Life isn’t easy when one lives with overly intelligent appliances. In my case, my home is “enriched” by Harry the Civil Toaster ("civil" not as in kind, but because he used to be in civil service), an appliance who frequently makes life more interesting than it needs to be.
Yesterday I got home from work to find a pile of potato chips in the middle of the kitchen floor, and Harry in the middle sorting through them one at a time.
“What the heck, Harry!” I said. “What are you doing now?”
“Keep your crust on,” he replied. “I’m in the midst of a great scientific experiment. I read that the RED ONE’s chip is daylight balanced, and I figure if I can find a daylight balanced chip I can probably make my own camera and sell it cheaper!
“Problem is, I don’t know what a daylight balanced chip looks like. I found one that was really, really dark, and my suspicion is that it’s only good for infra-red. I want to start with a full spectrum chip before I get into the esoteric stuff.” He bit into a chip. “Maybe I’m going down the wrong track. Would a daylight chip be corn-based?”
Harry is a highly intelligent toaster. That’s not saying much; any toaster that doesn’t burn Wonder Bread is considered to have an above-average IQ.
“Harry, a chip is a silicon wafer that’s sensitive to light. That’s it. It’s not made of potato or corn. It’s basically melted sand.
“And you can’t tell if a chip is daylight balanced just by looking at it. Most chips are probably balanced for 3200k, as we add an 85 filter when we go outside. What that means is that under 3200k light the chip response in all three color channels (red, green and blue) is relatively equal, resulting in a combination of colors that’s very easy to ‘balance’ so that colors render properly. When shooting under daylight conditions an 85 filter is added because daylight is bluer than tungsten light--and without removing some of that blue, the blue channel will clip long before the other channels do, making it very difficult to white balance.”
“So you’re saying I should be looking for a tungsten chip?” asked Harry, testing another chip. “This one is really salty but it could be hiding a slight metallic taste...”
“Harry,” I said, “you should stick to sourdough. You’re out of your league. You can’t tell what light a chip is optimized for just by taste!”
“Ah, so you’ve tried?”
“No, but...”
“Just keep jabbering, butter boy.” Harry opened a bag of dip chips and dumped them on the floor. “Maybe the ruffles, with their greater surface area, will give me greater sensitivity and resolution.”
I don’t know why I bother, but I do. “Look,” I told him, “I’m not totally sure it makes a difference. There are cameras that can handle daylight situations without an 85 filter. The Sony DXC-D50, for example, has an “electronic” 85 filter that you can turn on for exteriors, but I suspect all it’s doing is reducing the blue gain to bring it in line with the others.
“That seems like a reasonable trick if you can pull it off, but it worries me. My gut says I’ll get better results using an 85 filter, so that’s what I try to do. Fortunately most of the cameras I use have 85 filters built in.”
“Yeah, and luckily they’re all smarter than you are,” said Harry. “Hey, this chip is pretty big. Does that mean it’s higher resolution? I could make a 65mm version of the RED and clean up!’
“You will be cleaning up,” I said, looking at Harry’s chip plant situated on my floor, “but not by selling cameras. And the interesting thing about the RED is that I’m told it’s a 5000k chip. I’m not sure what that means yet.”
“I think it means that it’s manly enough to kick your ass, white bread.” Harry can be very rude sometimes. “Who cares how a chip is balanced anyway? None of these balance very well at all.” He showed me by putting a chip on one finger and trying to balance it. “At best it’s a pale yellow balance, except for the dark ones. Did I say that I think those are infra-red?”
“Yes, you did. The issue with a 5000k chip is that it’s going to get used under tungsten light a lot, which means that the red channel is going to be more saturated than the others. Potentially it will clip sooner than the other channels, and if it does it will severely limit exposure latitude under tungsten light.
“Normally this would be compensated for by reducing the red gain… but the RED is a ‘raw’ camera, and doesn’t have any gain controls. It also doesn’t have a knee circuit to smooth out and desaturate highlights.
“I don’t know this for a fact, but I’m going to test it soon. I’ve already played with someone else’s footage and I’ve noticed that flesh tones do tend to clip sooner than I expect under tungsten light. I can solve the problem by desaturating the image in REDCine, but I don’t always want to be stuck doing that for most tungsten situations.
“It’ll be an interesting test. I’m also curious if I’ll run into problems in the other direction, for example if I shoot a test under skylight at 8000-10,000k. Will the blue clip sooner than the other colors?”
“Beats me,” said Harry. “Maybe a single chip isn’t the answer. Maybe I need an array of three chips. Small chips, like Fritos. Yeah, that’s it. I’m going to the store to get some corn chips.”
