Raw Myths. You Can’t Change The white Balance Of the Camera Or ISO in Post.

advertise-here-275 Raw Myths. You Can't Change The white Balance Of the Camera Or ISO in Post.

It’s amazing how often people will tell you how easy it is to change the white balance or adjust the ISO of raw footage in post. But can you, is it really true and is it somehow different to changing the ISO or white balance of Log footage?

Let’s start with ISO. If ISO is sensitivity, or the equivalent of sensitivity how on earth can you change the sensitivity of the camera once you get into post production. The answer is you can’t.

But then we have to consider how ISO works on an electronic camera. You can’t change the sensor in a video camera so in reality you can’t change how sensitive an electronic camera is (I’m ignoring cameras with dual ISO for a moment). All you can do is adjust the gain or amplification applied to the signal from the sensor.  You can add gain in post production too. So, when you adjust the exposure or using the ISO slider for your raw footage in post all you are doing is adjusting how much gain you are adding. But you can do the same with log or any other gamma. 

One thing that makes a difference with raw is that the gain is applied in such a way that what you see looks like an actual sensitivity change no matter what gamma  you are transforming the raw to. This makes it a little easier to make changes to the final brightness in a pleasing way. But you can do exactly the same thing with log footage.  Anything you do in post must be altering the recorded file, it can never actually change what you captured.

Changing the white balance in post: White Balance is no different to ISO, you can’t change in post what the camera captured. All you can do is modify it through the addition or subtraction of gain.

Think about it. A sensor must have a certain response to light and the colours it sees depending on the material it’s made from and the colour filters used. There has to be a natural fixed white balance or a colour temperature that it works best at.

The Silicon that video sensors are made from is almost always more sensitive at the red end of the spectrum than the blue end. So as a result almost all sensors tend to produce the best results with light that has a lot of blue (to make up for the lack of blue sensitivity) and not too much red. So most cameras naturally perform best with daylight and as a result most sensors are considered daylight balanced.

If a camera produces a great image under daylight how can you possibly get a great image under tungsten light without adjusting something? Somehow you need to adjust the gain of the red and blue channels.

Do it in camera and what you record is optimised for your choice of colour temperature at the time of shooting. But you can always undo or change this in post by subtracting or adding to whatever was added in the camera.

If the camera does not move away from its native response then if you want anything other than the native response you will have to do it in post and you will be recording at the cameras native white balance. If you want a different colour temp then you need to add or subtract gain to the R & B channels in post to alter it.

Either way what you record has a nominal white balance and anything you do in post is skewing what you have recorded using gain. There is no such thing as a camera with no native white balance, all cameras will favour one particular colour temperature. So even if a manufacturer claims that the white balance isn’t baked in what they mean is they don’t offer the ability to make any adjustments to the recorded signal. If you want the very best image quality, the best method is to adjust at the time of recording. So, as a result a lot of camera manufacturers will skew the gain of the red and blue channels of the sensor in the camera when shooting raw as this optimises what you are recording. You can then skew it again in post should you want a different balance. 

With either method if you want to change the white balance from what was captured you are altering the gain of the red and blue channels. Raw doesn’t magically not have a white balance, so shooting with the wrong white balance and correcting it in post is not something you want to do. Often you can’t correct badly balanced raw any better than you can correct  incorrectly balanced log.

How far you can adjust or correct raw depends on how it’s been compressed (or not), the bit depth, whether it’s log or linear and how noisy it is. Just like a log recording really, it all depends on the quality of the recording. 

The big benefit raw can have is that the amount of data that needs to be recorded is considerably reduced compared conventional component or RGB video recordings.  As a result it’s often possible to record using a greater bit depth or with much less compression. It is the greater bit depth or reduced compression that really makes a difference. 16 bit data can have up to 65,536 luma gradations, compare that to the 4096 of 12 bit or 1024 of 10 bit and you can see how a 16 bit recording can have so much more information than a 10 bit one. And that makes a difference. But 10 bit log v 10 bit raw, well it depends on the compression, but well compressed 10 bit log will likely outperform 10 bit raw as the all important colour processing will have been done in the camera at a much higher bit depth than 10 bit.

12 thoughts on “Raw Myths. You Can’t Change The white Balance Of the Camera Or ISO in Post.”

  1. RAW format for stills then is different than RAW format for digital cinema…. you are able to manipulate the WB from tungsten to daylight in a RAW image when shooting stills no problem.

