Firmware v1.2.0 seems to have solved a banding issue that was present in the Canon EOS R

[jeffa4444]

I have not mentioned specifically Canon and I think that the modern sensors of all companies are more or less comparable in usable DR. One stop does not help in critical situations therefore I mentioned the 16-20 stops. Which no camera company delivers for the standard consumer/pro photography market as far as I know.

No company produces a camera with 16-20 stops period including very expensive high end movie cameras (dont believe all the marketing hype).

Latest research suggests certain humans can see up to 20 stops of DR when our eyes adjust but not at a single point of time (we have been helping a researcher on this for over ten years in Scotland).
Our eyes can adapt, a sensor cannot.

The best high end camera we have tested is claimed at 16 stops but actually is more like 15. No system of checking is a 100% guaranteed either because small external factors can affect the results and at high sensitivity (including heat).

Manufacturers over egg real world DR and always have.

 

[mb66energy]

Just after some thinking of how a firmware update can resolve banding issues: How does it work? Replacing the sensor by a FW update seems impossible.

My first substantial idea is that this is a timing issue of two digital high frequency signals at slightly different frequencies - one of the ADC, another from an other source which interfere. This might be solved by programming - if you synchronize both the interference is cancelled out and the resulting signal (here the stream of ADC values) is clean.

 

[mb66energy]

No company produces a camera with 16-20 stops period including very expensive high end movie cameras (dont believe all the marketing hype).

Latest research suggests certain humans can see up to 20 stops of DR when our eyes adjust but not at a single point of time (we have been helping a researcher on this for over ten years in Scotland).
Our eyes can adapt, a sensor cannot.

The best high end camera we have tested is claimed at 16 stops but actually is more like 15. No system of checking is a 100% guaranteed either because small external factors can affect the results and at high sensitivity (including heat).

Manufacturers over egg real world DR and always have.

Thanks for confirmation about camera DR availability for non-prototypes.

I derived the miniumum of 20 stops for the human from the fact that you have 1 000 W per m² on a bright sunlit day at noon and that a 70mW (electrical input) LED (10mW light output @ 15% efficiency) is bright enough to illuminate a room wall (roughly 10 m²) resulting in 0,001 W per m² and just there is maybe three or 4 stops of DR ...

And yes, manufacturers push their spec tables to the best case scenarios (like gas consumption of cars which is 20-30% higher than the data in their spec sheets.) So I would like to have a 18 bit DR in the spec sheet to get maybe 16 stops in real use.

 

[Viggo]

This made my day! I’m so glad they listened to us I haven’t edited any shots since I updated, but did today and can confirm that the banding is gone

 

[Wim Hendriks]

also the up-date for the canon RP

 

[Kit.]

Just after some thinking of how a firmware update can resolve banding issues: How does it work? Replacing the sensor by a FW update seems impossible.

There are probably several "dark" (unused) pixels at the beginning and the end of each line. Their averaged values can give an offset to subtract from the values in the line to reduce banding.

 

[mb66energy]

There are probably several "dark" (unused) pixels at the beginning and the end of each line. Their averaged values can give an offset to subtract from the values in the line to reduce banding.

I always wanted to know what the purpose for these non-net pixels is: Thanks for your idea!

Maybe they have found a better way to implement the offset correction - it's not always easy to understand simple systems and modern cameras are lightyears from simple with these different sensors, actors, electronics and its firmware which work together in a delicate way (or sometimes not).

 

[QuisUtDeus]

Are you too one of those who does NOT like these video only things where you need a lot of bandwith to get some bits of information? - IMO text rules But maybe I am too old for this modern "life".

If someone won't take the time to write down the information they have, I won't take the time to listen to their blather. Video is a plague and a crutch.

And I'm a millennial, for what it's worth.

 

[flip314]

No company produces a camera with 16-20 stops period including very expensive high end movie cameras (dont believe all the marketing hype).

But what if I SPEAK REALLY LOUDLY about a camera from a company that I can't name that will soon release a camera that will BLOW THE LID off of anything you've ever seen. Really, this will make everything you've ever heard about obsolete. You will get 16K video at 3000fps. 300,000 megapixel stills pulled out of video. 50 stops of DR and 80b color depth.

I have a prototype sitting on my desk RIGHT NOW, and my employer has kindly allowed me to post on internet forums about it, but this is all they will allow me to say. I can't tell you which company is producing it or when it will be available, only that it will BLOW YOUR MIND. The entire thing is in a 1" cube and weighs just grams. It will FUNDAMENTALLY change videography and photography forever and will be here SOON.

But I've already said too much. You will just have to wait and see, and then you will feel stupid for doubting me.

 

[mb66energy]

But what if I SPEAK REALLY LOUDLY about a camera from a company that I can't name that will soon release a camera that will BLOW THE LID off of anything you've ever seen. Really, this will make everything you've ever heard about obsolete. You will get 16K video at 3000fps. 300,000 megapixel stills pulled out of video. 50 stops of DR and 80b color depth.

I have a prototype sitting on my desk RIGHT NOW, [...]

Hardly believable. 50 stops of DR is roughly 1 000 000 000 000 000 photoelectrons full well capacity. EOS 5D has 70 000 x 13 000 000 total of full well capacity or ~ 1 000 000 000 so one pixel of your camera has to be ~ 1 000 000 times larger than the full full frame area resulting in 36 x 24 meters. No, I don't believe that it sits on your desk, no desk is large and stable enough.