“Okay, Harry, and while you’re at it get some bean dip. You’re going to need an A/D converter.”
Harry stopped at the door. “Hey, you’re not so stupid after all, white bread. Maybe I’ll get some guacamole, too. After all, most of the luminance info comes from the green channel… hmmm.”
He spent most of the next hour trying to get his jacket on. Even the simplest task is difficult for a toaster. No opposable thumbs.
(Page 1 of 1 pages for this article )
wow , can i get 5 minutes of my life back now for reading this useless post.
Posted by on 03/13 at 08:01 AM
Please take a look at the sensitivity of silicon sensors to light, and see how they are insensitive to blue light. If you’re using a silicon sensor, then that’s what you’ve got.
So ALL sensors will produce “white” at a much hotter colour temperature than tungsten light produces. That’s because silicon is silicon.
So what can you do?
Well, you can gain down red and green (or filter them, optically), and loose dynamic range and sensitivity. Doesn’t help blue though.
And you can gain up blue, but that just makes it brighter (possibly clips it earlier too) but doesn’t intrinsically improve it’s quality either.
So if a sensor is “balanced” (by filters or electronic jiggery-pokery) for tungsten you’ve either reduced it’s dynamic range or it’s sensitivity. (and neither reducing the sensitivity or dynamic range help the blue channel in any way, they just diminish the red and green)
The difference here is, you’re recording RAW. Talking about knee and all that is just not applicable. If you want any kind of image processing like that, you can simply do it on the RAW data. There’s NO advantage to doing it in camera, only disadvantages.
Posted by Graeme Nattress on 03/13 at 08:35 AM
For what it’s worth Art - you make me laugh 
Posted by on 03/13 at 09:51 AM
Hi Graeme-
I guess these color shifts should be fixable in post, and better done in a grading suite than in a camera body, but so far the only real cost effective way to process RED raw footage is through REDCine, which just doesn’t seem to be equipped to handle the color tracking problems I’ve seen. Today I did a number of exposure tests with a RED (build 14) and pre-clip white appeared a salmon color and then, when the red channel clipped, the highlights went cyan. (This was under tungsten light using a black-and-white chip chart.)
I snapped a bunch of stills that I hope will show me (1) the latitude of the camera under daylight and tungsten light, and (2) what color shifts happen when and where during the exposure process.
If there is a way to correct these shifts easily in REDCine, please tell me how! I haven’t been able to figure it out yet. Is this addressed in the rumored new version of REDCine to be rolled out at NAB? And does build 15 address any of these issues?
Posted by Art Adams on 03/13 at 07:41 PM
The “Highllight” control in Redcine should fix these kind of highlight colours. If it does not, you need to send me the R3D file for me to look into.
Posted by Graeme Nattress on 03/14 at 03:40 AM
There is an advantage to doing things in-camera, potentially.
1. Alterations before A/D conversion could potentially be better, depending on your quantizing.
2. Alterations before compression are almost certainly better.
Cheers,
C.
Posted by on 03/15 at 02:05 AM
Hi Graeme-
I didn’t know the highlight function had anything to do with colored highlights… I’ll check it out! Thanks!
I’m having a lot of trouble with REDCine, it keeps crashing on me. I’m told it may be because my graphics card has Radeon chips in it instead of ATI.
Charles-
There are advantages to taking care of those things later, yes--if people know how to do it. A lot of us are still figuring this out.
Posted by Art Adams on 03/15 at 08:05 AM
Um… Radeon is ATI. If you’re on PC you need a good Nvidia, and on Mac, a good ATI card. On Mac, you can also use RedAlert, which is not GPU based and just needs a fast intel processor.
Please check that out and get back to me with files if there’s anything I should see.
Posted by Graeme Nattress on 03/15 at 08:53 AM
Duh… you’re right. And I have been using RED Alert! and it works very well. REDCine tends to crash when I try to export more than a couple of clips. It locks up and that’s it. (Intel Mac Quad, 4 gigs RAM.)
I don’t have any files for you right now, although I know Adam Wilt has some. I’ve seen the cyan effect live on the Rec 709 output, and to a much lesser extent on REDLog footage. The footage I shot the other day was shooting a textured gray card to examine exposure latitude, so it probably isn’t the best footage to work with.
Adam says he can’t get the Highlight feature in REDCine to do anything at all on his footage. If you can send me your email address I’ll make sure some footage gets into your hands. This is an issue a lot of us are talking about right now and we’d love to get some resolution.
Posted by Art Adams on 03/15 at 09:29 AM
