Lightroom: Purple fringing in bottom RH corner of images from Canon 5D Mk3, could it be the Lightroom import algorithm?

  • 1
  • Problem
  • Updated 4 years ago
  • (Edited)
This is an odd scenario, but here goes....

I have been shooting a lot of “low light” images on my Canon 5D mk3. After importing them into Lightroom 5.3 I noticed a purple fringing in the bottom right hand corner of images with an ISO of 12800 or more. It happens in all sorts of lighting and compositions, so my initial reaction is that it is a camera issue. (I did ensure that lens correction utility in Lightroom was off too, but that didn't change much) (See screen shot attached for a typical example)

I have taken this up with Canon Canada saying I think it’s a sensor issue (There are other examples of a similar issue on various photography forums online). They asked me to supply the original “straight out of camera” RAW files, and said that in their software (Canon digital Photo Professional which comes bundled with Canon Cameras) the images are fine, and after checking they do indeed look fine in their software!

So, is Canon correct? And if so is there some sort of algorithm problem in Lightroom when it imports higher ISO Raw images that Adobe needs to correct? It seems odd though that it’s just in one corner time after time, because to me that suggests a camera sensor issue.

Is there a way to definitely test if the issue is “baked” into the RAW file, or if it’s being created on import to Lightroom? (I do not have Photoshop BTW)

I will gladly supply the sample RAW files to Adobe for test. here is a link to a dropbox of sample images direct from camera, that produce the error when viewed in Lightroom: https://www.dropbox.com/sh/lkd2muinfa...

Thanks in advance for your help in getting to the bottom of this issue.
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
  • confused, frustrated,

Posted 5 years ago

  • 1
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
The purple in the corner is there in the raw file. It is called Amp Glow or thermal noise, not purple fringing, which is a type chromatic aberration where purple is visible along high contrast edges of a photo, typically tree branches against the sky.

Amp Glow along a corner or edge of a sensor is caused by the heat of another component adjacent to the sensor becoming hot and that heat adding thermal noise to the photosites near it. Photosites in the camera sensor convert photons to electrons, and heat also causes extra electrons to travel into a photosite, where that heat is counted as light, instead, when the sensor is read out at the end of the exposure.

The reason it is purple is because there are twice as many green photosites as either red or blue photosites and so the random thermal noise in the green signal is averaged out compared to the red and blue signal.

Other raw converters show purple, too:


The reason the purple is much more obvious in LR than in the Camera JPGs is that LR is brightening the corners as part of its vignetting corrections from the lens profile.

You can just turn down the vignetting slider to undo some or all of the vignetting correction on these photos where the purple thermal noise is a problem.

Here is a side-by-side that shows the difference between no vignetting correction and 100% vignetting correction:
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
Hi Steve
Many thanks for looking into this so thoroughly.

Can I ask:
Which Raw Converters did you try? I will go back to Canon on this, and it would be really helpful to know what steps you took to investigate please.

Would you consider this a flaw in the camera design then (as you mention it might be from heat from other components near that corner of the sensor)

Do you think this is something that changes as a camera "heats up" after being used for long periods?

Thanks again.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
Besides LR, I only used raw-digger which is a raw analysis program that doesn't do anything special to the data.

There is a faint purple corner in the camera embedded jpg preview in the raw file, so I have no doubt this is something the camera is doing and the vignetting correction is enhancing the visibility due to lightening.

If you want to investigate for yourself the camera origins of this purple corner, set the highest ISO the camera has, set the camera to RAW+JPG, turn off any noise-reduction, especially long-exposure noise-reduction, put a lens cap on the lens or a body cap on the camera body and put in a dark place, then take a 30-second exposure in complete darkness. You should be able to see quite a bit of the purple glow in the camera JPG.

The higher the air temperature the worse it'll be. The higher the internal temperature the worse it'll be. The higher the ISO the worse it'll be, because higher ISO is just magnifying the signal from the sensor so any noise is magnified along with the data. If you have used the camera in a hot environment and for an extended period the internal temperature will be higher and the amount of thermal noise will be higher.

Was the gym hot and/or had you used your camera quite a bit beforehand that would cause the electronics to get hot due to frequency of use?

Is this a new camera to you or have you taken pictures at a similar ISO in the past and cannot detect any purple corner? It is possible something is overheating differently than before.

If you want to contact Canon, make sure you can see the purple with the corner when using Canon's own software, Digital Photo Professional, DPP, which should be on a disc you received with the camera.
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
Back to Canon next. Thanks again.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
It is possible that this amount of thermal noise is expected, or it is possible that the camera can be recalibrated to somewhat correct the amount of noise, or that some component of the camera is failing and overheating more than expected.

Perhaps another 5D.3 owner can share their experience with shots similar to where you're seeing the purple in the corner.
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
I did the test you suggested above:
Raw + JPEG in camera, High ISO, lens cap on, 30 second exposure, NR off, and here are the results of each image:

JPEG shot


RAW shot


They did look like this on the back of the camera too, so at least we know it is being created in camera as suspected.