    1. Just like different film stocks, the cameras sensor will have just one sensitivity to light and that’s what is recorded or saved. As you say you can manipulate the data with a computer to alter the way the data looks, but it really isn’t different to modifying the data in a jpeg. Raw is a single set of data about the scene just as a jpeg is. The difference is the 8 bit data and high compression makes the jpeg look bad if you push it too far compared to the much less compressed, higher bit depth raw.

  2. Very interesting. This is something that has always plagued my brain. I am not so much caught up in say the ProRes Raw upgrades that are coming to a majority of smaller cinema and dslr style cameras, I have always been more interested in the 12-bit color space that comes with it. Although your last comment about 10-bit raw vs log confuses me. But alas, this was still very informative. Your knowledge is always super beneficial. Thanks!

    1. Colourspace and bit depth are two different things and independent of each other. A colourspace is just a description of a colour range and how the colours in that range are stored, organised or reproduced. The same colour space can exist at almost any reasonable bit depth, for example S-gamut is the same colour space whether the file is 8 bit or 16 bit. Increasing bit depth does not normally give you more range or a bigger colourspace. Increased bit depth gives you greater tonal resolution.

      One of the most critical processes in a bayer camera is the conversion of the brightness only bitmap (the so called “raw data”) into a full colour image. You have to take the brightness information captured for each filtered colour and apply a lot of complex mathematics including things such as edge detection, colour estimation to create a colour image. I think we all know that whenever you do any heavy duty image processing that oversampling or starting off with as much original information as possible will produce the best results. It’s long been known by camera manufacturers that if you want a really good image you need to do your image processing at a bit depth greater than the bit depth of the output. Most high end video cameras take the data from the sensor at a minimum of 12 bit, if not 14 or even 16 bit. Then the camera uses an image processor operating at a similar or even higher bit depth to create the colour image. What you must remember with raw is that what we record is not a traditional colour image, but a record of the intensity of the different colours in the scene. If that raw image is only 10 bit and perhaps contains some compression artefacts before it’s even been turned into a colour image this is not good. The 10 bit colour image from the camera will probably be of better quality because it was derived from a higher quality original source. This is why until recently all professional raw cameras had at the very least a 12 bit log output if not 16 bit. It has long been accepted that you need at least 12 bit raw data to match 10 bit component log and the 12 bit raw must be free of any significant compression artefacts.

      Now that raw video has become the latest must have camera function everyone and their dog is adding raw video to cameras. But because it is actually quite difficult to record good quality 12 bit log raw or 16 bit linear raw in a small enough file for low cost, compact and reliable media, they are adding much simpler and easier formats such as 10 bit log or 12 bit linear. Just so they can put a raw sticker on the box.

      Everyone knows that there is a big difference between an 8 bit codec and a 10 bit one. We also know that highly compressed codecs don’t normally look as good as more lightly compressed codecs. Well these things apply to raw too. I would liken 10 bit raw to an 8 bit conventional recording. It works, used carefully you can get a good final image, but it’s not something you want to push about a lot, it’s a compromise and it will never be the best. When we get to 12 bit log it’s comparable to a 10 bit conventional recording. It should work well, should withstand a moderate amount of pushing and pulling in post. But I think what everyone has in their minds when they are thinking raw is really 16 bit raw. There used to be a saying regarding raw – to determine how many “bits” you need, take the dynamic range you wish to capture then add 2 and that’s the number of “bits” you need. So for a 14 stop camera, you want 16 bit raw. But in the race to be able to say “I’ve got raw” we suddenly now have a raft of 10 bit and 12 bit highly compressed raw formats that often offer no real benefit over a good quality 10 bit component log recording.

  3. I think you are incorrect about WB. This may be the case with SONY RAW because they “cheat” and use post processed white balance as a method to keep the final file smaller. ie Not true RAW. Have a look at other RAW codecs and they don’t do white balance in camera – they just tell you what it was as shot. WB is not applied until JPEG stage. The sensor doesn’t physically adjust it’s colour sensitivity with WB, it’s a post processing function. And if you have uncompressed RAW off the sensor it’s no different to doing it in camera.
    What can happen is the metering of the camera can be upset by incorrect WB when shooting.

    1. With video cameras because they almost all provide a simultaneous ability to output a high quality image via SDI, HDMI or for internal conventional recording, to avoid excess noise in daylight and/or reduced dynamic range under tungsten lights the R and B gain is adjusted in camera to maintain constant high performance under a wide range of lighting conditions.

