you are right that it could be related to the partial spectrum sources. That frame you have is really intresting however as it seems to be in all the car headlights but only the middle one of the streetlamps suffers from it. The thing is that middle street light looks really nasty to my eyes (much worse than the car headlights). I'm guessing that street lamp is not putting out the same kind of light as the other two. I think it's intresting that there is such a relationship between the lens and the colour temp in this way.

 

I just watched "Thief", shot partially in Chicago on Zeiss Super Speeds I believe, and there is a lot of that blue CA around headlamps and whatnot.

 

Were you also shooting on zeiss lenses?

I'll try and keep an eye out for theif now too.

 

I try to not get hung up on little optical artifacts like that, particular in a moving image rather than a still frame.

 

I suspect you wouldn't notice it in a moving image at all. This is the first time, such a thing has bothered me.

 

Thanks for the reply I think it's intresting.

 

The other shots look great to me BTW. I especially like the train in the station shot. You should definitely try and work that train and station into the pilot. ;)

 

love

 

Freya

[Phil Rhodes]

What surprises me slightly is that it's seeing both the overhead streetlighting and the stuff on the poles as the same mid yellow - was that low-pressure sodium orange in reality?

 

P

[David Mullen ASC]

What surprises me slightly is that it's seeing both the overhead streetlighting and the stuff on the poles as the same mid yellow - was that low-pressure sodium orange in reality?

 

P

 

Chicago went nuts installing sodium-vapor everywhere - the city glows in that color. Apparently some residents have complained about the degree of light pollution. I noticed it when I drove in from the airport, though my first thought was "this is good for shooting in terms of having a lot of available light to work with." As long as the director understands that it would be near impossible to time all of that yellow-gold-green color cast out because these are partial spectrum sources.

[Guest Stephen Murphy]

Chicago went nuts installing sodium-vapor everywhere - the city glows in that color. Apparently some residents have complained about the degree of light pollution. I noticed it when I drove in from the airport, though my first thought was "this is good for shooting in terms of having a lot of available light to work with." As long as the director understands that it would be near impossible to time all of that yellow-gold-green color cast out because these are partial spectrum sources.

 

Looks really interesting . The 800 asa 180 degree stuff looks really good. What colour temp did you have the camera set at for these?

[Tim Carroll]

Chicago went nuts installing sodium-vapor everywhere - the city glows in that color. Apparently some residents have complained about the degree of light pollution. I noticed it when I drove in from the airport, though my first thought was "this is good for shooting in terms of having a lot of available light to work with." As long as the director understands that it would be near impossible to time all of that yellow-gold-green color cast out because these are partial spectrum sources.

 

Having lived in Chicago, on and off, since 1987, I can tell you the "new" sodium-vapor streetlights have drastically changed the feel of the city at night. It went from this romantic warm glow with lots of dramatic shadows; to a blindingly hot-white blast with no secrets and mystery. Really a shame.

 

Best,

-Tim

[David Mullen ASC]

Looks really interesting . The 800 asa 180 degree stuff looks really good. What colour temp did you have the camera set at for these?

 

I set the camera to 3700K.

[Keith Walters]

I'm not keen on the blue fringing in this image. It's on one of the lampposts (middle one), in the car headlights and on the furthest of the two building lights. None of the other images seem to be exhibiting this effect.

 

I'm curious what is causing that?

 

love

 

Freya

 

Hmm. That effect is caused by quantum tunneling in the sensor. I remember first hearing this about from a Philips engineer about 30 years ago, when Bayer-Masked CCD sensors were first appearing in practical cameras! I'm surprised RED don't have algorithms in place to detect and correct this well-known phenomenon.

 

As I may have mentioned one or two (dozen) times here before, the transistor and photodiode structures that are depicted on diagrams of the sensors have little more physical reality than the rope barricades they put up to control queues of people. What is actually in the silicon is a ghostly pattern of electrostatic charges, that within strictly defined limits, can control the flow of electrons surprisingly well.

 

However if you go outside those limits, the whole thing can collapse like a Jello dessert under hot water. If you heavily overload any of the photodiodes, stray electrons start flowing through previously impassable barriers. The analogy is often made to "tunneling" through the potential barrier rather than "climbing over the top", hence the name.

 

Apart from temporarily destroying the function of any nearby transistors, the excess charge can also leak into adjacent photodiodes, producing the phenomenon of "Blooming" where a relatively small overloaded spot can start to grow in size.

 

In the case of Bayer-Mask sensors, the blue fringing is caused by the fact that that silicon photosensors are much less sensitive to blue light than they are to red. When a spot of white light strikes a Bayer sensor in the normal manner, the red photosites will acquire the most charge, green a bit less, and blue, quite a bit less.

This is normally corrected in the signal processing by boosting the green and blue signals so that they match the red.

 

However when a patch of photosites gets its charge by tunneling from adjacent overloaded photosites, sensitivity is not an issue because no photoelectric phenomenon is involved: The red-, the green- and the blue-masked photosites all get the same amount of charge.