If there are any other 5D mk3 users out there, it would be great to get your experiences on this issue too.
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
Hi
I wanted to share with the community the latest reply from Canon Canada (and by proxy Canon USA) about the amp glow issue. I am really quite staggered by their writhing and deceit over this issue. It really does make you question their honesty when they so blatantly twist the situation (and my words!) like this to protect themselves. Enjoy!

Email dated 31st March 2014 from Canon Canada:

There is no known issue with the 5D Mark III when shooting at ISO 12800, however there is an explanation for the behaviour that you have been observing. From the information that you have provided, you had indicated that you had shot images at ISO 12800 and had subsequently viewed those images within a third party software and that it was within this software that you noticed some noise at the bottom right hand corner of the image. You had also indicated that the phenomenon with the images shot at ISO 12800 did not appear when viewing the images within Canon’s Digital Photo Professional software (DPP). We have confirmed on our end that there is no issue present within DPP when viewing the ISO 12800 image that you provided. This difference suggests that your third party application may be displaying the image differently somehow. To investigate this issue further we recommend contacting your software vendor.

The phenomenon that you have observed and your particular case details have been discussed with the appropriate departments and management at Canon and it was determined that the unit is performing correctly as per the product specifications and manual with regards to the phenomenon you have reported.

Please be assured that Canon strives to market our products in the most ethical and professional business manner possible, and we sympathize for any inconvenience this situation may have caused you. The fact that you took the time to write to us is indeed appreciated. It is only through our customer's comments that we are able to provide quality products and service.

ENDS

They just won't even acknowledge the cold hard facts of the situation; that they have built a camera with a fundamental design flaw in it! It's not so much that the error exists as their lame attempts to deny it that are just so intolerable! I have trusted Canon and now they have broken that trust without a hint of remorse.

I welcome your comments on this thread.
Photo of Chris Cox

Chris Cox

  • 20280 Posts
  • 813 Reply Likes
Yeah, that looks like amplifier/heat noise -- you see this pretty commonly in astrophotography (LONG exposure) images. The normal solution for that is to take several similar exposures with the lens cap on, average those, then use that as a black field sample and subtract it from the raw images before doing additional processing.

White field/vignette correction will amplify the noise and make it more visible.

And yes, that Canon response is rather discouraging. It appears that someone failed to investigate the issue correctly.
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
Honestly I think they know about it, and are not willing to formally acknowledge the issue for fear of opening up calls for a swap out / recall. They have been so evasive both on the phone and in emails, carefully avoiding any responsibility. I've been a loyal purchaser of lots of high-end Canon gear, but this kind of thing has really opened my eyes to how they will avoid taking responsibility and just go all "corporate" when challenged. Shameful.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2645 Posts
  • 337 Reply Likes
In my opinion, what you're seeing is normal, and commonly happens with long exposures. It is not a bad-camera or something for Canon to fix, it just happens with long exposures, especially if you don't cool your camera.

Astrophotographers go to elaborate lengths to mitigate this sort of issue by doing, as Chris mentions, averaging several dark-frames together and then subtracting those from the light frames--the astrophotographs, to subtract out any camera-added variations.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2645 Posts
  • 337 Reply Likes
Obviously turning on long-exposure noise-reduction will do this same dark-frame subtraction, so Canon does anticipate and have a solution for this particular issue if you're shooting JPGs, or do raws also have this subtraction done to them, too, on your model of camera?
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
Hi Steve, but it doesn't just happen with long exposures; it happens in regular shots when the ISO is say 12800 or more, and given the camera is promoted to shoot as ISOs up to 102400 that an issue. Also if it was generalised "noise" I'd get that but this is localised to one area.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
What did they say about the JPG showing purple, and there being a little more along the bottom? Obviously the JPG was developed in camera so they can't blame any third-party software.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
Obviously turning on long-exposure noise-reduction will do this same dark-frame subtraction, so Canon does anticipate and have a solution for this particular issue if you're shooting JPGs, or do raws also have this subtraction done to them, too, on your model of camera?
Photo of Edward

Edward

  • 194 Posts
  • 4 Reply Likes
I've sent them everything including the RAW & JPEGs direct from the CF cards, but they still insist its in whatever software I am using! I have also offered to send them a photograph of the back of the camera when the image is produced but they said no need! The y just hold to the BS about it being the RAW engine in Lightroom and/or the Rawdigger software and/or A.N.Other software. Ridiculous.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
If they saw the JPG then it doesn't matter what any other software did to the raw, unless they thought, maybe, the JPG had been tampered with by other software despite anything you might have said to the contrary.
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
Canon cameras do record how hot the camera was when each picture was taken. You may have to use something like EXIFtool, or maybe http://regex.info/exif.cgi to view that sort of info.

How hot was the original boxing picture and how hot was the dark frame?
Photo of Steve Sprengel

Steve Sprengel, Champion

  • 2647 Posts
  • 337 Reply Likes
To answer my own question, using the regex site, for the dark frame, the camera was 24C = 75F which is a normal sort of temperature.