Reading of a fully saturated chip at 3000 fps @ 16k (132 000 000 pixels x 3000 frames x 1 000 000 000 000 000 electrons ) results in a current of 70 000 000 Ampere ... needs its own small powerplant.

 

[neuroanatomist]

Hardly believable. 50 stops of DR is roughly 1 000 000 000 000 000 photoelectrons full well capacity. EOS 5D has 70 000 x 13 000 000 total of full well capacity or ~ 1 000 000 000 so one pixel of your camera has to be ~ 1 000 000 times larger than the full full frame area resulting in 36 x 24 meters. No, I don't believe that it sits on your desk, no desk is large and stable enough.

Reading of a fully saturated chip at 3000 fps @ 16k (132 000 000 pixels x 3000 frames x 1 000 000 000 000 000 electrons ) results in a current of 70 000 000 Ampere ... needs its own small powerplant.

@HarryFilm is all the explanation necessary.

 

[lawny13]

It's a very good shot but I think the necessary DR isn't that vast because the gray surface has around 20% reflection (looks like a Kodak gray card) and the sun is blown out. Do you have used a mask to make the person brighter?

The sun in the frame in a deep forest is impossible right now with available photographic cameras so I would like to have 16 stops or more but I know very well that it will not always help if you try to compress that dynamics into a 6 or 8 stop wide print.

The human eye has way more DR, and yet in a scene like that the sun would be blown out to our eyes too.

So an image where the sun was not blown out would essentially look unrealistic.

 

[deleteme]

I would prefer to increase the full well capacity to 1 000 000 photoelectrons - I think this is the way to have data for all light levels. And it also helps to get a more precise image due to lower statistical errors. As physicist I always like to have the best data I can get.

The challenge with gathering that much data is that the cost-benefit ratio stops working at a point well short of what you may think will be useful.

 

[Policar]

Hoping this fix gets ported to the C300 Mk II etc as someone mentioned earlier... I have seen this banding problem show up in "real world" circumstances with Canon's cinema cameras, not just Xyla charts (where it's quite obviously worse than what other manufacturers have). It's been a known issue they've been intending to address for a while and I'm disappointed that it persists.

Everyone has their own needs, which don't necessarily correlate with their ability, but I do think it's presumptuous to assume others share your needs and wants. I'd be very happy with an EOS-R for landscapes and don't need anyone telling me I shouldn't be. Then again, my favorite landscape photographers are mostly shooting Velvia, which has maybe six stops of dynamic range. But you can't really do proper black and white zone system without a full 10+ stops, so I acknowledge the need for more.

These jokes about 16-20 stops of DR are just that for now, but I don't know if they will be for long. High end prints I believe have at most five stops of contrast, and typically maybe four, and even backlit LCD monitors often have 1000:1 contrast ratios, about ten stops. Fitting too much capture dynamic range into that space, even with dodging and burning/tone mapping/power windows will start to feel artificial–or just very very very flat, but 10,000 nit and even 4,000 nit micro-LED screens can comfortably display 15-17 stops in one scene. It's a lot like looking out a window. Glowing reports from CES, imo, are still underselling the technology.

Such displays are a ways away from the consumer space, but when they're commonplace, so too will be a need for greater dynamic range in capture. The C300 Mk II was designed with HDR video in mind, which is another reason it's annoying to me that there's such severe CMOS smear in that sensor.

 

[LSXPhotog]

I can confirm that this 100% resolved the issue on my camera. I almost feel like it has added much more usable dynamic range and I'm honestly impressed by the result.

 

[jolyonralph]

He answered this @ 2:06 in his video: He updated the camera with the older FW to the current version and the banding issue was resolved.

Are you too one of those who does NOT like these video only things where you need a lot of bandwith to get some bits of information? - IMO text rules But maybe I am too old for this modern "life".

Bandwidth is plentiful, time isn't!

 

[HarryFilm]

Hardly believable. 50 stops of DR is roughly 1 000 000 000 000 000 photoelectrons full well capacity. EOS 5D has 70 000 x 13 000 000 total of full well capacity or ~ 1 000 000 000 so one pixel of your camera has to be ~ 1 000 000 times larger than the full full frame area resulting in 36 x 24 meters. No, I don't believe that it sits on your desk, no desk is large and stable enough.

Reading of a fully saturated chip at 3000 fps @ 16k (132 000 000 pixels x 3000 frames x 1 000 000 000 000 000 electrons ) results in a current of 70 000 000 Ampere ... needs its own small powerplant.

===

Me thinks you took my camera body sarcasm just a tad too much to heart -- THAT post truly WAS SARCASM!!!
(be it at it's best or it's worst!) trying to illustrate the almost pointless nature of spec-wars when even a $100 smartphone is good enough these days for most photography!

...BUT... I will still note that a 50 megapixel Medium Format Sensor combined stills/video monster IS COMING .... rather soon!
.
from who? ... Well... I'm pretty sure you will probably be quite shocked at who that is!
.

 

[SecureGSM]

...BUT... I will still note that a 50 megapixel Medium Format Sensor combined stills/video monster IS COMING .... rather soon!

Put a bridle on thy tongue; set a guard before thy lips, lest the words of thine own mouth destroy thy peace... on much speaking cometh repentance, but in silence is safety.

William Drummond