 

However, if the de-Bayer software doesn't have the smarts to detect the presence of adjacent overloaded set of pixels, it applies the normal blue-boost to the surrounding ring of contaminated pixels and voila blue fringing. (Another way of looking at it is that, normally, the only time all three sensor colours get the same amount of charge is when they are imaging bluish-white light, so that's what the decoder outputs).

 

By the way, this phenomenon only happens with single-chip sensors; 3-chip cameras are free from this effect.

 

Quantum tunneling is the reason all these clever schemes people keep coming up with for boosting dynamic range by have multiple photosensor sizes never work out in practice: an overloaded highlight area is going to contaminate all the photodiodes around it. There have been various schemes to electrically isolate the photosites better, but inevitably this causes so much sensitivity loss that the improvement is largely negated. You get one or two stops improvement at best, and usually less.

 

If somebody actually figures out a practical way of doing this, they will revolutionize camera manufacturing technology, but it hasn't happened yet.

[Phil Rhodes]

I find the best way to understand this is to visualise lots of little Graeme Nattresses standing by a velvet rope in a tuxedo, consulting a list and saying "If your name's not down, you're not coming in".

 

And then being beaten up by four hundred thousand excited electrons. Aaargh, I'm melting!

 

P

[Satsuki Murashige]

Hi Keith, thanks for the detailed explanation.

 

Do you think part of the reason for the blue fringing is because the light source was sodium vapor with little to no blue in it? The effect of 'quantum tunneling' as you've described it wouldn't be much different than halation on celluloid if the halo were white, no?

Edited February 24, 2010 by Satsuki Murashige

[Keith Walters]

Do you think part of the reason for the blue fringing is because the light source was sodium vapor with little to no blue in it? The effect of 'quantum tunneling' as you've described it wouldn't be much different than halation on celluloid if the halo were white, no?

No, the tunneling electrical charges are effectiveley "colorless". That's the whole point: it doesn't matter what color the light is that causes the pixels to overload, the leaked charges fill up adjacent photosites equally. The presence of a layer of colored gel on top of a particular photosite has no bearing on how much charge leaks into it. Just like it doesn't matter what sort of roof your house has if there is ground water seeping into your basement...

 

Halation on film is mostly due to light scattering sideways in the emulsion and Mylar base. It;s a completely different phenomenon.

[John Sprung]

Do you think part of the reason for the blue fringing is because the light source was sodium vapor with little to no blue in it?

 

As Keith says, that's not the way the physics works. But it may be a little more obvious and objectionable because you know it's not supposed to be there.

 

Inside the image of the light source, everything gets blown out white. Outside, everything stays black as it should. Near the boundary, the yellow light of sodium should pass thru the red and green filters and fill up those photosites, but not the blues, making a yellow/orange fringe, if anything. But the glut of electrons from the nearby blown out part of the image tunnels thru from the reds and greens into the blues. The downstream boost to correct the low blue sensitivity then tilts the output and makes the fringe blue.

 

 

 

 

-- J.S.

[Graeme Nattress]

From my experiments, purple fringing is actually a lens effect and can be eliminated by stopping down the lens. You can easily see it's not a photosite overload effect by shooting into a bright light covered by a slit. First check you're seeing the purple fringing then take two more shots - one stopped down until the purple fringes have gone and one ND'd down by the same amount as you stopped down. When I did exactly the above, the stopped down version showed that indeed the purple fringes had gone, but they were still there on the ND'd version. This shows it's not the light level hitting the sensor causing an overload which is the issue, but that it is happening in the lens.

 

Graeme

[Phil Rhodes]

I think Graeme is right.

 

P

[Graeme Nattress]

I've also seen the effect on film too - there was a couple of shots in GF Newman's Law and Order that exhibited it.

 

Graeme

[David Mullen ASC]

I've also seen the effect on film too - there was a couple of shots in GF Newman's Law and Order that exhibited it.

 

Graeme

 

Like I said, I've seen plenty of film-shot movies with a red edge/fringe around bright lights when shot with Cooke S4's and a blue one when shot with Zeiss lenses. The old Super-Speeds and old Panavision Ultra Primes were really prone to a blue fringe around headlamps and whatnot.

[Keith Walters]

From my experiments, purple fringing is actually a lens effect and can be eliminated by stopping down the lens. You can easily see it's not a photosite overload effect by shooting into a bright light covered by a slit. First check you're seeing the purple fringing then take two more shots - one stopped down until the purple fringes have gone and one ND'd down by the same amount as you stopped down. When I did exactly the above, the stopped down version showed that indeed the purple fringes had gone, but they were still there on the ND'd version. This shows it's not the light level hitting the sensor causing an overload which is the issue, but that it is happening in the lens.

 

Graeme

 

But why does it only happen with those car headlights, and to a smaller extent, one of the street lamps?

 

The QT effect is very real and is most commonly seen with cheap digital stills cameras. Objects silhouetted against a bright sky often have a characteristic blue border on the inside.

 

I'm perfectly willing to accept that there might be multiple phenomena that produce similar-looking artifacts.

It might also simply be that that car has those awful blinding blue-white Xenon headlights that so many people seem convinced they need these days :rolleyes: