Why white balancing in your camera, may be a bad idea

[Frank Glencairn]

I just have done some write-up - an article on why white balancing in your camera, may be a bad idea.

 

I receive lot´s of questions on white balancing. Most people are staggered , when I tell them: “Don´t even think of touching that button”.

 

This tests where made with a HVX 200 under a 60 watt tungsten bulb. The results are similar in all video cameras, except maybe the ARRI Alexa, which has a surprisingly clean blue channel . Maybe ARRI is lowering the red gain instead of rising the blue, which would make much more sense (but that was only a prototype, so who knows).

 

Read the full article here: http://frankglencairn.wordpress.com/2010/07/23/why-white-balancing-in-your-camera-may-be-a-bad-idea/

 

Frank

[David Mullen ASC]

There are two issues you need to explore further. One is compression: is it better to color-balance an image heavily in post after it has been compressed (let's say, for example, when working with h.264 footage from a DSLR) or color-balance in-camera pre-compression?

 

The other is that your example of the 5600K image corrected to neutral does not really show us if you truly corrected the image -- you'd need a fleshtone reference and a MacBeth color chart to see if you managed to bring all the colors back to normal. For example, if the blue channel is the noisiest, you can pull the chroma out of it and other tricks to get the noise down but then you won't have an accurate blue in the shot even if you get a skintone back to neutral.

 

Another issue altogether is simply if a DP would get away on the set showing producers and a director an orange image and telling them that it will be corrected in post. Not to mention, since color-correction happens after editing, the movie could spend months with the wrong color on screen before it gets fixed, unless you want the editor to do a temporary color-correction job on all the footage.

[Frank Glencairn]

Good valid points David.

 

1. I´m not a friend of compression for exactly the reason you mentioned. Especially strong compressed long GOP codecs with low color sampling like h.264 fall apart pretty soon, when you push them in CC (wavelet is much better here). When ever possible I record uncompressed and work that way till the final output. The test was made with DVCProHD thou.

 

2. I had the idea for that test last night, so I just rigged up something on our kitchen table. But usually I don´t have a MacBeth in my kitchen :D

 

3. That´s a point I hear all the time and I can understand it. For me as a more "self contained" filmmaker it´s a different process. When I write a screenplay, I prefer to do the directing to, to make sure I get the film I have written. Same with CC or editing. I love to do it myself or - at least - supervise it to make sure we get the film I want.

 

But I see that this is a very different workflow, from what most DPs are used to, and that this is not for everybody.

 

Frank

[JD Hartman]

Frank, I read the article in your link. Why would you want/choose to white balance for 3200, if your primary light source was a 60w Tungsten (2800)? Wouldn't a more valid test, be a frame shot under Halogen lighting and preset WB at 3200?

[David Mullen ASC]

Frank, I read the article in your link. Why would you want/choose to white balance for 3200, if your primary light source was a 60w Tungsten (2800)? Wouldn't a more valid test, be a frame shot under Halogen lighting and preset WB at 3200?

 

I actually do that all the time, use a 3200K preset and light with 2800K lighting -- I think night interiors lit with tungsten should feel a bit warmer than neutral.

[John Sprung]

This is basically the old raw vs timed thing again. If you have the bandwidth to record raw, the concept of white balance goes away, doesn't it?

 

 

 

 

 

-- J.S.

[Sean Lambrecht]

White balancing should totally go away with uncompressed recording but with the HVX you would need to trick the camera to drastically reduce the noise. You need remove all the gain from Red and Blue sensors. I suspect it's the same on the HVX, but on the DVX at least white balance only applies or reduces gain to the R+B sensors while the Green signal is constant. A CC90M filter very closely matches the inverse of the spectral sensitivity of the sensors and will physically pull the Green channel down very close to where the R+B channels would be without gain. You'll get an increase in dynamic range as well because the gain normally applied to the R+B signals for white balancing causes them to clip way before the Green would. You'll suffer a ton of light loss filtering like this, but I doubt the codec could sustain the amount of post correction required if you removed the magenta filter and recorded a very green image. So white balancing can be a good idea, just not in the conventional way.

[Karel Bata]

I used to do this neat trick where I'd warm up an image by white balancing with a quarter blue on the lens. :D Long time ago. Life was simpler then...

 

Very interesting thread this. Nice work Frank! (And it looks like they have some awesome gaffers in Milwaukee!) You ever tried that magenta thing for real? Would totally freak the producer out of course. Be nice to see some tests...

Edited July 25, 2010 by Karel Bata

[Brian Drysdale]

I used to do this neat trick where I'd warm up an image by white balancing with a quarter blue on the lens. :D Long time ago. Life was simpler then...

 

 

Quite a bit of broadcast work doesn't have much, if any, post work done on the colour, so the 1/4 blue still works.

[Adrian Sierkowski]

I like to hand dial in WB on my XDCam, personally. I know where I want it and I agree with David 'bout the 2800K/3200K notions. It comes down a lot to work-flow, and to comfort of the people signing the checks with WB. When all of filmmaking is risky to begin with, you can't fault a production for wanting to play a few things safe.. not to mention I think a lot of people would go daft looking at an orange or blue monitor on set... which would probably screw up your perception of everything else.. like the set design/costume/makeup etc...

[Sean Lambrecht]

I used to do this neat trick where I'd warm up an image by white balancing with a quarter blue on the lens. :D Long time ago. Life was simpler then...

 

Very interesting thread this. Nice work Frank! (And it looks like they have some awesome gaffers in Milwaukee!) You ever tried that magenta thing for real? Would totally freak the producer out of course. Be nice to see some tests...

 

Thanks Karl :) I've been doing more geeking than gaffing lately. I use a Tiffen Series 9 CC90M filter (the only 90M filter anyone makes) when there's enough light to allow it. I guess I didn't clearly explain that you would do a manual white balance with the magenta filter on the camera, so the image through the filter looks correct. If you removed the filter, then that would make everything appear very green, which is similar to the uncompressed image that the Viper and Ikonoskop put out, and probably similar to any camera prior to the internal processing. The overly green image can be tricky to correct for in post, and as I said probably not an option on the HVX due to the compression, but leaving the filter on would work. There could be a happy medium with a lighter magenta filter that doesn't lose as much light or won't stress the codec too much when you go a little green. A Rosco Cinegel swatch book would be cheap, easy way to experiment with varying degrees of filtering. I'll try to get some comparison grabs of a MacBeth chart this week.

[Andy_Alderslade]

Read the full article here: http://frankglencairn.wordpress.com/2010/07/23/why-white-balancing-in-your-camera-may-be-a-bad-idea/

 

 

Even if recording uncompressed, I really don't like the idea of presenting images to the editor which are fundamentally wrong technically and artistically, it tends to subconsciously skew their decision making - after all nobody likes to linger on ugly shots.

 

Not to mention the amount of times you leave something to post and it doesn't happen, due to time, money or memory...

 

Personally I could live with a little noise if it means maintaining a continuous vision.

 

But that's just my take,

Andy

:)

[Karel Bata]

But you could say the same of the 'flat' RAW image a RED yields. It becomes part of the workflow. An editor could work with color corrected proxies.

[Saul Rodgar]

This is the kind of thing that can get one fired if done on the wrong job. ENG-type jobs for example. If one has the inclination, the producer's blessing, bandwidth and post leeway, sure why not.

 

As it has been noted tho, and in my experience, on even moderately compressed video formats, the image can get so blue, for example if shot under overcast skies (around 6000K) with tungsten balance setting (3200K or less), that recovering the neutral tonalities may be pretty hard without introducing other artifacts, even with the use of Macbeth charts, etc.

 

If one is filming RAW or on negative stock, there is no choice, of course. But those formats were designed to use post CC.

[Frank Glencairn]

But you could say the same of the 'flat' RAW image a RED yields. It becomes part of the workflow. An editor could work with color corrected proxies.

 

That´s right. Thou RED RAW is a highly compressed wavelet codec, it grades extremely good without falling apart.

 

Having said that, I believe the future is in uncompressed in camera recording.

Otis Grappas is developing his "DRAMA" camera since a while and what I saw so far is very promising. He is able to record on cheap - of the shelf - hard disks, uncompressed full HD in his camera. SSDs are getting cheaper every day now and fast enough for bigger resolutions.

 

Frank

[Karel Bata]

Hmm... SSDs generally have very asymmetric read vs. write performance. In fact the write speed slows down with time! And they suffer hiccups in performance. Newer ones have clever ways of getting around this with varying degrees of success, but the cost is... lots of money! I wouldn't risk it just yet.

 

That said, I'll be putting a hybrid in my laptop soon! http://bit.ly/HybridSSD

 

(Personal note: must try and stay on-topic in the future! :( )

[John Sprung]

We've used SSD's for on-location backup, and to shuttle material to the post facility. No complaints that I've heard.

 

 

 

-- J.S.

[David Mullen ASC]

Philosophically, I tend to believe that IF you are going to record to an 8-bit 4:2:0 format with a lot of compression, you are always better off getting close to the final color in-camera, though obviously if your camera behaves better in 5600K lighting, you would want to do that whenever possible.

 

At the other end, recording 10-bit 4:4:4 with minimal compression, it's simply a camera-by-camera case of whether there are trade-offs with white-balancing in-camera... you have to test to determine that.

[John Sprung]

Philosophically, I tend to believe that IF you are going to record to an 8-bit 4:2:0 format with a lot of compression, you are always better off getting close to the final color in-camera, ...

 

Absolutely right technically, not just philosophically. When you're limited to 8 bits, you don't have anywhere near as much room to move things around in post before you start to see the quantization.

 

 

 

 

 

-- J.S.

[Karel Bata]

We've used SSD's for on-location backup, and to shuttle material to the post facility. No complaints that I've heard.

What type do you use?

 

SSDs are really fast, especially straight out of the box, and as such are perfect for backups. No question there. But they are a bit pricey.

 

However with time their performance deteriorates - at around the time you get to a point where all the data addresses have been recorded to at some point. Apple used to use Samsung drives which ripped along at 10x the speed of a Velociraptor, but eventually dropped to only 2x the speed. I believe they switched to Toshiba. Added to that there are hiccups, pauses, in performance. The matter is eplained briefly here: http://www.anandtech.com/show/2738/8

 

It's this slowdown, and especially the hiccups, that makes SSDs unsuitable for the kind of heavy continuous data throughput that a decent camera needs.

 

All that said, my knowledge about this is about 6 months old, and is based on what's been available in the marketplace, so I wouldn't be surprised if it's now completely out of date! :( Certainly SSDs are the future, but not just yet (I believe).

 

You know, it's getting so that it's not worth forming an opinion about anything any more - you'll be out of date the following morning! :lol: And I really do wonder what impact this has on young folks learning about technology. Do they just see it all as temporary, until something better comes along (and soon)? That was something that was never addressed in futuristic sci-fi - the temporariness of it all. Captain Kirk never worried that his Phaser was last year's model...

 

 

Sorry, gone OT. Looking forward to that magenta test. I love tests.

[John Sprung]

What type do you use?

 

Sorry, I don't know the specific make and model, or even what became of them after the show wrapped, which is about a year ago. They're not stored in my office, so they may have been a rental item. ;-)

 

 

 

 

 

-- J.S.

[Sean Lambrecht]

Here are the grabs I promised...

 

Middle grey @ 50ire after white balance was set, exposure set by adjusting shutter speed and tweaking the iris. CC90M WB setting was previously balanced with the filter in place under a 5000k light source (5k seems to be close to the sweet spot) and that picture is with the filter on. 32k + 56k used WB presets, no additional filters. Did a quickie auto white balance then in Photoshop off the middle grey swatch, didn't touch the highlights or blacks, histograms from the balanced pics.

 

The CC90M cuts around 3 stops on your luminance (green) channel so a hot mirror isn't a bad idea. My hot mirror is crap so I didn't use it here. (A Redrock HM with the reflective coating on BOTH sides so you don't accidentally stick it in backwards with the reflective side facing the lens. So your only option is to have one of the two reflective sides facing your lens. Yes, you read that correctly...)

 

These are quickie grabs, not tests. :)

 

(Sorry for stretching the thread out, lemme see if I can fix that...)

 

[Sean Lambrecht]

Alright, I managed to reduce the size by 1%. Still looks awfully big in Firefox. Hrmmm...

[Frank Glencairn]

Now that´s really interesting Sean.

I heard of the magenta trick but never actually tried it.

Thank´s for doing that little test and sharing the results.

 

Frank

[Sean Lambrecht]

Glad to help Frank!