      There is a very big difference between the effect changing the gain of a cameras A to D converters and changing gain anywhere else has. For example changing gain prior to recording alters the dynamic range that can be captured. Changing the gain of the A to D’s changes both DR and noise performance in ways very different to any changes made once the pixel output has been digitised. If you want the best DR and least noise, under daylight you want to keep the blue gain low and not use more R gain than you absolutely need. Under Tungsten lighting you need to raise the blue gain but lower the red to avoid a reduction in DR in the red channel. If you want the best possible performance across a broad range of colour temperatures you need to do these adjustments as early in the camera chain as possible. Once you have recorded the sensor output you can’t then alter the DR, if your reds are clipped because you had too much red gain you can’t ever recover that.

      It is very common to adjust the A to D gain to maintain the best dynamic range and noise performance under different colour temperatures.

      But – importantly, however you do it once the signal has been recorded in a file that does not have unlimited range or unlimited resolution, you have a fixed, single white balance determined by however the camera was setup to respond to light. Whether that’s a stills camera or a video camera, what you record has to have been optimised for just one single white balance, exactly as it can also only ever be optimised for one sensitivity. So anything you do in post is simply a correction to what was recorded and is not the same thing as an in camera white balance adjustment (unless it is a very crude camera). This is absolutely no different to to making a colour correction to log. If you were to record using component or RGB with the same gamma, bit depth and compression ratio as raw you would be able to correct and grade it in exactly the same way as you can raw. Raw isn’t magic, it’s just efficient.

  4. Alister, I agree with several others that your info about white balance is incorrect. I’m sure there is an ideal white balance for a sensor that will produce the best results/ dynamic range, but trying to adjust white balance in any gamma encoded video in post is not the same thing as doing it with Raw material. With log footage if you only use gain to adjust white balance you will never get the shadows, mids and highlights all balanced. If you get the highlights balanced the mids will be off. If you get mids balanced shadows will be off., etc. I have done extensive tests with this using Sony cameras. You need to use several controls like lift, gamma, gain, or offset as well as more focused controls (such as log shadows and highlights in Resolve) to even get reasonable results. With Raw material, the temp and tint controls achieve perfect white balance in post throughout the entire tonal range – identical to white balancing in camera. Exposure can be mimicked pretty accurately on log footage with a single control (offset), but white balance can not. You can see this for yourself by shooting a test card.

    1. Gamma is reversible. It is simply a mathematical formula applied to the footage and as a result is totally reversible. Provided you know the formula, with the right tools you can easily mimic the way raw behaves in post. Take S-Log material into ACES and use the colour temp slider and you can push and pull the white balance in just the same way as raw because ACES knows the S-Log formula and takes the log back to linear. Most colour managed workflows will allow you to do this.

      And don’t forget – many raw formats also have gamma. 10 and 12 bit log raw have a log gamma curve applied. So to do any changes to this you must be able to apply a transform to it or you will have the same issues.

      It’s not that you can’t adjust log as well as raw, it’s simply that people either don’t know how to do it properly or are using the wrong tools.

      White balancing raw in post can never be the same as white balancing in camera as you don’t have access to the sensors A to D converters. This is easy to show. Shoot the same test chart under tungsten and daylight. White balance correctly for both in camera. Then do the same but shoot the Tungsten chart using a daylight balance and the daylight chart using tungsten. Then take this into post and correct back to the right colour temperature. Then observe the noise. The noise will be much greater in the blue channel of the tungsten lit chart. Then there is also the dynamic range issue. When you white balance in camera you are adjusting your gain prior to recording and this will give you the optimum DR. Shoot with the wrong in camera white balance and the DR is not optimised and nothing you do in post can ever change it.

  5. Great article.
    There is however a noticeable difference in WB adjustment for RAW and debayered video.
    In Assimilate I did a test with ProRes Raw. I shot 3 images with a Varicam LT, outputting 10-bit log RAW. One file is at a correct in-camera WB of 5140, the second at 15000, the third at 2000. So those two are way off.
    Yet it is fairly easy to get those two images to a very acceptable look with the WB and Tint sliders. Not everything matches up identically with those simple adjustments. But with more tweaking, I can get them to match.
    If I try those adjustments with the same shots converted to 10-bit log RGB it becomes very hard to get an acceptable result.
    So even at that lower bitrate the RAW gives more wiggle room than the debayered file.

    1. There is no reason why that should be. Whatever you shoot does not have “every” white balance, it can only have one white balance as you only record one set of data, so it all depends on how the sensor was set. The only reason for the difference will be down to workflow, not the material itself.

Leave a Reply

Your email address will not be published. Required fields are marked *

*

This site uses Akismet to reduce spam. Learn how your comment data is processed.